Which is most important
EFF
0%
0% [ 0 ]
PER
100%
100% [ 16 ]
Total Votes : 16
Author Message
Chronz1
Joined: 22 May 2006
Posts: 201
PostPosted: Wed May 16, 2007 4:39 am Post subject: PER vs EFF Reply with quote
Which is the most credible measurement of statistical achievements and why?
Back to top
View user's profile Send private message
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Wed May 16, 2007 8:19 am Post subject: Reply with quote
Quote:
Which is the most credible measurement of statistical achievements and why?
All are credibles. Among the highly "economics yield minded" WProduced, The offensive biased PER, and the "supposed" highly subjective ON/OFF; If I were a GM, I would try WP. But, If I were a coach, then the last two would be more useful for me.
Back to top
View user's profile Send private message
holymoly
Joined: 30 Jul 2005
Posts: 63
PostPosted: Thu May 17, 2007 9:04 am Post subject: Reply with quote
I think the more interesting comparison is between ORtg/DRtg and PER. Their are many players with low ORtg and a negative net Rtg but a high PER (usually for players who use a high % of poss). Which one do you look at?
_________________
.....
60% of the time, it works every time
Back to top
View user's profile Send private message Send e-mail MSN Messenger
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Thu May 17, 2007 9:59 am Post subject: Reply with quote
NBA Efficiency? Lol.
It weighs field goal attempts and ft attempts equally, which is kind of strange in my book.
Berri has an interesting post on it here
http://dberri.wordpress.com/2006/11/09/ ... -rebounds/
The basic takeaway is that Efficiency drastically overvalues scoring.
Back to top
View user's profile Send private message
Mike G
Joined: 14 Jan 2005
Posts: 3605
Location: Hendersonville, NC
PostPosted: Thu May 17, 2007 11:20 am Post subject: Reply with quote
Flint: LOL, indeed. The Babbler stats indicate that among the top 30 NBA players (in PAWS/Min) are Esteban Batista, Jeff Foster, Chuck Hayes, Amir Johnson, Jamal Sampson, Andris Biedrins, Renaldo Balkman, Kyle Lowry, Reggie Evans, Tyson Chandler, and Dikembe Mutombo. At least 1/3 of the league's "best" do not even start for their teams.
In PER rankings, at 100 minutes minimum, only Amir stands out (at #31) as an oddity. He does not rank in my top 100. He's at #19 in NBA.com's EFF/48M, where Sean May is #24, Biedrins #31, Cardinal #42, Lowry #49.
I don't see how any of these names can enter into a 'top 50 players' list; Chandler does make my top 100. But among any known measures, the WOW stuff really strains credibility.
Why does Berri credit FG and subtract 'unmade' FGA? Why not subtract 'ungotten' rebounds? Please don't ask me to read his blogs; he's quite unreadable to me. If you understand what he's saying, please translate.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Thu May 17, 2007 1:22 pm Post subject: Reply with quote
Mike -
Batista, Johnson, Sampson, and Lowry barely played this year, so its sort of silly to bring them up.
Biedrins, Hayes, Mutombo, Evans, Foster, and Balkman are all excellent players. I dont know who is on your top 100, but at least four of the above guys should be in there. None are scorers, but all have a profound influence on games through their impact on possession. Mutombo, for instance, proved his value this year when Yao went down. The Rockets posted a higher winning percentage with him than with Yao before his injury. Anecdotally, this tells me that perhaps a system that doesn't include him in the TOP 100 players is more flawed than Berri's.
Vis a vis NBA Efficiency, Berri has a perfectly valid point. A made FG produces two or three points. A made ft produces one. It's ridiculous to weight them equally in a formula.
The reason Berri subtracts for a missed fg is because regression analysis suggests that a missed fg on the team level has the same impact on winning as a point scored. A missed shot and a made 2 pt field goal have the same impact on Wins therefore.But it's also just common sense. When you take a shot you expend a resource, a possession. The cost has to be subtracted from the benefit of points scored. To do it any other way is kind of cock-eyed. NBA Efficiency used in a casino would suggest that if you risked 100 in blackjack, and won qnother 100, that your profit amounted to something more than 100 dollars. You only need to shoot 33% to show a profit on your scoring with NBA Efficiency, which is why players like antoine walker come out so well, despite not being very good.
How did you arrive at the weights for your formula?
Back to top
View user's profile Send private message
Chronz1
Joined: 22 May 2006
Posts: 201
PostPosted: Thu May 17, 2007 6:21 pm Post subject: Reply with quote
Reason I ask is because Im having a NEVERENDING debate on why Dirk has better numbers than KG. EFF is his biggest supporter, I tried endlessly to tell him EFF doesnt take into account important factors like pace or minutes played but he insists the pace of the game doesnt affect star players only the role players.
And when I use EFF/48 and Dirk comes out on top he backpedals and says KG has better numbers because the only thing Dirk has him beet on is scoring/efficient. As if its not the most important aspect and that the rest of the numbers compensate for that.
Back to top
View user's profile Send private message
kjb
Joined: 03 Jan 2005
Posts: 865
Location: Washington, DC
PostPosted: Fri May 18, 2007 9:47 am Post subject: Reply with quote
Berri's approach -- subtracting ALL field goal attempts -- seems strange to me. A made shot is a postive thing, but in Berri's system, a made 2pt FG is +1 (2 points - 1 FGA). NBA efficiency, and other systems, make more intuitive sense because a guy doesn't get penalized for doing something his team has to do that trip down the floor anyway -- a 2pt FG is 2 points. He gets "demerits" only for "bad stuff".
There's another thought -- maybe shooting a low percentage can be a net benefit because of offensive rebounding. Last season, roughly a quarter of missed field goal attempts went to the offense -- often to a guy in a good position to get a high percentage follow-up opportunity.
In Berri's system -- looking ONLY at scoring -- a guy who shoots 50-100 contributes nothing (100 pts - 100 FGA = 0).
Subtracting only for misses suggest the guy contributes 50 credits (100 points - 50 misses = 50 credits).
If we consider the value of offensive rebounding, on average out of 50 missed shots, 12-13 will be rebounded go back to the offense. Looked at that way, the player shooting 50-100 gets 62.5 "credits".
Or, if we look at it using Dan Rosebaum's tweak -- subtracting 1 for every made FG, and .75 for every missed FG (25% going back to the offense) -- the player gets 12.5 credits.
Back to top
View user's profile Send private message AIM Address Yahoo Messenger
Mike G
Joined: 14 Jan 2005
Posts: 3605
Location: Hendersonville, NC
PostPosted: Fri May 18, 2007 10:27 am Post subject: Reply with quote
kjb wrote:
In Berri's system -- looking ONLY at scoring -- a guy who shoots 50-100 contributes nothing (100 pts - 100 FGA = 0).
If true, that's pretty amazing. My question was :
Quote:
...Why not subtract 'ungotten' rebounds?
This doesn't seem like any more absurd a suggestion. In fact, rebounds are 'automatic', resulting from missed shots. A rebound goes to one team or the other, depending on effort, etc. Points, however, result from the effort required to overcome defenses. They aren't automatically going to happen to one team or to either.
An average rebounder should get 10% of the rebounds while he's on the court. So should anything less than a RebRt of 10.0 be counted as negative? Should a RebRt of 15 be counted as 5 ?
The same argument -- that says missing more than 1/2 your shots makes your scoring contribution negative -- would seem to say getting fewer rebounds than your counterpart makes you a negative rebounder. A team that gets outrebounded should total negative 'rebounds added', or whatever you may call it.
Over-valuing rebounds does not fix the problem of over-valued scoring.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Fri May 18, 2007 11:06 am Post subject: Reply with quote
KJB - Here is the relevant excerpt from the book on this topic. http://www.wagesofwins.com/WOWCh7.htm
I dont understand your post. Especially:
NBA efficiency, and other systems, make more intuitive sense because a guy doesn't get penalized for doing something his team has to do that trip down the floor anyway -- a 2pt FG is 2 points. He gets "demerits" only for "bad stuff".
Look Berri's approach is really quite simple. He did a regression analysis and found that the value of a shot attempt, whether made or missed, and a point scored, are the same in terms of their impact on team wins. And he designed his formula to reflect these weights.
This makes perfect sense to me, I dont know why its such a stumbling block for many people.
Back to top
View user's profile Send private message
Mike G
Joined: 14 Jan 2005
Posts: 3605
Location: Hendersonville, NC
PostPosted: Fri May 18, 2007 11:14 am Post subject: Reply with quote
In other words, because Berri said so?
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Fri May 18, 2007 11:16 am Post subject: Reply with quote
Quote:
kjb wrote:
In Berri's system -- looking ONLY at scoring -- a guy who shoots 50-100 contributes nothing (100 pts - 100 FGA = 0).
If true, that's pretty amazing.
This isn't really true. It's true to some extent for a player who only shoots two point field goals, say Eddy Curry. But you can't leave free throws out of the picture, they are an extremely crucial part of overall shooting efficiency.
And Winscore is just a quick and dirty metric. In the final analysis, with Wins Produced, you show a "profit" on your scoring if your TS% is above the average for your position.
Back to top
View user's profile Send private message
kjb
Joined: 03 Jan 2005
Posts: 865
Location: Washington, DC
PostPosted: Fri May 18, 2007 11:19 am Post subject: Reply with quote
I've read the excerpt, and I read his blog daily. Berri's explanation does sound simple, but what "resource" is being used when a guy makes a shot?
By the way -- I'm not necessarily opposed to this viewpoint. I'd like to understand the thinking behind it.
Back to top
View user's profile Send private message AIM Address Yahoo Messenger
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Fri May 18, 2007 11:21 am Post subject: Reply with quote
Mike - How did you arrive at the weights in your formula? Berri used regression analysis to determine what the impact of a rebound, steal, turnover, shot attempt etc were at the team level in terms of wins. How did you make your decision? Do you link your formula to actual team wins?
And does is make sense to you that a shot attempt and a ft attempt are weighted equally by the NBA Efficiency formula? Are you really going to defend that particular metric. Isn't it obvious to you that scoring must be overvalued by an approach which implies that that a shot attempt has one value if it goes in and a different value if it doesn't?
Back to top
View user's profile Send private message
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Fri May 18, 2007 11:22 am Post subject: Reply with quote
KJB - The resource that is being expended is the teams possession of the ball.
kjb
Joined: 03 Jan 2005
Posts: 864
Location: Washington, DC
PostPosted: Fri May 18, 2007 12:06 pm Post subject: Reply with quote
Flint wrote:
KJB - The resource that is being expended is the teams possession of the ball.
Interesting, and it makes some sense now. Wouldn't it make sense then to include something for the effect of offensive rebounding? Because approximately 25-30% of the time, a missed shot is not a spent possession -- it becomes an offensive rebound, which preserves possession of the ball. Basically, to incorporate Dan Rosenbaum's tweak to Berri's system.
Back to top
View user's profile Send private message AIM Address Yahoo Messenger
Mike G
Joined: 14 Jan 2005
Posts: 3559
Location: Hendersonville, NC
PostPosted: Fri May 18, 2007 1:39 pm Post subject: Reply with quote
Flint wrote:
Mike - How did you arrive at the weights in your formula? Berri used regression analysis to determine what the impact of a rebound, steal, turnover, shot attempt etc were at the team level in terms of wins. How did you make your decision? Do you link your formula to actual team wins?
And does is make sense to you that a shot attempt and a ft attempt are weighted equally by the NBA Efficiency formula? Are you really going to defend that particular metric. Isn't it obvious to you that scoring must be overvalued by an approach which implies that that a shot attempt has one value if it goes in and a different value if it doesn't?
I'm not defending EFF at all. It's very 'simple', but that doesn't make it good. However, it is possible to create a weaker metric, and Berri seems to have done that.
Every shot I've ever seen has one value if it goes in and another if it doesn't. In fact, they're mutually exclusive events. Like winning or losing, you play a game until one or the other happens. Then you move up or down one game in the standings.
This thread isn't supposed to be about MikeStats, and it's way too complex to describe. I take scoring rate and multiply it by (TS%/.530). Other factors are applied, dealing with Min/G, % of games started, opponent PPG, % of FG assisted, and maybe some others. All these bump a 20 Pts/36 scorer up or down.
Scoring is then defined to have a weight of 1. The other weights are subject to the season (or postseason, or playoff series) involved. Typical weights are:
Reb .98
Ast 1.32
PF -.23
Stl 1.6
TO -1.6
Blk 1.4
These are determined just as you suggest, by correlations to team (pythagorean) wins. I give 'high' credit to scorers: A 20 ppg guy who shoots .400 gets 2/3 the scoring credit as a 20 ppg guy who shoots .600 -- all other factors being equal. This recognizes a couple of facts:
- As KJB says, a missed shot is not a lost possession, it's 3/4 of a lost possession. A made shot, however, gives the opponent 100% of 1 possession.
- Since a 40% shot is much better than no shot, it's a great deal better to be able to get a shot, even if it isn't a high-% shot. Of the 60% which are missed, you get 1/4 back as OffReb. That's 15% of all shots taken by the 40% guy.
Calling <50% shooters non-contributors is as nonsensical as calling rebounders who get less than 10% of rebounds non-contributors. It leads you to ludicrous conclusions like "McGrady was worse than the average SG", and "Iverson lost the series".
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Fri May 18, 2007 4:53 pm Post subject: Reply with quote
KJB - Not sure which tweak you mean. Clearly you need to account for offensive rebounding. But that isn't too hard is it? You give the rebounder credit for securing a board, and you move on. I don't know. You seem to be suggesting that the shooter deserves some sort of credit for creating an offensive rebounding opportunity for his teammate. That seems like a strange method of accounting.
Mike - For you to say that his metric is weaker than NBA Efficiency, well that's pretty amazing. This is a metric that as I have said weighs foul shots and fgs equally, which is just ridiculous. Furthermore, any player who shoots better than 33% on two point shots and 25% on three point shots is rewarded for each additional shot he takes. That's just dumb. The net effect of NBA Efficiency is to drastically overvalue scorers.
I dont really understand your weights. It also doesn't seem like you have derived them in the same manner as Berri. How is it possible that an assist is worth 1.32 points. This would suggest that the passer is getting 66% fo the credit for a two point hoop. Berri's analysis suggest that an assist is actually worth much less than a point. Is a rebound really worth less than a block? That seems amazing to me. With a rebound you capture or retain possession for your team. This would seem to a much more valuable action than a block, which frequently fails to secure possession for a team.
Is there a thread you can refer me to that explains your system in more depth. Or perhaps a paper you have written about it?
Back to top
View user's profile Send private message
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Fri May 18, 2007 5:28 pm Post subject: Reply with quote
"A missed shot is 3/4 of a lost possession". That's true (by RebR%), but ratings punish the missed shot full way and then prize back the offensive rebounder apart. By the same logic, a rebound is 1 full possession, but the rebounder must be credited 2/3 and 1/3 to the shot defender ( if it's another teammate), but teammates are never prized back, the rebound steal all the defensive action, regressions talk about that. That's old stuff.
It's true that FG% has a correlation about a 40% of a won or lost game, but, the problem is that Berri thinks you must allways try to give the ball to a 60%FG scorer, because he thinks that high percentage is just because his isolated scoring skills. A lot of unidimensional high percentage scorers are overrated by numbers. They score only at special and opportune situations and a lot of help . It doesn't mean they deserve more ball touches. Although is true that 53% TS must be the limit for a player to get license to shoot, scoring totals must be a wider part of the usage equation.
WOW critics usage and creating, but they overrate high assists players as much as PER, then they are prizing it.
It's true 20ppg is the limit that separates good scorers from the rest, no matter the FG%. I think this Production means about a 50% from scoring IQ. Ballhandling and efficiency another 20% each, and off the ball 10%.
Last edited by Harold Almonte on Fri May 18, 2007 7:57 pm; edited 3 times in total
Back to top
View user's profile Send private message
Mike G
Joined: 14 Jan 2005
Posts: 3559
Location: Hendersonville, NC
PostPosted: Fri May 18, 2007 7:35 pm Post subject: Reply with quote
Flint wrote:
... How is it possible that an assist is worth 1.32 points. This would suggest that the passer is getting 66% of the credit for a two point hoop. ...Is a rebound really worth less than a block? ..., which frequently fails to secure possession for a team.
Blocks may be a proxy for inside defense, altered shots, intimidation, etc. Counted blocks may each represent several good defensive plays. Likewise, a counted assist may represent more than one good pass, ability to penetrate, bringing the ball up the court, etc.
I give less than full credit for FG which were assisted. In any case, 1.32 credits is less than 2 points' credit. With many FG, perhaps the assist man should get more than 50% credit for the score. For a 3-ptr, the assister only gets about 1/3 of the credit.
The defensive rebound is another feat which by itself doesn't secure possession. Before there is a rebound, there is a missed shot; before that, there is some defensive pressure. The rebound only seals the deal.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Fri May 18, 2007 7:43 pm Post subject: Reply with quote
I still dont get it. Can you refer me to a post which explains your system in more detail? I think its really interesting that you did your own regression analysis to find the value of the various box score stats to team wins, but arrived at very different values than Berri.
Back to top
View user's profile Send private message
kjb
Joined: 03 Jan 2005
Posts: 864
Location: Washington, DC
PostPosted: Fri May 18, 2007 8:02 pm Post subject: Reply with quote
Flint wrote:
KJB - Not sure which tweak you mean. Clearly you need to account for offensive rebounding. But that isn't too hard is it? You give the rebounder credit for securing a board, and you move on. I don't know. You seem to be suggesting that the shooter deserves some sort of credit for creating an offensive rebounding opportunity for his teammate. That seems like a strange method of accounting.
I'm not sure what thought of mine you think you're responding to because I wrote nothing that resembles what you're describing. I'm not talking about how the offensive rebounder should be credited for getting the board, I'm talking about the value of a missed shot. At no point did I suggest crediting the guy who missed the shot for creating an offensive rebounding opportunity. I'm questioning how much the shooter should be penalized for missing.
Back to top
View user's profile Send private message AIM Address Yahoo Messenger
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Fri May 18, 2007 8:09 pm Post subject: Reply with quote
Quote:
I still dont get it. Can you refer me to a post which explains your system in more detail? I think its really interesting that you did your own regression analysis to find the value of the various box score stats to team wins, but arrived at very different values than Berri.
There is a post outhere where Dan R. talks about Berri didn't found all of his stats weights by regressions. Some of them are just as by logic as a primitive tendex formula.
Back to top
View user's profile Send private message
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Fri May 18, 2007 8:12 pm Post subject: Reply with quote
Quote:
There's another thought -- maybe shooting a low percentage can be a net benefit because of offensive rebounding. Last season, roughly a quarter of missed field goal attempts went to the offense -- often to a guy in a good position to get a high percentage follow-up opportunity.
KJB - That's what I was referring to. I guess I misunderstood you.
To me, what happens after a shot doesn't change the accounting at the individual level. If Ben Gordon misses a shot and Tyrus Thomas grabs an o rebound, Ben Gordon should receive the same deduction as if the other team had gained possession. Its not like you can just erase the missed shot.
Back to top
View user's profile Send private message
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Fri May 18, 2007 8:26 pm Post subject: Reply with quote
Maybe you must obviate offensive rebounds around missed shots like defense around a defensive rebound is obviated and tried to repair with a team defense adjust., and then you just use a team offense adjust. I think is fair.
Back to top
View user's profile Send private message
Mike G
Joined: 14 Jan 2005
Posts: 3559
Location: Hendersonville, NC
PostPosted: Fri May 18, 2007 8:37 pm Post subject: Reply with quote
Flint wrote:
To me, what happens after a shot doesn't change the accounting at the individual level. If Ben Gordon misses a shot and Tyrus Thomas grabs an o rebound, Ben Gordon should receive the same deduction as if the other team had gained possession. Its not like you can just erase the missed shot.
A missed FGA isn't worth 0% of a successful (2 point) possession because of a potential for an OffReb. A turnover is a zero. Tying up the opponent (jump ball) is somewhere between 0 and 100%, as is a blocked shot, a deflected pass, and lots of other plays, counted and uncounted.
The value of the play, for better or worse, shouldn't depend on what stats are being counted. The missed shot ends up being counted most of the time in an O or D Reb. Or there may be a foul called in the rebounding scrum. Seldom is a missed shot worth 0 points or possessions before the rest of the play has transpired.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
DLew
Joined: 13 Nov 2006
Posts: 224
PostPosted: Fri May 18, 2007 8:41 pm Post subject: Reply with quote
Well, by this logic a missed shot is as bad as a turnover. However, you have acknowledged that a missed shot can be rebounded. A turnover cannot be. So are these two really equally bad events?
I think it is reasonable to suggest that an missed shot only costs you 75% of a possession because that is the probability that you lose the ball, while a turnover costs you 100% of a possession (obviously).
The rebounder should get credit for converting a 25% situation (missed shot) into a possession. This would imply that an offensive rebound is worth .75 of a possession and a defensive rebound .25.
That wages of wins treats a missed shot as equal to a turnover puzzles me, clearly, indisputably, a guy who turns the ball over every time he touches it hurts the team more than a guy who misses a shot every time he touches it (all else equal) because at least the missed shot guy creates an opportunity for his team to get the offensive rebound.
Back to top
View user's profile Send private message
kjb
Joined: 03 Jan 2005
Posts: 864
Location: Washington, DC
PostPosted: Fri May 18, 2007 8:42 pm Post subject: Reply with quote
Flint wrote:
Quote:
There's another thought -- maybe shooting a low percentage can be a net benefit because of offensive rebounding. Last season, roughly a quarter of missed field goal attempts went to the offense -- often to a guy in a good position to get a high percentage follow-up opportunity.
KJB - That's what I was referring to. I guess I misunderstood you.
I guess you did. Smile What I was trying to say there is what I've been trying to say throughout my participation in this thread -- all missed shots are not created equal. A missed shot that's rebounded by a teammate has the same effect on the team as a pass -- except, of course, that the shot-clock resets.
Quote:
To me, what happens after a shot doesn't change the accounting at the individual level. If Ben Gordon misses a shot and Tyrus Thomas grabs an o rebound, Ben Gordon should receive the same deduction as if the other team had gained possession. Its not like you can just erase the missed shot.
But, if Ben Gordon misses a shot, and Thomas grabs the rebound and scores, the net effect is still the same for the Bulls -- points on the scoreboard. The effect of Gordon's missed shot is nothing. It has spent no resource -- the possession is still alive because of Thomas's rebound.
Back to top
View user's profile Send private message AIM Address Yahoo Messenger
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Fri May 18, 2007 9:05 pm Post subject: Reply with quote
Then, the last word would be that the (missed FG-O Reb) 70/30 relationship is about the same as 70/30 (D Reb-shot guarding). Then you get your 70/30RebR% and FG missed by scorer effect- on the shot defense effect 70/30%.
And remember even without defending you can be defending. Just letting a scorer to shoot from his weak zone and regress to his efficiency mean, you are defending. Defense zones are something about it.
Back to top
View user's profile Send private message
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Fri May 18, 2007 9:18 pm Post subject: Reply with quote
Quote:
But, if Ben Gordon misses a shot, and Thomas grabs the rebound and scores, the net effect is still the same for the Bulls -- points on the scoreboard. The effect of Gordon's missed shot is nothing. It has spent no resource -- the possession is still alive because of Thomas's rebound.
KJB - Do you really believe this? Well you wrote it, so you must. I don't know, it doesnt make any sense to me to say that because a missed shot was rebounded by your own team, then the shooter hasn't been unproductive, but if the other team gets the ball, then he has. That just seems highly illogical to me. A missed shot is not like a pass in any way. Most passes dont go to the opponent. Most shot you miss do.
I mean, would you say to a guy who survived a round of russian roullette, hey, you didn't blow your head off, so therefore sticking a gun in your mouth and pulling the trigger had no effect.
Mike - Berri doesnt deal in missed fg's. He deals in points and shot attempts. You have an attempt whether it goes in or not. How can you reasonably judge shooting efficiency if, as NBA Efficiency does, you only count missed fg's and not all shot attempts. As I have said before, and as Berri has said many times, doing so results in a player showing a profit on his scoring as long as he shoots above 33% on two points field goals and above 25% on three pointers. Berri's approach, which essentially compares a players TS% to the average for his position seems to be a much better approach. NBA Efficiency clearly overrates scoring. Does anyone dispute this? I mean a metric that has rated Antoine Walker so high for so many years has to be fundamentally flawed.
Dlaw - I dont see any reason why a shot attempt and a turnover shouldn't have the same value.
PER vs EFF (Chronz1, 2007)
PER vs EFF (Chronz1, 2007)
Last edited by Crow on Tue May 10, 2011 9:27 pm, edited 1 time in total.
Re: PER vs EFF
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Fri May 18, 2007 9:31 pm Post subject: Reply with quote
DLaw - Doesn't it seem to you that you can't both credit a player for regaining possession by capturing an offensive rebound, and also reduce the value of a missed fg to reflect the fact that shot are often rebounded. This seems to be double counting.
Back to top
View user's profile Send private message
DLew
Joined: 13 Nov 2006
Posts: 224
PostPosted: Sat May 19, 2007 2:46 am Post subject: Reply with quote
Seems to me that the value of missing a shot and getting an offensive rebound should sum to zero, because there is no change in possession. The values I suggested (-.75, .75) sum to zero, so there is no double counting.
WOW on the other hand does double count. When you miss a shot and they get a rebound you lose a WOW point and they gain one, suggesting they are leading by two, when in fact the swing has only been one expected point, the expected point that you did not realize offensively. This is because when you score the opposing team gets the ball.
To elaborate further, if the opposing team scores the expected possession value of one point (let's say they hit a free throw) you get the ball and your expected value for that possession is one, so you are essentially tied.
If instead of allowing them to score you stopped them and got the rebound you would hold the ball with the expectation of scoring one point while in possession, the opposing team would still have a zero on the scoreboard. Clearly your expected lead is 1. Why then, does WOW say you are ahead by two at this juncture?
So, you think that a team that makes a two point field goal 50% of the times down the floor and throws the ball directly out of bounds the other half of the time would be equal to a team that takes a shot every time but makes only 50%? This is factually untrue. The team that has missed shots instead of turnovers will be better because they will get some offensive rebounds and second chance points. Berri gives all of the value to the person who gets the rebound and none to the person who created the rebound by missing a shot. It is inarguable that turning the ball over instead of missing a shot costs a team about a quarter of point.
Back to top
View user's profile Send private message
kjb
Joined: 03 Jan 2005
Posts: 865
Location: Washington, DC
PostPosted: Sat May 19, 2007 6:40 am Post subject: Reply with quote
Flint wrote:
Quote:
But, if Ben Gordon misses a shot, and Thomas grabs the rebound and scores, the net effect is still the same for the Bulls -- points on the scoreboard. The effect of Gordon's missed shot is nothing. It has spent no resource -- the possession is still alive because of Thomas's rebound.
KJB - Do you really believe this? Well you wrote it, so you must. I don't know, it doesnt make any sense to me to say that because a missed shot was rebounded by your own team, then the shooter hasn't been unproductive, but if the other team gets the ball, then he has. That just seems highly illogical to me. A missed shot is not like a pass in any way. Most passes dont go to the opponent. Most shot you miss do.
I mean, would you say to a guy who survived a round of russian roullette, hey, you didn't blow your head off, so therefore sticking a gun in your mouth and pulling the trigger had no effect.
Either I'm having a uniquely difficult time expressing myself clearly on this thread, or you have a remarkable gift for misunderstanding. I'll try once more before I give up trying to have a conversation with you.
Ben Gordon passes to Tyrus Thomas, who dribbles around, makes a move and scores a basket. Net TEAM effect -- 2 points.
Ben Gordon misses a shot, Tyrus Thomas gets the offensive rebound and scores. Net TEAM effect -- 2 points.
In this example, Gordon's missed shot has the identical TEAM effect as his pass. In this example, Gordon's miss has no negative effect on his TEAM.
Back to top
View user's profile Send private message AIM Address Yahoo Messenger
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Sat May 19, 2007 8:45 am Post subject: Reply with quote
There's some kind of basketball "Fibonacci" 70/30 proportion outhere (even with Assist (pass-screen-auto)/AssistedFG), but it's difficult to accept it without a consistent and definetly demonstration. Even for me.
Back to top
View user's profile Send private message
Mike G
Joined: 14 Jan 2005
Posts: 3583
Location: Hendersonville, NC
PostPosted: Sat May 19, 2007 11:16 am Post subject: Reply with quote
I forget who Fibonacci was, or what this thread started out to be. But I wonder about the ballpark estimate that 25% of missed FG become ORb. Consider these team averages from this past season:
FG - 36.2
FGA - 78.9
FGM - 42.8 (missed = FGA-FG)
ORb - 11.0
DRb - 29.6
ORb/TRb = 11.0/40.7 = .258
This may be the origin of the 25% figure we've been throwing about. But consider this other source of rebounds:
FT - 19.4
FTA - 25.8
FTM - 6.4 (=25.8-19.4)
Of 6.4 FTM/G, how many are DReb, and how many are OReb? I've estimated 53% are D and 3.5% are O (The other 43.5% are 1st-of-2 FT or other non-reboundables). (There is probably a better estimate, but this gets to the point.)
So from missed FT, we typically see (.53*6.4 = ) 3.4 DRb, and 0.2 ORb.
Therefore, from missed FG, we get (29.6-3.4 = ) 26.2 DRb, and 10.8 ORb.
The total is (26.2+10.8) 37.0 Reb from FGM.
Then we can guess (10.8/37.0 = ) .292 of FGM are ORb.
29.2% is much closer to the Fibonacci/Almonte standard of 30%.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Sat May 19, 2007 11:29 am Post subject: Reply with quote
KJB - I can see the problem now. The thread begin with a poll about what the best statistical method for assigning credit at the level of the individual is. You jumped there to talking about the action at the team level. Totally different. Yes, if the end result is that the Bulls score two points, then Ben Gordon's miss didnt end up hurting their team offensive efficiency. That doesn't mean his action doesn't have a cost and doesnt require a "demerit." At the individual level you have to deduct Ben Gordon for missing a shot. His shot is not like a pass. If it were, you couldn't credit Tyrus Thomas for salvaging the possession with a rebound, correct? In order to give Thomas the credit for rebounding there has to be a deduction somewhere else.
DLaw - We agree that a shot attempt that is missed and a rebound have equal value. They should sum to zero. Where we diverge is that I believe that a point scored has the same value as a rebound. You seem to be saying that a point is more valuable than a rebound or shot attempt. That a point is worth 1 and a rebound worth .75. I dont get this.
You pose a question about how a sequence should be scored. Here is how I would do so.
My team misses shot = - 1
Opposing team gains rebound = 1
Opposing team misses shot = -1
My team gets rebound = 1
My teams net winscore = 0
Opposing teams winscore = 0
Points on the scoreboard - equal
I wouldn't agree that "your expected lead is 1," since you have possession. You could only say that if you were guaranteed to have one more possession than the other team. In a basketball game, the number of possessions enjoyed by each team is equal. (BTW, I really dont think this is really a useful way of thinking.) Therefore, I would say that the expected value of the opposing team's possession which will inevitably follow yours, cancels out your "expected lead."
Vis a vis this:
Quote:
So, you think that a team that makes a two point field goal 50% of the times down the floor and throws the ball directly out of bounds the other half of the time would be equal to a team that takes a shot every time but makes only 50%? This is factually untrue.
To be honest, I find this incredibly murky and confusing. I dont really know what you are saying. In Berri's system a point scored, a turnover, a rebound, and a shot attempt have the same value. That's because that is the result suggested by regression analysis.
From the perspective of winscore, a team that came down and scored every time it shot would gain one unit of winscore for each two point field goal. It would be deducted a unit for every turnover. Its net for winscore would be 0 if it turned the ball over half the time and scored half the time. A team that committed no turnovers and hit 50% of its shots would similarly have a winscore of 0, at least from its scoring activity. However, it would, as you suggest, gain positive credit for offensive rebounds collected and for extra points scored off those boards. So, its "team winscore" whatever that means, would be higher. Players would be credited for the rebounds and extra points scored. And this makes sense, doesn't it, since Team 2 would, as you suggest, end up scoring on more than 50% of its possessions.
I would also point out that you can have shot attempts before you have turnovers. A team can come down, miss ten shots, collect ten offensive rebounds, then have a turnover. It's net for the possession in that case is still -1, just as it would be if it came down and missed a shot, and it was rebounded by the other team.
Dlaw - I think you have to admit that the value of a rebound, and the value of a point scored must be the same. If the expected value of one possession is a point, which we agree on, than the value of capturing a rebound and thereby creating or salvaging a possession must be one point also. It should not be .75 of a point. Your thinking that we ought to credit "the person who created the rebound by missing a shot" is just barney.
Back to top
View user's profile Send private message
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Sat May 19, 2007 12:14 pm Post subject: Reply with quote
Quote:
the value of a rebound, and the value of a point scored must be the same. If the expected value of one possession is a point, which we agree on, than the value of capturing a rebound and thereby creating or salvaging a possession must be one point also. It should not be .75 of a point. Your thinking that we ought to credit "the person who created the rebound by missing a shot" is just barney.
Do you really think the rebounder created alone his rebound stat? Equalizing rebounds (and missed shots) to a possession are shortcuts for easy calcs, and mostly because the straitjacket imposed by boxscore. Possession and Win are absolut things. You can force or adjust every variable to fit with them and to be what this topic started: almost credible.
MikeG- Does exist some stat study about what % of deflections and blocks are grabbed back by the offensive team? Sorry for my socratic way of discussion.
Last edited by Harold Almonte on Sat May 19, 2007 12:34 pm; edited 1 time in total
Back to top
View user's profile Send private message
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Sat May 19, 2007 12:32 pm Post subject: Reply with quote
Yes Harold. I actually think a rebounder deserves full credit for salvaging a possession, and that splitting that credit between the shooter, who "created the opportunity for the rebound," and the rebounder who actually did the work, doesn't make any sense.
Back to top
View user's profile Send private message
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Sat May 19, 2007 12:44 pm Post subject: Reply with quote
Well, you are from the school that consider the offensive rebounds a new possession and not a (70/30) offensive continuation play. The same school consider DReb. is a new possession and not a (70/30) offensive-defensive continuation play. WOW is in that school and by this, rebounders are suposed to be overrated. It's just basketball philosophy, you can be also credible.
Last edited by Harold Almonte on Sat May 19, 2007 12:57 pm; edited 1 time in total
Back to top
View user's profile Send private message
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Sat May 19, 2007 12:49 pm Post subject: Reply with quote
I am not one of those people. An offensive rebound extends a possession, it doesnt create a new one.
Back to top
View user's profile Send private message
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Sat May 19, 2007 12:59 pm Post subject: Reply with quote
I'm lost here. I don't know how you can extend one possession play (the missed FG punish) to a second (the OReb prize) and end being just one. The punish and the prize is not an extending, is a double counting.
Back to top
View user's profile Send private message
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Sat May 19, 2007 1:21 pm Post subject: Reply with quote
Harold - I have seen your posts over at the WOW. You read his stuff, you must understand where I am coming from. A shot attempt is a deduction against the shooter. This represents the cost of giving up possession and putting the ball in play. However, the possession doesn't end unless the opposing team grabs the rebound. A team can extend its possession by grabbing an offensive rebound. If this happens, the credit belongs solely to the rebounder. And the net result of this sequence is zero overall. The rebound cancels out the missed shot and we are back at the same place we were at the beginning of the possession. On the team level, nothing has changed. On the individual level however the rebounder's stats have improved, the shooter's have gone the other way.
That make sense? Where is the double counting?
The fact that offensive rebounds occur at some general rate, 30% of the time is what you suggest, doesn't mean anything to how we should account for individual performance. Mathematically, its just strange to think about it this way.
Honestly, I am kind of mystified.
Back to top
View user's profile Send private message
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Sat May 19, 2007 2:02 pm Post subject: Reply with quote
Quote:
A shot attempt is a deduction against the shooter. This represents the cost of giving up possession and putting the ball in play. However, the possession doesn't end unless the opposing team grabs the rebound. A team can extend its possession by grabbing an offensive rebound. If this happens, the credit belongs solely to the rebounder. And the net result of this sequence is zero overall. The rebound cancels out the missed shot and we are back at the same place we were at the beginning of the possession. On the team level, nothing has changed. On the individual level however the rebounder's stats have improved, the shooter's have gone the other way.
-What I understand is you have (-1 possession punish) a missedFG + (1 possession prize) an OReb = 1 (absolute) possession and 0 points, and you have two rated individual plays. I know the trick is separate the team total possession attempt from the two individual punish-prize plays, and you don't rate neither the team nor the possession, but the two players involved.
-What the other school says is (0) punish for scorer if there is (+1 prize) to OReb= 1 team possession.
What I think is possessions don't belong to players, but team, and you can't talk about possessions and prize and punish just one player each, that's a shortcut. What I'm trying to say is all FGA possessions are a (70/30) relationship between two players, and one player can't be entirely punished or prized.
Back to top
View user's profile Send private message
mateo82
Joined: 06 Aug 2005
Posts: 211
PostPosted: Sat May 19, 2007 2:27 pm Post subject: Reply with quote
For me it's neither. I like win shares for the "all encompassing" stat, but i'm not a huge fan of any of them. Win shares at least can make predictions, so in that way it is falsifiable whereas neither EFF nor PER are. My feeling with any metric is that it must prove it's own value. If not I see no reason to use it, even if it's widely used.
Back to top
View user's profile Send private message
jkubatko
Joined: 05 Jan 2005
Posts: 702
Location: Columbus, OH
PostPosted: Sat May 19, 2007 3:38 pm Post subject: Reply with quote
mateo82 wrote:
I like win shares for the "all encompassing" stat [...]
Hear, hear!
_________________
Regards,
Justin Kubatko
Basketball-Reference.com
NickS
Joined: 30 Dec 2004
Posts: 384
PostPosted: Sat May 19, 2007 3:52 pm Post subject: Reply with quote
Flint wrote:
You pose a question about how a sequence should be scored. Here is how I would do so.
My team misses shot = - 1
Opposing team gains rebound = 1
Opposing team misses shot = -1
My team gets rebound = 1
My teams net winscore = 0
Opposing teams winscore = 0
Points on the scoreboard - equal
What about a situation in which the other team hits their shot:
My team misses shot = - 1
Opposing team gains rebound = 1
Opposing team hits shot = +1
My teams net winscore = -1
Opposing teams winscore = +2
Points on the scoreboard - +2 for opposing team.
Unless you give your team a +1 credit for receiving the ball after an opponents made shot, this doesn't add up.
Back to top
View user's profile Send private message
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Sat May 19, 2007 4:00 pm Post subject: Reply with quote
Nick - its not supposed to add up. Winscore is not meant to equal the points on the scoreboard. I think thats what you mean right?
As I said before, winscore isn't really an instrument designed to work at the team level.
Back to top
View user's profile Send private message
NickS
Joined: 30 Dec 2004
Posts: 384
PostPosted: Sat May 19, 2007 4:10 pm Post subject: Reply with quote
Flint wrote:
Nick - its not supposed to add up. Winscore is not meant to equal the points on the scoreboard. I think thats what you mean right?
As I said before, winscore isn't really an instrument designed to work at the team level.
That doesn't make sense. I thought the appeal of Berri's system was that it was derived from the value events have to a team.
If that isn't the case, what were you trying to argue in the bit I quoted?
Trying to make sense of this, for a moment, it seems like any rating system is going to have similar problems with the difference between a TO that's counted as a steal for the other team, and a TO that that isn't. Assuming that you count steals as positive value, the former will result in a greater differential of value than the latter -- and it's just an artifact of the type of TO.
At the same time, it seems like the two most common ways for a possession to end are either a made basket or a miss and a rebound and that it would make sense for those two events to have equal value in a rating system.
This is a little scattershot, because I'm still thinking this through, but my intuition is to say that DLew's comment about double counting isn't precisely the problem (because of the analogy to steals above) but that something is a little odd here.
Back to top
View user's profile Send private message
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Sat May 19, 2007 4:18 pm Post subject: Reply with quote
MikeG wrote:
Quote:
29.2% is much closer to the Fibonacci/Almonte standard of 30%
It's not any Fibonacci, it's just the famous Rosembaum's sharing (I apologize by not give his credits), and one of his regressions obtained this "near to 70/30" link between missedFG and DReb.
My THEORY is just an extention. I'm talking about a longer chain of 70/30 links that begin with the assist at the offensive end, and finish with the DReb at the defensive end.
Back to top
View user's profile Send private message
NickS
Joined: 30 Dec 2004
Posts: 384
PostPosted: Sat May 19, 2007 4:21 pm Post subject: Reply with quote
Consider this example, a game of 1 on 1 between two players (so we aren't worried about the team aspect).
Player A shoots a 3-pointer on every possession and shoots 33.33%.
Player B shoots a 2-pointer on every possession and shoots 50%.
Both players get 20% of their own offensive rebounds.
At the end of 100 trips up and down the floor (possessions) player A has shot 115 3 pointers, hit 39, and gotten 15 offensive rebounds.
Player B has shot 111 shots, hit 56 of them, and gotten 11 offensive rebounds.
Player A's total Win Score:
117 points scored = 117
100 possessions used = -100
44 defensive rebounds = +44
15 offensive rebounds = +15
= 76
Player B's total Win Score:
112 points scored = 112
100 possessions used = -100
61 defensive rebounds = +61
11 offensive rebounds = +11
= 84
Player B has a higher total WS despite losing the game.
Tell me where that's wrong?
Last edited by NickS on Sat May 19, 2007 4:24 pm; edited 1 time in total
Back to top
View user's profile Send private message
NickS
Joined: 30 Dec 2004
Posts: 384
PostPosted: Sat May 19, 2007 4:24 pm Post subject: Reply with quote
Actually I think this is wrong:
NickS wrote:
Player A's total Win Score:
117 points scored = 117
100 possessions used = -100
44 defensive rebounds = +44
15 offensive rebounds = +15
= 76
Player B's total Win Score:
112 points scored = 112
100 possessions used = -100
61 defensive rebounds = +61
11 offensive rebounds = +11
= 84
This should be:
Player A's total Win Score:
117 points scored = 117
115 shot attempts = -115
44 defensive rebounds = +44
15 offensive rebounds = +15
= 61
Player B's total Win Score:
112 points scored = 112
111 shot attempts = -111
61 defensive rebounds = +61
11 offensive rebounds = +11
= 73
So the difference is larger than I first thought. But I may well be missing something else obvious.
Back to top
View user's profile Send private message
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Sat May 19, 2007 4:26 pm Post subject: Reply with quote
I was responding to Mike's post and trying to show how Berri's system worked differently from how he described it. I suppose it may have appeared that I was trying to show that winscore and the scoreboard should add up, but that is not the case and it wasn't my intention to suggest that.
There are quirky situations that you have to incorporate into the system. For instance, if a shot caroms of the rim and out of bounds. This results in a team rebound which is incorporated into the system as part of the infamous "team adjustment. " A turnover that doesn't correspond to a steal, for instance if someone throws to Eddy Curry and it bounces off his hands out of bounds, I suppose would be dealt with in the same manner I believe. I am not sure exactly, but either way, thats a relatively infrequent occurrence.
Let me know what you think is odd, will try to answer your question...
Back to top
View user's profile Send private message
NickS
Joined: 30 Dec 2004
Posts: 384
PostPosted: Sat May 19, 2007 4:37 pm Post subject: Reply with quote
Flint wrote:
Let me know what you think is odd, will try to answer your question...
Try the example I give above with player A and player B. They are identical in their eFG% and their ability to get offensive rebounds. If they play each other, on average Player A will win the game and player B will have a higher WS.
Doesn't this make it seem like something is being under/over valued? That WS is, in other words, a biased rather than unbiased estimate in that circumstance?
Back to top
View user's profile Send private message
NickS
Joined: 30 Dec 2004
Posts: 384
PostPosted: Sat May 19, 2007 5:33 pm Post subject: Reply with quote
NickS wrote:
This should be:
Player A's total Win Score:
117 points scored = 117
115 shot attempts = -115
44 defensive rebounds = +44
15 offensive rebounds = +15
= 61
Player B's total Win Score:
112 points scored = 112
111 shot attempts = -111
61 defensive rebounds = +61
11 offensive rebounds = +11
= 73
I think I'm getting somewhere here. Let's try this again with the following weights: Shot attempt (-.7), offensive rebound (.7), defensive rebound (.3), so that shot attempt and offensive rebound counterbalance, and shot attempt + defensive rebound = the cost of a lost possession.
In that case we would have:
Player A's total Win Score:
117 points scored = 117
115 shot attempts = -80.5
44 defensive rebounds = +13.2
15 offensive rebounds = +10.5
= 60.2
Player B's total Win Score:
112 points scored = 112
111 shot attempts = -77.7
61 defensive rebounds = +18.3
11 offensive rebounds = +7.7
= 60.3
At this point the two players have identical scores. Which is closer but still not perfect. But, if add a penalty to made shots equivalent to the value of a defensive rebound (to account for the fact that a made shot is a complete change of possession, rather than just .7 of a change of possession we would have.
Player A's total Win Score:
117 points scored = 117
115 shot attempts = -80.5
39 made shots = -11.7
44 defensive rebounds = +13.2
15 offensive rebounds = +10.5
= 48.5
Player B's total Win Score:
112 points scored = 112
111 shot attempts = -77.7
56 made shots = -16.8
61 defensive rebounds = +18.3
11 offensive rebounds = +7.7
= 43.5
The two players WS differ by exactly the difference on the scoreboard!
Obviously .7/.3 is arbitrary. The last method would work whatever weights were used as long as the value of an offensive rebound + defensive rebound equaled 1.
In fact, this would also (somewhat) work with the WoW forumla as long as the penalty for a made shot still equalled a defensive rebound (1 point in this case). That would have the odd effect, however, of claiming that a made 2 point basket had 0 value.
Ultimately I think DLew is correct that in having the sum of the value of an offensive rebound and the value of a defensive rebound equal 2 rather than 1 that a form of double counting is going on.
Back to top
View user's profile Send private message
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Sat May 19, 2007 5:38 pm Post subject: Reply with quote
Lol. That's complicated. Let me hack at it.
Each player on average should score one point for every shot he takes.
The way I make it Player A scores 113.32 points, B scores 110. This is because A rebounds 13.32 of his 66.6 misses. He takes 13.32 additional shots, and this translates into 13.32 additional points scored at his established rate. B would score 50 points on his first 100 shots, an additional ten from the ten offensive rebounds collected from his fifty misses.
What their winscore would be seems quite complicated, even though their scoring and shot attempts should cancel out.
A gets 40 rebounds, five defensive rebounds off of B's ten secondary shots, and 13.32 offensive rebounds. So 58.32 total rebounds?
B gets 53.28 misses off A's first shots, and another 8.79 rebounds off of the A's secondary misses. He also collects ten offensive rebounds. His rebounding's contribution to winscore is 72.07.
Total winscores
A. 113.32 + 58.32 = 171.64
B. 110 + 72.07= 182.07
I dont really know what that means. There is a difference of 10 there. I feel like I am missing something. I must have made a mistake. I dont know, check my work, let me know what you think, I am going to go ponder this.
Back to top
View user's profile Send private message
NickS
Joined: 30 Dec 2004
Posts: 384
PostPosted: Sat May 19, 2007 5:55 pm Post subject: Reply with quote
I haven't worked through your numbers yet, but here's something else to consider, that you should translate into the appropriate WS numbers as appropriate.
Imagine two teams A and B, both always shoot 100%, team A shoots only 3 pointers, team B shoots only 2 pointers.
Imagine they both go down the court 10 times and take ten shots.
Team A: 30 points - 10 possessions used = +20
Team B: 20 points - 10 possessions used = +10
At that point the difference in their WS is the same as the difference in points on the scoreboard (10 points).
Now imagine same two teams, same 10 trips up and down the court, but in this case each team takes 5 shots, and has 5 unforced TOs.
Team A: 15 points - 10 possessions used = +5
Team B: 10 points - 10 possessions used = +0
Again, at this point the difference in WS equals the difference on the scoreboard (5 points).
This reinforces my opinion that it isn't the difference between 2 and 3 point shots that's the problem, but that it's something about the way WS counts rebounds.
If difference in WS is equal to difference in points when you have no rebounds, but not equal when rebounds are involved, there's soemthing up with rebounding.
Again, please correct me if I'm mischaracterizing WS.
Back to top
View user's profile Send private message
NickS
Joined: 30 Dec 2004
Posts: 384
PostPosted: Sat May 19, 2007 6:44 pm Post subject: Reply with quote
NickS wrote:
f [we] add a penalty to made shots equivalent to the value of a defensive rebound (to account for the fact that a made shot is a complete change of possession, rather than just .7 of a change of possession we would have.
Player A's total Win Score:
117 points scored = 117
115 shot attempts = -80.5
39 made shots = -11.7
44 defensive rebounds = +13.2
15 offensive rebounds = +10.5
= 48.5
Player B's total Win Score:
112 points scored = 112
111 shot attempts = -77.7
56 made shots = -16.8
61 defensive rebounds = +18.3
11 offensive rebounds = +7.7
= 43.5
As a point of possible interest, I was trying to figure out what the scale on these numbers were. It seems like it should be, based on the theory, points scored above 1 point/possession.
but, if that were the case, the totals would be 17 and 12 since the two players score respectively, 117 and 112 points in 100 possessions.
I figured out that the scale is points scored at a rate above 1 point / possession + the value that's created by defensive rebounds.
As an easy way to see this, imagine two teams that shoot 0%, and don't get offensive rebounds playing 10 possessions.
Team A: 10 shot attempts (-7) + 10 rebounds (+3) = -4
Team B: the same.
If the value of an offensive rebound was set to 1 and the value of a defensive rebound was set to 0 than the two offensively inept teams would have a value of:
Team A: 10 shot attempts (-10) + 10 rebounds (+0) = -10
Team B: the same.
And the value would be exactly equal to the difference between points scored and possessions used.
This is just to say that the weights that are chosen for offensive and defensive rebounding will also act as a scaling factor on the final values.
Note, going back to the quoted example, that if you took each team's positive value for defensive rebounds and, instead, applied it as a penalty to the other team's score (to turn the value of a missed shot into -1/0 rather than -.7/.3) you would get:
Team A: 48.5 - 13.2 -18.3 = 17
Team b: 43.5 - 18.3 -13.2 = 12
Again, equal to the points scored above base efficiency.
So, there you go, that's what the numbers mean. Just something that had been bothering me.
Back to top
View user's profile Send private message
NickS
Joined: 30 Dec 2004
Posts: 384
PostPosted: Sat May 19, 2007 11:36 pm Post subject: Reply with quote
I've posted a lot on this topic, but I've continued to think about it and I want to summarize.
What I have shown is that, if Flint's calculations are correct, that there is a simple example of a case in which the win score calculation will rank the worse player above the better player.
Flint wrote:
Total winscores
A. 113.32 + 58.32 = 171.64
B. 110 + 72.07= 182.07
I want to emphasize that the example is not complicated. There are no assists, steals, blocks, or turn overs involved. It just involves Shots, made and missed baskets, and rebounds.
I believe that this example will be a problem for any linear weight system in which the sum of the values for an offensive and defensive rebound are greater than the value of a possession, and I think I can explain why.*
Take this quote from Flint earlier.
Flint wrote:
[Y]ou must understand where I am coming from. A shot attempt is a deduction against the shooter. This represents the cost of giving up possession and putting the ball in play.
consider the last phrase, "putting the ball in play". If we penalize someone who shoots and misses for having given up possession of the ball, that means that at the point that the ball is coming off the rim, from a WinScore perspective, neither team has possession of the ball. At that point the offensive team has given up possession and both teams are trying to gain possession of the ball.
It's obvious that, at that point, there are two possible outcomes, an offensive or defensive rebound, and that the difference between those two outcomes is one possession.
If the difference in credit that is available to the two teams is greater than the value of one possession, that is not an accurate description of the situation because what is at stake is, precisely, possession of the ball. It can't be worth more than that.
Flint, seriously, I appreciate your getting me to think through this example, it has really helped me clarify my thoughts.
*[Technical note: the other way to look at this, is where we started, that any system in which a missed shot and a defensive rebound combine for more than the value of one possession will have a problem. Because the (negative) value of a missed shot has to equal the (positive) value of an offensive rebound this means the same thing as a system in which the values of an offensive and defensive rebound add up to more than the value of a possession]
Back to top
View user's profile Send private message
NickS
Joined: 30 Dec 2004
Posts: 384
PostPosted: Sun May 20, 2007 11:20 am Post subject: Reply with quote
Yet another thought on this topic. This has now gotten me thinking about the possible goals of player ranking system.
First of all I want to say that the argument above does not show, by itself, that WinScore is a bad linear weights system. Compare, for example, to a system that gave players credits for points scored, and for nothing else. That system would pass the test of always having the difference in credit be equal to the difference on the scoreboard, but would be a much worse system.
This just shows, I think, that WinScore demonstrably overvalues rebounds, as many people have argued. I'm not a fan of WinScore for other reasons, mostly because of the arguments Dan Rosenbaum has made which I find convincing, but this is a demonstration of what errors does one accept in using WinScore as a valuation.
I'm inclined to believe that any linear weight system will have errors, and the value of the system depends, in part on recognizing when those errors occur and being able to take that into account.
So it got me thinking about, descriptively what different linear weight systems attempt to accomplish. That will be a long post, and I may make it a separate topic.
But, in brief, a linear weight system is an attempt to distribute credit, and the two questions to be asked are (1) how much credit / debit is there to be distributed and (2) how should it be distributed (for example, if you have two teammates one of whom takes most of the shots, and the other who gets most of the rebounds how do you share credit for scoring? The shooter wouldn't have possession of the ball without the rebounder getting rebounds, but the rebounder would, presumably, have a harder time turning possessions into points without the shooter).
This series of posts got me thinking about the first question -- how much credit is there to be distributed. PER is closer to saying that the total amount of credit is equal to the total number of points scored -- regardless of how much the opponent scores. WinScore is closer to saying that total credit is the amount by which the team would outscore an average opponent. Two very different idea of "credit."
This just get me thinking again about the value of trying to identify the "marginal" or "replacement" (two different quantities) scoring rate and assinging credit for actions that help a team score more efficiently than the marginal rate.
For example, if you thought the replacement scoring rate was .7 points/possessions the weights could look like this (with a .7/.3 split between offensive and defensive rebounds):
Made basket: 2 or 3 points
Offensive rebound: .49 points
Defensive rebound: .21 points
Shot attempt: -.49 points
Made basket: -.21 points
Back to top
View user's profile Send private message
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Sun May 20, 2007 1:58 pm Post subject: Reply with quote
Nick - The problem here is that Win Score, as noted before, is just a quick and dirty metric. It omits team defense. That's why you observe an anomaly. The more complete Wins Produced metric, does incorporate team defense. And if you look at it through Wins Produced, both player's "teams" do come out equal.
Joined: 25 Mar 2007
Posts: 112
PostPosted: Fri May 18, 2007 9:31 pm Post subject: Reply with quote
DLaw - Doesn't it seem to you that you can't both credit a player for regaining possession by capturing an offensive rebound, and also reduce the value of a missed fg to reflect the fact that shot are often rebounded. This seems to be double counting.
Back to top
View user's profile Send private message
DLew
Joined: 13 Nov 2006
Posts: 224
PostPosted: Sat May 19, 2007 2:46 am Post subject: Reply with quote
Seems to me that the value of missing a shot and getting an offensive rebound should sum to zero, because there is no change in possession. The values I suggested (-.75, .75) sum to zero, so there is no double counting.
WOW on the other hand does double count. When you miss a shot and they get a rebound you lose a WOW point and they gain one, suggesting they are leading by two, when in fact the swing has only been one expected point, the expected point that you did not realize offensively. This is because when you score the opposing team gets the ball.
To elaborate further, if the opposing team scores the expected possession value of one point (let's say they hit a free throw) you get the ball and your expected value for that possession is one, so you are essentially tied.
If instead of allowing them to score you stopped them and got the rebound you would hold the ball with the expectation of scoring one point while in possession, the opposing team would still have a zero on the scoreboard. Clearly your expected lead is 1. Why then, does WOW say you are ahead by two at this juncture?
So, you think that a team that makes a two point field goal 50% of the times down the floor and throws the ball directly out of bounds the other half of the time would be equal to a team that takes a shot every time but makes only 50%? This is factually untrue. The team that has missed shots instead of turnovers will be better because they will get some offensive rebounds and second chance points. Berri gives all of the value to the person who gets the rebound and none to the person who created the rebound by missing a shot. It is inarguable that turning the ball over instead of missing a shot costs a team about a quarter of point.
Back to top
View user's profile Send private message
kjb
Joined: 03 Jan 2005
Posts: 865
Location: Washington, DC
PostPosted: Sat May 19, 2007 6:40 am Post subject: Reply with quote
Flint wrote:
Quote:
But, if Ben Gordon misses a shot, and Thomas grabs the rebound and scores, the net effect is still the same for the Bulls -- points on the scoreboard. The effect of Gordon's missed shot is nothing. It has spent no resource -- the possession is still alive because of Thomas's rebound.
KJB - Do you really believe this? Well you wrote it, so you must. I don't know, it doesnt make any sense to me to say that because a missed shot was rebounded by your own team, then the shooter hasn't been unproductive, but if the other team gets the ball, then he has. That just seems highly illogical to me. A missed shot is not like a pass in any way. Most passes dont go to the opponent. Most shot you miss do.
I mean, would you say to a guy who survived a round of russian roullette, hey, you didn't blow your head off, so therefore sticking a gun in your mouth and pulling the trigger had no effect.
Either I'm having a uniquely difficult time expressing myself clearly on this thread, or you have a remarkable gift for misunderstanding. I'll try once more before I give up trying to have a conversation with you.
Ben Gordon passes to Tyrus Thomas, who dribbles around, makes a move and scores a basket. Net TEAM effect -- 2 points.
Ben Gordon misses a shot, Tyrus Thomas gets the offensive rebound and scores. Net TEAM effect -- 2 points.
In this example, Gordon's missed shot has the identical TEAM effect as his pass. In this example, Gordon's miss has no negative effect on his TEAM.
Back to top
View user's profile Send private message AIM Address Yahoo Messenger
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Sat May 19, 2007 8:45 am Post subject: Reply with quote
There's some kind of basketball "Fibonacci" 70/30 proportion outhere (even with Assist (pass-screen-auto)/AssistedFG), but it's difficult to accept it without a consistent and definetly demonstration. Even for me.
Back to top
View user's profile Send private message
Mike G
Joined: 14 Jan 2005
Posts: 3583
Location: Hendersonville, NC
PostPosted: Sat May 19, 2007 11:16 am Post subject: Reply with quote
I forget who Fibonacci was, or what this thread started out to be. But I wonder about the ballpark estimate that 25% of missed FG become ORb. Consider these team averages from this past season:
FG - 36.2
FGA - 78.9
FGM - 42.8 (missed = FGA-FG)
ORb - 11.0
DRb - 29.6
ORb/TRb = 11.0/40.7 = .258
This may be the origin of the 25% figure we've been throwing about. But consider this other source of rebounds:
FT - 19.4
FTA - 25.8
FTM - 6.4 (=25.8-19.4)
Of 6.4 FTM/G, how many are DReb, and how many are OReb? I've estimated 53% are D and 3.5% are O (The other 43.5% are 1st-of-2 FT or other non-reboundables). (There is probably a better estimate, but this gets to the point.)
So from missed FT, we typically see (.53*6.4 = ) 3.4 DRb, and 0.2 ORb.
Therefore, from missed FG, we get (29.6-3.4 = ) 26.2 DRb, and 10.8 ORb.
The total is (26.2+10.8) 37.0 Reb from FGM.
Then we can guess (10.8/37.0 = ) .292 of FGM are ORb.
29.2% is much closer to the Fibonacci/Almonte standard of 30%.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Sat May 19, 2007 11:29 am Post subject: Reply with quote
KJB - I can see the problem now. The thread begin with a poll about what the best statistical method for assigning credit at the level of the individual is. You jumped there to talking about the action at the team level. Totally different. Yes, if the end result is that the Bulls score two points, then Ben Gordon's miss didnt end up hurting their team offensive efficiency. That doesn't mean his action doesn't have a cost and doesnt require a "demerit." At the individual level you have to deduct Ben Gordon for missing a shot. His shot is not like a pass. If it were, you couldn't credit Tyrus Thomas for salvaging the possession with a rebound, correct? In order to give Thomas the credit for rebounding there has to be a deduction somewhere else.
DLaw - We agree that a shot attempt that is missed and a rebound have equal value. They should sum to zero. Where we diverge is that I believe that a point scored has the same value as a rebound. You seem to be saying that a point is more valuable than a rebound or shot attempt. That a point is worth 1 and a rebound worth .75. I dont get this.
You pose a question about how a sequence should be scored. Here is how I would do so.
My team misses shot = - 1
Opposing team gains rebound = 1
Opposing team misses shot = -1
My team gets rebound = 1
My teams net winscore = 0
Opposing teams winscore = 0
Points on the scoreboard - equal
I wouldn't agree that "your expected lead is 1," since you have possession. You could only say that if you were guaranteed to have one more possession than the other team. In a basketball game, the number of possessions enjoyed by each team is equal. (BTW, I really dont think this is really a useful way of thinking.) Therefore, I would say that the expected value of the opposing team's possession which will inevitably follow yours, cancels out your "expected lead."
Vis a vis this:
Quote:
So, you think that a team that makes a two point field goal 50% of the times down the floor and throws the ball directly out of bounds the other half of the time would be equal to a team that takes a shot every time but makes only 50%? This is factually untrue.
To be honest, I find this incredibly murky and confusing. I dont really know what you are saying. In Berri's system a point scored, a turnover, a rebound, and a shot attempt have the same value. That's because that is the result suggested by regression analysis.
From the perspective of winscore, a team that came down and scored every time it shot would gain one unit of winscore for each two point field goal. It would be deducted a unit for every turnover. Its net for winscore would be 0 if it turned the ball over half the time and scored half the time. A team that committed no turnovers and hit 50% of its shots would similarly have a winscore of 0, at least from its scoring activity. However, it would, as you suggest, gain positive credit for offensive rebounds collected and for extra points scored off those boards. So, its "team winscore" whatever that means, would be higher. Players would be credited for the rebounds and extra points scored. And this makes sense, doesn't it, since Team 2 would, as you suggest, end up scoring on more than 50% of its possessions.
I would also point out that you can have shot attempts before you have turnovers. A team can come down, miss ten shots, collect ten offensive rebounds, then have a turnover. It's net for the possession in that case is still -1, just as it would be if it came down and missed a shot, and it was rebounded by the other team.
Dlaw - I think you have to admit that the value of a rebound, and the value of a point scored must be the same. If the expected value of one possession is a point, which we agree on, than the value of capturing a rebound and thereby creating or salvaging a possession must be one point also. It should not be .75 of a point. Your thinking that we ought to credit "the person who created the rebound by missing a shot" is just barney.
Back to top
View user's profile Send private message
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Sat May 19, 2007 12:14 pm Post subject: Reply with quote
Quote:
the value of a rebound, and the value of a point scored must be the same. If the expected value of one possession is a point, which we agree on, than the value of capturing a rebound and thereby creating or salvaging a possession must be one point also. It should not be .75 of a point. Your thinking that we ought to credit "the person who created the rebound by missing a shot" is just barney.
Do you really think the rebounder created alone his rebound stat? Equalizing rebounds (and missed shots) to a possession are shortcuts for easy calcs, and mostly because the straitjacket imposed by boxscore. Possession and Win are absolut things. You can force or adjust every variable to fit with them and to be what this topic started: almost credible.
MikeG- Does exist some stat study about what % of deflections and blocks are grabbed back by the offensive team? Sorry for my socratic way of discussion.
Last edited by Harold Almonte on Sat May 19, 2007 12:34 pm; edited 1 time in total
Back to top
View user's profile Send private message
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Sat May 19, 2007 12:32 pm Post subject: Reply with quote
Yes Harold. I actually think a rebounder deserves full credit for salvaging a possession, and that splitting that credit between the shooter, who "created the opportunity for the rebound," and the rebounder who actually did the work, doesn't make any sense.
Back to top
View user's profile Send private message
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Sat May 19, 2007 12:44 pm Post subject: Reply with quote
Well, you are from the school that consider the offensive rebounds a new possession and not a (70/30) offensive continuation play. The same school consider DReb. is a new possession and not a (70/30) offensive-defensive continuation play. WOW is in that school and by this, rebounders are suposed to be overrated. It's just basketball philosophy, you can be also credible.
Last edited by Harold Almonte on Sat May 19, 2007 12:57 pm; edited 1 time in total
Back to top
View user's profile Send private message
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Sat May 19, 2007 12:49 pm Post subject: Reply with quote
I am not one of those people. An offensive rebound extends a possession, it doesnt create a new one.
Back to top
View user's profile Send private message
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Sat May 19, 2007 12:59 pm Post subject: Reply with quote
I'm lost here. I don't know how you can extend one possession play (the missed FG punish) to a second (the OReb prize) and end being just one. The punish and the prize is not an extending, is a double counting.
Back to top
View user's profile Send private message
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Sat May 19, 2007 1:21 pm Post subject: Reply with quote
Harold - I have seen your posts over at the WOW. You read his stuff, you must understand where I am coming from. A shot attempt is a deduction against the shooter. This represents the cost of giving up possession and putting the ball in play. However, the possession doesn't end unless the opposing team grabs the rebound. A team can extend its possession by grabbing an offensive rebound. If this happens, the credit belongs solely to the rebounder. And the net result of this sequence is zero overall. The rebound cancels out the missed shot and we are back at the same place we were at the beginning of the possession. On the team level, nothing has changed. On the individual level however the rebounder's stats have improved, the shooter's have gone the other way.
That make sense? Where is the double counting?
The fact that offensive rebounds occur at some general rate, 30% of the time is what you suggest, doesn't mean anything to how we should account for individual performance. Mathematically, its just strange to think about it this way.
Honestly, I am kind of mystified.
Back to top
View user's profile Send private message
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Sat May 19, 2007 2:02 pm Post subject: Reply with quote
Quote:
A shot attempt is a deduction against the shooter. This represents the cost of giving up possession and putting the ball in play. However, the possession doesn't end unless the opposing team grabs the rebound. A team can extend its possession by grabbing an offensive rebound. If this happens, the credit belongs solely to the rebounder. And the net result of this sequence is zero overall. The rebound cancels out the missed shot and we are back at the same place we were at the beginning of the possession. On the team level, nothing has changed. On the individual level however the rebounder's stats have improved, the shooter's have gone the other way.
-What I understand is you have (-1 possession punish) a missedFG + (1 possession prize) an OReb = 1 (absolute) possession and 0 points, and you have two rated individual plays. I know the trick is separate the team total possession attempt from the two individual punish-prize plays, and you don't rate neither the team nor the possession, but the two players involved.
-What the other school says is (0) punish for scorer if there is (+1 prize) to OReb= 1 team possession.
What I think is possessions don't belong to players, but team, and you can't talk about possessions and prize and punish just one player each, that's a shortcut. What I'm trying to say is all FGA possessions are a (70/30) relationship between two players, and one player can't be entirely punished or prized.
Back to top
View user's profile Send private message
mateo82
Joined: 06 Aug 2005
Posts: 211
PostPosted: Sat May 19, 2007 2:27 pm Post subject: Reply with quote
For me it's neither. I like win shares for the "all encompassing" stat, but i'm not a huge fan of any of them. Win shares at least can make predictions, so in that way it is falsifiable whereas neither EFF nor PER are. My feeling with any metric is that it must prove it's own value. If not I see no reason to use it, even if it's widely used.
Back to top
View user's profile Send private message
jkubatko
Joined: 05 Jan 2005
Posts: 702
Location: Columbus, OH
PostPosted: Sat May 19, 2007 3:38 pm Post subject: Reply with quote
mateo82 wrote:
I like win shares for the "all encompassing" stat [...]
Hear, hear!
_________________
Regards,
Justin Kubatko
Basketball-Reference.com
NickS
Joined: 30 Dec 2004
Posts: 384
PostPosted: Sat May 19, 2007 3:52 pm Post subject: Reply with quote
Flint wrote:
You pose a question about how a sequence should be scored. Here is how I would do so.
My team misses shot = - 1
Opposing team gains rebound = 1
Opposing team misses shot = -1
My team gets rebound = 1
My teams net winscore = 0
Opposing teams winscore = 0
Points on the scoreboard - equal
What about a situation in which the other team hits their shot:
My team misses shot = - 1
Opposing team gains rebound = 1
Opposing team hits shot = +1
My teams net winscore = -1
Opposing teams winscore = +2
Points on the scoreboard - +2 for opposing team.
Unless you give your team a +1 credit for receiving the ball after an opponents made shot, this doesn't add up.
Back to top
View user's profile Send private message
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Sat May 19, 2007 4:00 pm Post subject: Reply with quote
Nick - its not supposed to add up. Winscore is not meant to equal the points on the scoreboard. I think thats what you mean right?
As I said before, winscore isn't really an instrument designed to work at the team level.
Back to top
View user's profile Send private message
NickS
Joined: 30 Dec 2004
Posts: 384
PostPosted: Sat May 19, 2007 4:10 pm Post subject: Reply with quote
Flint wrote:
Nick - its not supposed to add up. Winscore is not meant to equal the points on the scoreboard. I think thats what you mean right?
As I said before, winscore isn't really an instrument designed to work at the team level.
That doesn't make sense. I thought the appeal of Berri's system was that it was derived from the value events have to a team.
If that isn't the case, what were you trying to argue in the bit I quoted?
Trying to make sense of this, for a moment, it seems like any rating system is going to have similar problems with the difference between a TO that's counted as a steal for the other team, and a TO that that isn't. Assuming that you count steals as positive value, the former will result in a greater differential of value than the latter -- and it's just an artifact of the type of TO.
At the same time, it seems like the two most common ways for a possession to end are either a made basket or a miss and a rebound and that it would make sense for those two events to have equal value in a rating system.
This is a little scattershot, because I'm still thinking this through, but my intuition is to say that DLew's comment about double counting isn't precisely the problem (because of the analogy to steals above) but that something is a little odd here.
Back to top
View user's profile Send private message
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Sat May 19, 2007 4:18 pm Post subject: Reply with quote
MikeG wrote:
Quote:
29.2% is much closer to the Fibonacci/Almonte standard of 30%
It's not any Fibonacci, it's just the famous Rosembaum's sharing (I apologize by not give his credits), and one of his regressions obtained this "near to 70/30" link between missedFG and DReb.
My THEORY is just an extention. I'm talking about a longer chain of 70/30 links that begin with the assist at the offensive end, and finish with the DReb at the defensive end.
Back to top
View user's profile Send private message
NickS
Joined: 30 Dec 2004
Posts: 384
PostPosted: Sat May 19, 2007 4:21 pm Post subject: Reply with quote
Consider this example, a game of 1 on 1 between two players (so we aren't worried about the team aspect).
Player A shoots a 3-pointer on every possession and shoots 33.33%.
Player B shoots a 2-pointer on every possession and shoots 50%.
Both players get 20% of their own offensive rebounds.
At the end of 100 trips up and down the floor (possessions) player A has shot 115 3 pointers, hit 39, and gotten 15 offensive rebounds.
Player B has shot 111 shots, hit 56 of them, and gotten 11 offensive rebounds.
Player A's total Win Score:
117 points scored = 117
100 possessions used = -100
44 defensive rebounds = +44
15 offensive rebounds = +15
= 76
Player B's total Win Score:
112 points scored = 112
100 possessions used = -100
61 defensive rebounds = +61
11 offensive rebounds = +11
= 84
Player B has a higher total WS despite losing the game.
Tell me where that's wrong?
Last edited by NickS on Sat May 19, 2007 4:24 pm; edited 1 time in total
Back to top
View user's profile Send private message
NickS
Joined: 30 Dec 2004
Posts: 384
PostPosted: Sat May 19, 2007 4:24 pm Post subject: Reply with quote
Actually I think this is wrong:
NickS wrote:
Player A's total Win Score:
117 points scored = 117
100 possessions used = -100
44 defensive rebounds = +44
15 offensive rebounds = +15
= 76
Player B's total Win Score:
112 points scored = 112
100 possessions used = -100
61 defensive rebounds = +61
11 offensive rebounds = +11
= 84
This should be:
Player A's total Win Score:
117 points scored = 117
115 shot attempts = -115
44 defensive rebounds = +44
15 offensive rebounds = +15
= 61
Player B's total Win Score:
112 points scored = 112
111 shot attempts = -111
61 defensive rebounds = +61
11 offensive rebounds = +11
= 73
So the difference is larger than I first thought. But I may well be missing something else obvious.
Back to top
View user's profile Send private message
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Sat May 19, 2007 4:26 pm Post subject: Reply with quote
I was responding to Mike's post and trying to show how Berri's system worked differently from how he described it. I suppose it may have appeared that I was trying to show that winscore and the scoreboard should add up, but that is not the case and it wasn't my intention to suggest that.
There are quirky situations that you have to incorporate into the system. For instance, if a shot caroms of the rim and out of bounds. This results in a team rebound which is incorporated into the system as part of the infamous "team adjustment. " A turnover that doesn't correspond to a steal, for instance if someone throws to Eddy Curry and it bounces off his hands out of bounds, I suppose would be dealt with in the same manner I believe. I am not sure exactly, but either way, thats a relatively infrequent occurrence.
Let me know what you think is odd, will try to answer your question...
Back to top
View user's profile Send private message
NickS
Joined: 30 Dec 2004
Posts: 384
PostPosted: Sat May 19, 2007 4:37 pm Post subject: Reply with quote
Flint wrote:
Let me know what you think is odd, will try to answer your question...
Try the example I give above with player A and player B. They are identical in their eFG% and their ability to get offensive rebounds. If they play each other, on average Player A will win the game and player B will have a higher WS.
Doesn't this make it seem like something is being under/over valued? That WS is, in other words, a biased rather than unbiased estimate in that circumstance?
Back to top
View user's profile Send private message
NickS
Joined: 30 Dec 2004
Posts: 384
PostPosted: Sat May 19, 2007 5:33 pm Post subject: Reply with quote
NickS wrote:
This should be:
Player A's total Win Score:
117 points scored = 117
115 shot attempts = -115
44 defensive rebounds = +44
15 offensive rebounds = +15
= 61
Player B's total Win Score:
112 points scored = 112
111 shot attempts = -111
61 defensive rebounds = +61
11 offensive rebounds = +11
= 73
I think I'm getting somewhere here. Let's try this again with the following weights: Shot attempt (-.7), offensive rebound (.7), defensive rebound (.3), so that shot attempt and offensive rebound counterbalance, and shot attempt + defensive rebound = the cost of a lost possession.
In that case we would have:
Player A's total Win Score:
117 points scored = 117
115 shot attempts = -80.5
44 defensive rebounds = +13.2
15 offensive rebounds = +10.5
= 60.2
Player B's total Win Score:
112 points scored = 112
111 shot attempts = -77.7
61 defensive rebounds = +18.3
11 offensive rebounds = +7.7
= 60.3
At this point the two players have identical scores. Which is closer but still not perfect. But, if add a penalty to made shots equivalent to the value of a defensive rebound (to account for the fact that a made shot is a complete change of possession, rather than just .7 of a change of possession we would have.
Player A's total Win Score:
117 points scored = 117
115 shot attempts = -80.5
39 made shots = -11.7
44 defensive rebounds = +13.2
15 offensive rebounds = +10.5
= 48.5
Player B's total Win Score:
112 points scored = 112
111 shot attempts = -77.7
56 made shots = -16.8
61 defensive rebounds = +18.3
11 offensive rebounds = +7.7
= 43.5
The two players WS differ by exactly the difference on the scoreboard!
Obviously .7/.3 is arbitrary. The last method would work whatever weights were used as long as the value of an offensive rebound + defensive rebound equaled 1.
In fact, this would also (somewhat) work with the WoW forumla as long as the penalty for a made shot still equalled a defensive rebound (1 point in this case). That would have the odd effect, however, of claiming that a made 2 point basket had 0 value.
Ultimately I think DLew is correct that in having the sum of the value of an offensive rebound and the value of a defensive rebound equal 2 rather than 1 that a form of double counting is going on.
Back to top
View user's profile Send private message
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Sat May 19, 2007 5:38 pm Post subject: Reply with quote
Lol. That's complicated. Let me hack at it.
Each player on average should score one point for every shot he takes.
The way I make it Player A scores 113.32 points, B scores 110. This is because A rebounds 13.32 of his 66.6 misses. He takes 13.32 additional shots, and this translates into 13.32 additional points scored at his established rate. B would score 50 points on his first 100 shots, an additional ten from the ten offensive rebounds collected from his fifty misses.
What their winscore would be seems quite complicated, even though their scoring and shot attempts should cancel out.
A gets 40 rebounds, five defensive rebounds off of B's ten secondary shots, and 13.32 offensive rebounds. So 58.32 total rebounds?
B gets 53.28 misses off A's first shots, and another 8.79 rebounds off of the A's secondary misses. He also collects ten offensive rebounds. His rebounding's contribution to winscore is 72.07.
Total winscores
A. 113.32 + 58.32 = 171.64
B. 110 + 72.07= 182.07
I dont really know what that means. There is a difference of 10 there. I feel like I am missing something. I must have made a mistake. I dont know, check my work, let me know what you think, I am going to go ponder this.
Back to top
View user's profile Send private message
NickS
Joined: 30 Dec 2004
Posts: 384
PostPosted: Sat May 19, 2007 5:55 pm Post subject: Reply with quote
I haven't worked through your numbers yet, but here's something else to consider, that you should translate into the appropriate WS numbers as appropriate.
Imagine two teams A and B, both always shoot 100%, team A shoots only 3 pointers, team B shoots only 2 pointers.
Imagine they both go down the court 10 times and take ten shots.
Team A: 30 points - 10 possessions used = +20
Team B: 20 points - 10 possessions used = +10
At that point the difference in their WS is the same as the difference in points on the scoreboard (10 points).
Now imagine same two teams, same 10 trips up and down the court, but in this case each team takes 5 shots, and has 5 unforced TOs.
Team A: 15 points - 10 possessions used = +5
Team B: 10 points - 10 possessions used = +0
Again, at this point the difference in WS equals the difference on the scoreboard (5 points).
This reinforces my opinion that it isn't the difference between 2 and 3 point shots that's the problem, but that it's something about the way WS counts rebounds.
If difference in WS is equal to difference in points when you have no rebounds, but not equal when rebounds are involved, there's soemthing up with rebounding.
Again, please correct me if I'm mischaracterizing WS.
Back to top
View user's profile Send private message
NickS
Joined: 30 Dec 2004
Posts: 384
PostPosted: Sat May 19, 2007 6:44 pm Post subject: Reply with quote
NickS wrote:
f [we] add a penalty to made shots equivalent to the value of a defensive rebound (to account for the fact that a made shot is a complete change of possession, rather than just .7 of a change of possession we would have.
Player A's total Win Score:
117 points scored = 117
115 shot attempts = -80.5
39 made shots = -11.7
44 defensive rebounds = +13.2
15 offensive rebounds = +10.5
= 48.5
Player B's total Win Score:
112 points scored = 112
111 shot attempts = -77.7
56 made shots = -16.8
61 defensive rebounds = +18.3
11 offensive rebounds = +7.7
= 43.5
As a point of possible interest, I was trying to figure out what the scale on these numbers were. It seems like it should be, based on the theory, points scored above 1 point/possession.
but, if that were the case, the totals would be 17 and 12 since the two players score respectively, 117 and 112 points in 100 possessions.
I figured out that the scale is points scored at a rate above 1 point / possession + the value that's created by defensive rebounds.
As an easy way to see this, imagine two teams that shoot 0%, and don't get offensive rebounds playing 10 possessions.
Team A: 10 shot attempts (-7) + 10 rebounds (+3) = -4
Team B: the same.
If the value of an offensive rebound was set to 1 and the value of a defensive rebound was set to 0 than the two offensively inept teams would have a value of:
Team A: 10 shot attempts (-10) + 10 rebounds (+0) = -10
Team B: the same.
And the value would be exactly equal to the difference between points scored and possessions used.
This is just to say that the weights that are chosen for offensive and defensive rebounding will also act as a scaling factor on the final values.
Note, going back to the quoted example, that if you took each team's positive value for defensive rebounds and, instead, applied it as a penalty to the other team's score (to turn the value of a missed shot into -1/0 rather than -.7/.3) you would get:
Team A: 48.5 - 13.2 -18.3 = 17
Team b: 43.5 - 18.3 -13.2 = 12
Again, equal to the points scored above base efficiency.
So, there you go, that's what the numbers mean. Just something that had been bothering me.
Back to top
View user's profile Send private message
NickS
Joined: 30 Dec 2004
Posts: 384
PostPosted: Sat May 19, 2007 11:36 pm Post subject: Reply with quote
I've posted a lot on this topic, but I've continued to think about it and I want to summarize.
What I have shown is that, if Flint's calculations are correct, that there is a simple example of a case in which the win score calculation will rank the worse player above the better player.
Flint wrote:
Total winscores
A. 113.32 + 58.32 = 171.64
B. 110 + 72.07= 182.07
I want to emphasize that the example is not complicated. There are no assists, steals, blocks, or turn overs involved. It just involves Shots, made and missed baskets, and rebounds.
I believe that this example will be a problem for any linear weight system in which the sum of the values for an offensive and defensive rebound are greater than the value of a possession, and I think I can explain why.*
Take this quote from Flint earlier.
Flint wrote:
[Y]ou must understand where I am coming from. A shot attempt is a deduction against the shooter. This represents the cost of giving up possession and putting the ball in play.
consider the last phrase, "putting the ball in play". If we penalize someone who shoots and misses for having given up possession of the ball, that means that at the point that the ball is coming off the rim, from a WinScore perspective, neither team has possession of the ball. At that point the offensive team has given up possession and both teams are trying to gain possession of the ball.
It's obvious that, at that point, there are two possible outcomes, an offensive or defensive rebound, and that the difference between those two outcomes is one possession.
If the difference in credit that is available to the two teams is greater than the value of one possession, that is not an accurate description of the situation because what is at stake is, precisely, possession of the ball. It can't be worth more than that.
Flint, seriously, I appreciate your getting me to think through this example, it has really helped me clarify my thoughts.
*[Technical note: the other way to look at this, is where we started, that any system in which a missed shot and a defensive rebound combine for more than the value of one possession will have a problem. Because the (negative) value of a missed shot has to equal the (positive) value of an offensive rebound this means the same thing as a system in which the values of an offensive and defensive rebound add up to more than the value of a possession]
Back to top
View user's profile Send private message
NickS
Joined: 30 Dec 2004
Posts: 384
PostPosted: Sun May 20, 2007 11:20 am Post subject: Reply with quote
Yet another thought on this topic. This has now gotten me thinking about the possible goals of player ranking system.
First of all I want to say that the argument above does not show, by itself, that WinScore is a bad linear weights system. Compare, for example, to a system that gave players credits for points scored, and for nothing else. That system would pass the test of always having the difference in credit be equal to the difference on the scoreboard, but would be a much worse system.
This just shows, I think, that WinScore demonstrably overvalues rebounds, as many people have argued. I'm not a fan of WinScore for other reasons, mostly because of the arguments Dan Rosenbaum has made which I find convincing, but this is a demonstration of what errors does one accept in using WinScore as a valuation.
I'm inclined to believe that any linear weight system will have errors, and the value of the system depends, in part on recognizing when those errors occur and being able to take that into account.
So it got me thinking about, descriptively what different linear weight systems attempt to accomplish. That will be a long post, and I may make it a separate topic.
But, in brief, a linear weight system is an attempt to distribute credit, and the two questions to be asked are (1) how much credit / debit is there to be distributed and (2) how should it be distributed (for example, if you have two teammates one of whom takes most of the shots, and the other who gets most of the rebounds how do you share credit for scoring? The shooter wouldn't have possession of the ball without the rebounder getting rebounds, but the rebounder would, presumably, have a harder time turning possessions into points without the shooter).
This series of posts got me thinking about the first question -- how much credit is there to be distributed. PER is closer to saying that the total amount of credit is equal to the total number of points scored -- regardless of how much the opponent scores. WinScore is closer to saying that total credit is the amount by which the team would outscore an average opponent. Two very different idea of "credit."
This just get me thinking again about the value of trying to identify the "marginal" or "replacement" (two different quantities) scoring rate and assinging credit for actions that help a team score more efficiently than the marginal rate.
For example, if you thought the replacement scoring rate was .7 points/possessions the weights could look like this (with a .7/.3 split between offensive and defensive rebounds):
Made basket: 2 or 3 points
Offensive rebound: .49 points
Defensive rebound: .21 points
Shot attempt: -.49 points
Made basket: -.21 points
Back to top
View user's profile Send private message
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Sun May 20, 2007 1:58 pm Post subject: Reply with quote
Nick - The problem here is that Win Score, as noted before, is just a quick and dirty metric. It omits team defense. That's why you observe an anomaly. The more complete Wins Produced metric, does incorporate team defense. And if you look at it through Wins Produced, both player's "teams" do come out equal.
Re: PER vs EFF
NickS
PostPosted: Sun May 20, 2007 3:33 pm
Flint wrote:
Nick - The problem here is that Win Score, as noted before, is just a quick and dirty metric. It omits team defense. That's why you observe an anomaly. The more complete Wins Produced metric, does incorporate team defense. And if you look at it through Wins Produced, both player's "teams" do come out equal.
How does that work?
Team B has a defensive rating that is 3% worse than team A, but their score is ~6% higher. That implies that the adjustment is greater than the difference in their defenses.
Back to top
View user's profile Send private message
NickS
PostPosted: Sun May 20, 2007 3:51 pm Post subject:
Based on all of this I did a quick and dirty player rating based on points generated above a baseline scoring efficiency and it's interesting. I just through this together, so I will double check, but here are the top 10 on a per/minute basis for a couple of values of a possession.
It's interesting to see how usage becomes more important, compared to keeping and getting possessions as you change the baseline rate.
Looking at the numbers, I may be undercounting the cost of a TO (I'll double check that), but the lists are plausible. FWIW, David Lee (the WoW MVP) clocks in at #11 on the first list.
Possession = 1 point:
1: nash,steve
2: wade,dwyane
3: duncan,tim
4: camby,marcus
5: nowitzki,dirk
6: ginobili,manu
7: ming,yao
8: stoudemire,amare
9: garnett,kevin
10: gasol,pau
Possession = .8 points (so credit for scoring at better than .8pt/possesion)
1: wade,dwyane
2: nash,steve
3: bryant,kobe
4: ming,yao
5: nowitzki,dirk
6: arenas,gilbert
7: duncan,tim
8: ginobili,manu
9: stoudemire,amare
10: garnett,kevin
Possession = .65 points
1: wade,dwyane
2: bryant,kobe
3: arenas,gilbert
4: ming,yao
5: nash,steve
6: nowitzki,dirk
7: anthony,carmelo
8: ginobili,manu
9: mcgrady,tracy
10: james,lebron
Back to top
View user's profile Send private message
NickS
PostPosted: Sun May 20, 2007 4:10 pm Post subject:
One last note (and I do hope that anyone who isn't interested has stopped reading at this point). The rankings above are not pace adjusted which helps Nash and Camby. Also, just for fun I calculated a top 10 if you value possessions at 1.15 points and the rebounders and shot blockers shoot to the top at that point.
1: camby,marcus
2: mutombo,dikembe
3: lee,david
4: nash,steve
5: duncan,tim
6: chandler,tyson
7: okafor,emeka
8: marion,shawn
9: kidd,jason
10: millsap,paul
Back to top
View user's profile Send private message
Flint
PostPosted: Sun May 20, 2007 4:26 pm Post subject:
You have to subtract the value of made defensive fgs from each team's totals. A made fg is worth . .33. A made 3 point fg is worth .066. (see page 103 of the WOW) If you do so the numbers work. It's different from Wins Score, and more accurate.
Back to top
View user's profile Send private message
NickS
PostPosted: Sun May 20, 2007 5:20 pm Post subject:
Thanks, so let me run those numbers:
Player A
39 3 pointers made: +2.574
56 2 pointers given up: -1.792
44 Defensive Rebounds: +1.496
15 Offensive Rebounds: +.51
76 missed Field Goals: -2.58
Total: .204
Player B
56 2 pointers made: +1.848
39 3 pointers given up: -2.574
61 Defensive Rebounds: +2.074
11 Offensive Rebounds: +.374
44 missed Field Goals: -1.496
Total: .226
I still have player B ranking 10% better than player A.
That's before team adjustment, but using the weights on page 103.
Back to top
View user's profile Send private message
Flint
PostPosted: Sun May 20, 2007 6:21 pm Post subject:
Your example is too complicated.
Let's assume each players has twelve possessions and there are no offensive rebounds. That is throwing things out of whack. I don't exactly know how he calculates wins produced, but if i were to take a crack at it it would look something like this.
Player A (two point shooter)
Two point fgm 6 x .33 = .198
D. Rebounds 8 x .034= .272
FGs Missed 6 x .034= -.204
3FGs allowed 4 x .066= -.264
Net .002
Player B (Three Point Shooter)
3 pt Fg made 4 x .066 = .264
DRB 6 x .034 = .204
FG missed 8 x .034 = -272
2 fg allowed 6 x .032 = .-192
Net .004
They dont quite add up to zero, I would assume this is some sort of rounding error. But basically, it definitely works out. The players performance is identical, as you would expect.
I dont really feel like tackling your example right now. Clearly, Wins Produced is going to come out clean any way you slice it. There may be fatal flaws in his approach, but the numbers adding up is not going to be it. I dont think he gets the book published or an economics professorship without being able to get that right.
Vis a vis rebounding being overvalued, Berri had a long post at this link
http://dberri.wordpress.com/2006/11/09/ ... -rebounds/
He finds that even if he reduces the value of a rebound to .7 of a point his results are almost exactly the same. This shows, he says, that the difference between his system and others is entirely related to how it accounts for scoring and shooting efficiency.
I saw Paul Millsap on one of your lists. Berri looked at his college numbers and predicted he would be one of the best rookies. Turned out to be very prescient, although he thinks that either Roy or Rondo are most deserving of the Rookie of the Year award.
Back to top
View user's profile Send private message
NickS
PostPosted: Mon May 21, 2007 12:14 am Post subject:
Flint wrote:
Vis a vis rebounding being overvalued, Berri had a long post at this link
http://dberri.wordpress.com/2006/11/09/ ... -rebounds/
He finds that even if he reduces the value of a rebound to .7 of a point his results are almost exactly the same.
This is an interesting point. I went back and re-read the Dave Berri post and realized that it says something slightly different than what you describe and that it actually supports the opposite interpretation.
You say that changing the value of a rebound from 1 to .7 produces "results [that] are almost exactly the same." Dave Berri writes "The correlation between Win Score and the Kaufman inspired formula is 0.99"
If you have a number that is made of a variety of components and you change the value of one of the components and get a new number that correlates exactly with the original number that means one of two things. Either the factor for which you have changed the weighting is insignificant, and the new number is the same as the original number, or the factor completely dominates the rating and the new number is scalled to the orginal number by the amount which you have changed the valuation of that factor.
Given that we know that the total value of scoring among all players is 0. It seems very likely that for many players rebounding is the major component of their win score.
So this suggests that the second scenario (rebounding dominates win score) is more likely than the first scenario (rebounding is insignificant to win score).
So the fact that the correlation is extremely close to 1 after changing the value of rebounding is probably actually a sign that rebounding completely dominates the metric, rather than a sign that it is insignificant.
I never noticed that before.
Back to top
View user's profile Send private message
Mike G
PostPosted: Mon May 21, 2007 7:28 am Post subject:
I confess that I haven't poured my whole soul into trying to follow the zeroes and ones that make up this accounting system. But it's intuitively bizarre that "... the total value of scoring among all players is 0."
And if so, it's doubly bizarre that 'the total value of rebounding' would not be zero.
So I refer back to my much earlier question: If missed shots are assigned a deficit in these formulas, why aren't 'ungotten' rebounds also considered to be a negative?
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Harold Almonte
PostPosted: Mon May 21, 2007 10:36 am Post subject: Reply with quote
Quote:
Player A (two point shooter)
Two point fgm 6 x .33 = .198
D. Rebounds 8 x .034= .272
FGs Missed 6 x .034= -.204
3FGs allowed 4 x .066= -.264
Net .002
Player B (Three Point Shooter)
3 pt Fg made 4 x .066 = .264
DRB 6 x .034 = .204
FG missed 8 x .034 = -272
2 fg allowed 6 x .032 = .-192
Net .004
The only thing I can see here is: scoring totals are more important than efficiency, rebounds and TOs. The player B tied the game having less efficiency, less rebounds and more missed FGs. You can say, It's three points that have an extra point prize. Well, bad luck scoring 3p is a more valuable skill.
The same if you have an "unskilled" but careful worker A who only produces 9 articles with just 1 mistake a whole day of labor; and a "skilled" and experienced worker B who produces 30 with 7 mistakes the same day. Who give more money to the company? The "no mistake" or the "productive"?
You can say a DReb earns the possession, but did the DReb avoid the scoring?. It's DReb (1 poss. prize) the counterpart of missFG (-1 poss. punish)?. Even if the offensive team score, you'll also have your possession. The problem with WS is that DReb. suposedly did all the defense...This is old stuff.
Back to top
View user's profile Send private message
NickS
PostPosted: Mon May 21, 2007 11:00 am Post subject:
Harold Almonte wrote:
The problem with WS is that DReb. suposedly did all the defense...This is old stuff.
Actually, I'm arguing that, separate from any dispute over how WS assigns credit, that it is generating the wrong amount of credit to assign.
Think about this example, if we imagine the game of 1-on-1, the same player is forcing the miss and getting the rebound, so there isn't any problem with the assignment of credit. The problem is that the worse player accumulates more credits than the better one, showing that there's a problem with how credits are generated.
Back to top
Flint
Posts: 112
PostPosted: Mon May 21, 2007 11:04 am
Nick -
NBA Efficiency and the WOW value points scored and rebounds exactly the same. In each system a point scored and a rebound collected have the same value. If rebounding "dominated win score," then it should dominate NBA Efficiency also, and the two ratings should be basically identical. However, they are not. The reason they are not is that they treat shooting efficiency much differently. This would indicate to me that how you account for scoring is a major part of the player rating, probably more important than rebounding.
If you have read the book, you know that the major point of it is that scoring is wildly overrated in the NBA relative to its contribution to wins. Players are paid huge salaries and given all star and post season honors and shoe contracts on the basis of how many points they score, with little regard for shooting efficiency or their other contributions, or lack therof in other areas of the box score.
Given that this is the case, I find it surprising, to say the least, that you think shooting efficiency doesnt matter at all in Berri's rating and that rebounding is the only thing that counts.
Mike - I think what you mean is, why aren't you deducted for allowing an offensive rebound? In the same way that a shooter costs his team a possession (unless his team recovers it), doesn't someone cost the defensive team a possession when an offensive rebounding is allowed? I would say to that perhaps it comes out in the wash in Wins Produced when you deduct for opponent's made field goals. The "cost" of allowing this offensive rebound is spread throughout the team through the mechanism of reduced defensive efficiency. The "cost" of not capturing that rebound is the points scored off the offensive rebound, and this cost is deducted from each player. Would that make it even out in your view?
Back to top
View user's profile Send private message
Harold Almonte
PostPosted: Mon May 21, 2007 11:33 am
Quote:
The "cost" of not capturing that rebound is the points scored off the offensive rebound
How can we assure an offensive rebound will be automatically converted in score?
The only thing a DReb avoids is the other team don't rebound back the ball (and does that only 70% of the time). It doesn't avoid scoring, "defense around" avoided scoring, but then you come and give the full possession value to the rebound and even you add a team defense adjust. You are overrating defense. And I'm not denying frontcout "might" have a higher defensive value then backcourt. And even this need to be wide proved.
One metric is offensive biased, the other is defensive (rebound) biased. The utility of NBAeff. is none, PER is the usage concept, and WP the team win fit, but all of them are just "almost credible"
Back to top
View user's profile Send private message
NickS
PostPosted: Mon May 21, 2007 12:11 pm
Flint wrote:
Given that this is the case, I find it surprising, to say the least, that you think shooting efficiency doesnt matter at all in Berri's rating and that rebounding is the only thing that counts.
That's not what I'm saying, exactly. I was arguing that the Rebounding component of the WinScore/WinsProduced metric is the major source of variability in player rankings.
(noted that this is the simpler metric)
Points + Rebounds + Steals + ½Assists + ½Blocked Shots – Field Goal Attempts – Turnovers - ½Free Throw Attempts - ½Personal Fouls
There are no steals, assists, blocked shots, turnovers, free throws or fouls.
So that is
Points + Rebounds - Field Goal Attempts
You can divide that into a scoring portion (Points - Field Goal Attempts) and a rebounding portion (rebounds).
Consider our example of player A and player B:
Player A: Scoring (117 - 115) Rebounding (49)
= 2 + 49
Player B: Scoring (112 - 111) rebounding (72)
= 1 + 72
What you can see is that the rebounding portion of the total is much larger than the scoring portion. This means that if you have or double the value of rebounding it will correlate extremely well with the original.
2+ (49 / 2) ~ (2+49)/2
dividing the value of rebounding in half correlates well with just dividing the total score in half.
I'm saying that the high level of correlation that Berri mentions in that post means that the scoring potion (including positive value for points scored and negative value for possessions used) is being swamped by the rebounding portion.
Back to top
View user's profile Send private message
Mark
PostPosted: Mon May 21, 2007 4:29 pm
Mike G wrote:
And if so, it's doubly bizarre that 'the total value of rebounding' would not be zero.
So I refer back to my much earlier question: If missed shots are assigned a deficit in these formulas, why aren't 'ungotten' rebounds also considered to be a negative?
If the rating system is intended to represent both offense and defense fully then I agree the double entry accounting for rebounding (credit/debit) to each team should sum to zero- in the end. A defensive rebounding departure from average should get dealt with in the team defense adjustment. Perhaps some of the value of those defensive rebounds gets backed out? Seems like it has to to get back to balance.
To some extent "ungotten" rebounds can have an impact on position adjusted standings- if you get less than average it will affect your score relative to so called peers.
Looking at individual rebounding relative to position expectation and/or actual counterpart data and plugging it directly into WP score could improve ratings and reduce size of team adjustrment. Individual shot defense is another step that could help get to full double entry accounting completeness at player level but the quality of the database gets raised by some as a stumbling block as discussed before.
Last edited by Mark on Mon May 21, 2007 7:46 pm; edited 6 times in total
Back to top
View user's profile Send private message
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Mon May 21, 2007 5:32 pm
The only thing that can be done with linear formulas is: to shortcut or to make the team adjust (but not to do both things at the same time, then you are cheating).
We're talking about give credits of 2/3 or .75 to FGA and DReb. Give similar credit to the OReb. But, how do you punish scoring teammates by not offensive rebounding (their 30%) if you fail to score?. How do you punish defense by being scored?. How do you spare the supposed 0.25 of defense around a DReb?. All those 1/3s must necessearily be adjusted at both end. That's the best justice.
I don't like any linear formula, I don't see the big importance of an overall season performance rating (the team-win-$ fit of WP only-they don't want to repair). I would prefer another kind of "scouting point of view" metric where you define some skill-block and you use the block of metrics (almost like PER but with defense) you thing best qualify those skills. How would you weight those skill-block? 4 factors would be the guide, I think you don't need more regression than that. But even with this metric, the boxscore wouldn't help with somethings like screen-assists, potential assists, getting free, counted blocks, counted deflections, counted double teams, charges, zones, etc.
I would add emotionals factors to this metric: command and leadership, self esteem, egocentrism.
page 6 missing
jeffpotts77
PostPosted: Wed May 23, 2007 1:11 pm
Flint wrote:
You referenced numbers, but I don't think they show convincingly that Yao is better. A poster at the WOW broke down their numbers in a comment. Here is what he said:
Quote:
on a per 48 minute basis:
blocks : tied at around 3
turnovers: 1.5 for dikembe, 5.4 for yao
assists: .5 for dikembe, 3 for yao
rebounds: 18.5 for dikembe, 13.3 for yao
fg% 56.6 for dikembe, 51.5 for yao
personal fouls: 5.5 for dikembe, 5.0 for yao
So it appears that dikembes main advantages are in turnovers, and rebounds, and field goal percentage.
Your point about TS% is well taken. Yao obviously is the better scorer since he scores more pointa with the same efficiency. But Mutombo is a MUCH better rebounder than Yao, and he commits many fewer turnovers. Over 48 minutes Deke was worth nine more possessions. That's a big difference. I understand your point that the reason Mutombo doesn't commit turnovers is that he is a low usage player, and that actually he has a higher TO rate. But I am not sure that really translates into an impact on wins. The implication of Berri's work is that there is always room for at least one very low usage player out there. You can't play with five of them, but having one of them out there is a great thing. For this genre of ball player, the defensive center, Mutombo is the absolute prototype (well Bill Russell maybe) and Chandler is the newest version.
For every low usage player included in a lineup, the other players must absorb/use extra possessions. The fact that Yao can use so many possessions so efficiently, with a lower turnover rate and a higher TS%, makes him much more valuable than Deke, even with the disparity in their rebound rates. You keep harping on the fact that the Rockets were better with Yao out of the lineup but this argument ignores the fact that this event did not occur in a vacuum. So many things could have occured when Yao went on the shelf that can't be found in the boxscores. Was Tracy McGrady playing with more fire during this stretch? Were various players (other than Deke) playing with more intensity on the defensive end? Was their schedule easier during this stretch? Was this just a case of too small a sample size? I'm sure there are more variables that I haven't thought of. You cannot draw the (seemingly absurd) conclusion that Deke is better than Yao without first addressing those other possibilities.
Back to top
View user's profile Send private message Yahoo Messenger
Flint
PostPosted: Wed May 23, 2007 1:32 pm
Justin - Well, I object to the fact, as a WOW fan, that some people, Mike G I am looking at you, think its a weaker metric than EFF.
I would say there are two very good reasons to discuss NBA Efficiency. First, it's the official metric of the league. Second, money.
Berri has some cool stuff about salaries in th NBA. The controversy over his Wins Produced metric has obscured his (at least to me interesting) findings about compensation. I know a lot of studies have already been done. But He found he "could explain 63% of player salary when we employed the NBA's performance metric," while Wins Produced only explains 47% of salary. I wonder what the number is for PER.
He finds that points scored works a little bit better than NBA efficiency, at 63.5% and that:
"
Quote:
if we add a player's accumulation in the past season of rebounds, steals, blocked shots, assists, turnovers, personal fouls, and shooting efficiency," our explanatory power rises to 64%. Of all these statistics, only blocked shots were found to significantly impact compensation. And like our study of the All-Rookie team, the number of points scored has the largest impact on player salary. Furthermore, increases in rebounds, steals, assists, turnovers, personal fouls, and shooting efficiency on both field goals and free throw attempts were not found to have significant impact on player salary."
The basic take away is that "player salary in the NBA is dominated by how many points a player scores. And it does not seem to matter much if the player scores efficiently, or if he rebounds, generates steals, or avoids turnovers. Basically if a player can score, he will score a major payday. One should note that this same pattern is also seen in the coaches’ voting for the All-Rookie team."
I know others said that before Berri, but it was shocking to me.
Back to top
View user's profile Send private message
Statman
PostPosted: Wed May 23, 2007 1:48 pm
Wow - alot to respond here - let me give it a go:
1.
Quote:
Ok. You dont think Berri's evaluation of Mutombo passes the laugh test.
[quote="Flint"]
Quote:
Seriously - not a single objective (or even casual) observer or APBRmetric stats freak would EVER dare to say Dikembe was more valuable than Yao. Yet - Berri's numbers seem to say this. HUGE red flag for me.
It certainly does go against the grain. But you have to give the guy credit. He doesn't turn tail when the data coughs up strange results, he embraces these quirks. He posted on Mutombo-Yao a while back, and in my mind it was one of his bests post.
http://dberri.wordpress.com/2007/03/08/ ... mutombo/re
If you dont think Deke was as good as Yao this year, the burden is on you to explain how the Rockets had a better record without Yao after his injury, than they had with him before. Really, your laugh test conclusion is actually difficult to defend , given this fact. I posed this question on a previous thread I think, but its a great little puzzle. How do you explain it? How did the Rockets win more games without Yao?
Couldn't find that article. Anyway - WHERE do you get this info about the Rockets having a better record without Yao? The Rockets were 20-14 without Yao Ming,. They also lost the only game he played less than 19 minutes (6 minutes AT HOME against the LA Clippers). Only half those 34 games were versus playoff teams - and TWENTY of those games were at home. YET - they still were MUCH WORSE record wise without Yao than with him (20-14 to 32-16) despite favorable conditions (more home games) without him.
So - the Rockets were much better per possession when Yao was on the court than when he when he wasn't - and were slightly WORSE when Dikembe was on the court than when he wasn't. The Rockets were much worse record wise w/o Yao than with him considering the conditions. Is that good enough for the "burden of proof"?
2.
Quote:
And all this doesn't mean Mutombo is the better player right now to have on your team. If I were a GM and I had the choice between Mutombo and Yao, of course I would choose Yao. But it would largely because of future potential and his ability to play more minutes. Yao also was much better last year than this year in terms of his WOW numbers. But I would certainly rather have Mutombo in his prime than Yao and the performance he has provided so far in his career.
Uh - what would Mutambo's WoW FOR THIS YEAR have ANYTHING to do with Mutambo in his prime? Mutambo isn't CLOSE to the player he was in his prime - and he's nowhere CLOSE to the player Yao is now. But - Berri has him "rated" better.
3.
Quote:
You referenced numbers, but I don't think they show convincingly that Yao is better. A poster at the WOW broke down their numbers in a comment. Here is what he said:
Quote:
on a per 48 minute basis:
blocks : tied at around 3
turnovers: 1.5 for dikembe, 5.4 for yao
assists: .5 for dikembe, 3 for yao
rebounds: 18.5 for dikembe, 13.3 for yao
fg% 56.6 for dikembe, 51.5 for yao
personal fouls: 5.5 for dikembe, 5.0 for yao
So it appears that dikembes main advantages are in turnovers, and rebounds, and field goal percentage.
Your point about TS% is well taken. Yao obviously is the better scorer since he scores more pointa with the same efficiency. But Mutombo is a MUCH better rebounder than Yao, and he commits many fewer turnovers. Over 48 minutes Deke was worth nine more possessions. That's a big difference. I understand your point that the reason Mutombo doesn't commit turnovers is that he is a low usage player, and that actually he has a higher TO rate. But I am not sure that really translates into an impact on wins. The implication of Berri's work is that there is always room for at least one very low usage player out there. You can't play with five of them, but having one of them out there is a great thing. For this genre of ball player, the defensive center, Mutombo is the absolute prototype (well Bill Russell maybe) and Chandler is the newest version.
INTERESTING. Let's see - you completely don't include points. You use FG% INSTEAD of TS% - which incorporates getting to the line and hitting free throws - something Yao does much better than Dikembe.
Here are the COMPLETE stats - per 48 minutes:
TS% - exactly the same 60.1%
Points: Yao 35.5 to Dikembe 8.6 (Yao is 4.16 TIMES better)
Rebounds: Yao 13.4 to Dikembe 18.1 (Dikembe is about 1.36 times better)
Assists:: Yao 2.8 to Dikembe 0.5 (Yao is 5.74 times better)
Steals: Yao 0.5 to Dikembe 0.8 (Dikembe is 1.63 times better)
Turnovers: Yao 4.9 to Dikembe 1.4 (Dikembe is 3.49 times better)
blocks: Yao 2.8 to Dikembe 2.8 (Dikembe is 1.02 times better)
You call Dikembe MUCH better a rebounder. Well - Yao is three times a better SCORER (btw - scoring points still wins games) than Mutambo is a better rebounder. Yao turns the ball over much more - but garners assists and points at a MUCH higher rate the he commits turnovers in relation to Dikembe (can't have all that good without some bad).
Look at those stats. Now - ALSO consider that Yao's +/- is MUCH better than Dikembe's - and Houston was MUCH better when Yao started than when Dikembe started - do you actually THINK it's even a comparison?
In 90-91 - Mark Eaton & Michael Jordan had almost identical TS%. Eaton was a MUCH better rebounder, INFINITELY better shot blocker, turned the ball over MUCH less. Would anyone even remotely consider Eaton better than Jordan? Well - considering Dikembe better than Yao is almost (but not quite) on par to this statistically.
4.
Quote:
Vis a vis Eddy Curry, I have a visceral hatred for the man's game. Having watched him play most of this year, I wouldn't hesitate to say that its far more logical to call David Lee one of the best performers in the league on a per minute basis, than to call Eddy Curry an above average player. Berri doesnt actually rate Curry as slightly below average. He rates him way below average. In Berri's system, .100 is the average player, so Curry's .002 suggests that Curry contributed almost nothing in terms of wins this year, despite playing 35 minutes per.
I love this btw:
Quote:
His own coach didn't see fit to play him more than Curry (who to you was OBVIOUSLY horrible in every facet of the game besides shooting).
His coach is Isaiah Thomas!
Curry was "obviously horrible" in every facet of the game other than scoring. He was actually historically bad. He assist to turnover ratio was ..24.And he collected less than twice as many rebounds as he committed turnovers. He was second in the entire league in turnovers! He had more turnovers than Steve Nash. And he is one of the worst rebounding centers in the league. For a center, those are fairly historic numbers. He blocked fewer shots than Renaldo Balkman and his off box score defense could charitably be described as lousy.
What the numbers say to me is that Eddy Curry is one of the worst centers playing more than 30 minutes in the NBA. I can think of 20 players at the position who I would rather see in a Knicks uniform. There is no way he even approaches average, unless you consider, as PER and EFF do, scoring to be much the most important thing that players do on a basketball court.
David Lee is the man. He wasn't "well above average" in rebounding, he was, with Tyson Chandler, the best rebounder in the league per minute.
OK - you hate Curry (and Iverson from previous posts) - so it's understandable you latch onto this as "proof" Berri's system is on point. I understand. However - PER, Eff., and pretty much ALL rating systems have David Lee as a bit to much better than Curry per minute. None of them would have Lee (playing less than 28 minutes a game on a BAD team) as the BEST in the league - like Berri does. I agree - I don't think Curry should rate above average - BUT to denounce completely other systems because they do yet latch on in support of a system that has DAVID LEE as the best in the league per minute and DIKEMBE MUTAMBO as better than Yao Ming seems to be a HUGE leap of faith just because the system better fits your view of players like Curry and Iverson.
BTW - David Lee was 5th in rebounding per minute (not 2nd) - behind Evans, Mutambo, Chandler, & Foster. I'll be willing to bet ALL those guys rate VERY highly in Berri's system per minute - because the system obviously has a huge bias towards rebounders who don't ATTEMPT much of anything else. If you rebound alot and never take (thus possibly miss) a shot and never touch the ball (thus never turning the ball over) because you are an offensive zero - you will rate in the top 30 in the league.
5.
Quote:
Important difference. His Ts% was second in the league to Steve Nash. True, he didnt score much, but that's irrelavant and his 14.4 per 40 really isnt all that bad. Curry, Marbury, and Crawford dominated the ball. I can barely remember any plays being called for Lee. But having watched him closely, he seems like the kind of player who is more than low usage. One thing is for sure. Before he went down with injury, the Knicks had actually posted a respectable number of wins. This had nothing to do with Curry, and everything to do with Lee. Curry's numbers are exactly what you would expect given his previous performance, he just played more minutes this year.
I dont think its a pipe dream to imagine Lee next year playing 37 minutes, averaging 19 and 12 on 55% fg shooting, committing less than 3 turnovers per game, and chipping in 2.5 assists, a steal and block per. Those are certainly all-star numbers, even on a goofy looking white guy, and to me if he achieves those number he will have to enter the discussion of who the best players in the league are. Sound crazy? Does that pass the laugh test?
Well - to think David Lee THIS SEASON was the best player in the league per minute is definitely a pipe dream. You love the kid - we all do (not as much as you though) - but a very good rebounder who shoots quite efficiencently (while taking few shots) does not an All NBAer make. I'm glad Berri helps you feel he's at or near the top of the league - but the truth is he still was not a completely a full time player on a BAD team. He was the best of the Knicks per minute without a doubt - but anything beyond that is highly suspect I think (especially near - say - top 10 in the entire NBA).
Since Berri overvalues rebounds, as well as somewhat overpenalizes turnovers and definitely overpenalizes missed shots - David Lee fits PERFECTLY in this WoW.
BTW - don't get TOO excited about Lee's future until you take a look at Danny Fortson's early years (much better scorer and rebounder per minute than Lee - and I am CERTAIN would have rated at or near the top of the league in WoW per minute). He hasn't really made any all star teams.
_________________
Dan
My current ratings (go to the last page for the most recent stuff):
http://www.pointguardu.com/f136/statmans-ratings-56243/
Back to top
View user's profile Send private message Send e-mail
Statman
s
PostPosted: Wed May 23, 2007 1:53 pm
jeffpotts77 wrote:
For every low usage player included in a lineup, the other players must absorb/use extra possessions. The fact that Yao can use so many possessions so efficiently, with a lower turnover rate and a higher TS%, makes him much more valuable than Deke, even with the disparity in their rebound rates. You keep harping on the fact that the Rockets were better with Yao out of the lineup but this argument ignores the fact that this event did not occur in a vacuum. So many things could have occured when Yao went on the shelf that can't be found in the boxscores. Was Tracy McGrady playing with more fire during this stretch? Were various players (other than Deke) playing with more intensity on the defensive end? Was their schedule easier during this stretch? Was this just a case of too small a sample size? I'm sure there are more variables that I haven't thought of. You cannot draw the (seemingly absurd) conclusion that Deke is better than Yao without first addressing those other possibilities.
Exactly - not to mention (as I just did the post before) that Houston was WORSE in EVERY way you look at it without Yao - against weaker competition and a higher percentage of home games. I don't know where this "better record without Yao" came from.
_________________
Dan
My current ratings (go to the last page for the most recent stuff):
http://www.pointguardu.com/f136/statmans-ratings-56243/
Back to top
View user's profile Send private message Send e-mail
Kevin Pelton
Site Admin
PostPosted: Wed May 23, 2007 2:27 pm
Flint wrote:
The basic take away is that "player salary in the NBA is dominated by how many points a player scores. ..."
This conclusion is not as strong, in my opinion, as Berri makes it out to be. Keep in mind that he is using total points instead of points per minute, so playing time plays a role in this correlation as well.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Gordon
PostPosted: Wed May 23, 2007 10:51 pm
Quote:
Nick - Its not rebounding that makes the difference. Its the difference in the way the WOW accounts for shooting efficiency. Rebounding is not overvalued. Its valued the same way as it is by Hollinger and NBA Efficiency.
There is no debate here -- you are both right (or wrong!). If a metric undervalues scoring, then it must overvalue rebounds (and vice-versa). What matters here is the RELATIVE weight each variable has in the metric -- how much influence does it have on the ranking of players? If scoring is rewarded less in one metric, but rebounds the same, then as a proportion of total player value points will be less important AND rebounds more.
Flint: consider a simple experiment: take PTS, FGA, and FTA entirely out of the WS formula. It would still be true, as Berri notes, that the formula differs only in its handling of points, not rebounds. But the new formula would indeed overvalue rebounds, in the sense that they would essentially determine player rankings all by themselves. So clearly, changing the values for scoring variables can impact the relative weight of rebounds in a metric.
Now, does WS overvalue rebounds/undervalue scoring? Clearly, Berri is right to value efficiency. But all metrics give more value to higher-efficiency shooters. The question is whether it is correct to use average NBA performance as the baseline for scoring (avg = 0), while using zero as the baseline for rebounding. Essentially, WS assumes that being an average NBA shooter and making roughly half your shots has no value at all to your team -- if you don't shoot it, someone else will -- but being an average NBA rebounder -- which means pulling down half the missed shots -- has huge value. The reason this is done is presumably because there are no "missed rebounds" in the box score. But that doesn't make it right to consider opportunities consumed (FGAs) in one case, but ignore the issue when it comes to rebounding.
Consider that for a statistically average player, in WS virtually none of his value derives from scoring, while almost 100% of his value derives from rebounding. Does anyone believe that the ability to take shots and score at league average efficiency is an ubiquitous skill that has no value? Does that seem plausible or reasonable?
Back to top
View user's profile Send private message
Mike G
PostPosted: Thu May 24, 2007 6:45 am
Gordon, thanks for elucidating the question I was asking on page 1.
WoW apparently supposes that if I don't take this 45% shot, someone else on my team will, by default, get a 50% shot. Then if I don't go for this rebound that I have a 50% chance to get, will a teammate automatically appear with a 70% chance to get it?
Both of those assumptions are absurd, but WoW seems to make one and exclude the other. Note that Flint, when asked about these absurdities, tends to reply along the lines of "But scoring is overvalued"; as if that justifies anything.
It may really be about the notion that average play = zero contribution. It makes more sense to me to say that the 20 RebRt guy is twice the rebounder as the 10 RebRt guy. Not that he's infinitely better (10 being average). Nor is he contributing 5x as much as the 12 RbRt player.
As long as scoring has no value (or negligible values both sides of zero), then Eddy Curry just makes turnovers and occasionally rebounds. If scoring had value, then he'd be a go-to scorer who rebounds and has too many TO. And that's what he is.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Thu May 24, 2007 8:22 am Post subject: Reply with quote
Quote:
if you don't shoot it, someone else will -- but being an average NBA rebounder -- which means pulling down half the missed shots -- has huge value. The reason this is done is presumably because there are no "missed rebounds" in the box score. But that doesn't make it right to consider opportunities consumed (FGAs) in one case, but ignore the issue when it comes to rebounding.
Quote:
WoW apparently supposes that if I don't take this 45% shot, someone else on my team will, by default, get a 50% shot. Then if I don't go for this rebound that I have a 50% chance to get, will a teammate automatically appear with a 70% chance to get it?
I can't understand why if you prize a DReb, punish the missed FG automatically, but when you get an OReb, who do you punish at the defensive end? Why do you rate scoring but no defense?
I think, but I don't have the skills to prove it, that WOW can be very good predicting (with luck help) no prolific rebounders players trades (Miller-A.I.), but have troubles when a lot of rebounds are in the mid (B. Wallace). After all a player's total rebounds depends on his teammates's rebounding ability.
O.K. you would say, there's no missed Rebs in boxscore, I will make up with a team defense adjust. But, why don't you cut also missed FG and make up with a team offense adjust, and let's see what you'll get? Don't you say Reb=1 or 0.75 poss. and missedFG = 1 or 0.75 poss?, well, then 50-50.
Scoring is a matter of 5 people, although just one shoots the ball, in the end is supossed to be a (70/30) relationship between the usager and his teammates. The same, rebounding is a matter of 5 people, although in the end just one can catch the ball.
P.D. what you will get is that in the team adjust, the missed FG will be shared among all scorers, like missed Reb are. Is it fair? I think so, Is it real? Of course not, like is not real there is no one guiltier rebounder for every missed Reb.
OK. You will say there are more missed FG than FGM, and there are less missed Reb than DRebs, and in the end unidimensional low usage scorers and not-center defenders will be hurted more. That's basketball as I understand it.
Back to top
View user's profile Send private message
Mike G
Joined: 14 Jan 2005
Posts: 3552
Location: Hendersonville, NC
PostPosted: Thu May 24, 2007 10:32 am Post subject: Reply with quote
Statman wrote:
... - David Lee was 5th in rebounding per minute (not 2nd) - behind Evans, Mutombo, Chandler, & Foster. ..
Lee was 4th in RebRt., three spots ahead of Foster. In 'my' Reb rate, Boozer is #1, followed by Chandler, DHoward, Mutombo, Lee, and Evans.
Since there's a lot of agreement that one's teammates may inflate or depress one's RebRt, why don't we agree we should adjust this by the ratio of TmReb/OppReb, rather than strictly by (TmRb + OppRb) ?
Utah's league-best rebounding advantage -- led by Boozer -- suggests he'd have had an even higher RbRt on a weaker rebounding team, doesn't it?
David Lee started 12 games this year. In those, he averaged 13.3 Reb/40. In his (46) games off the bench, he had 14.1 Reb/40. Curry (8.0 Reb/40) started every game. Is lee a 74% better rebounder than Curry, or just 66% better?
_________________
`
36% of all statistics are wrong
Last edited by Mike G on Thu May 24, 2007 10:53 am; edited 1 time in total
Back to top
View user's profile Send private message Send e-mail
Gordon
Joined: 02 May 2007
Posts: 3
PostPosted: Thu May 24, 2007 10:46 am Post subject: Reply with quote
Perhaps this has been noted elsewhere, but one of the ironies in the WoW debate is that WS actually does less to penalize low percentage shooters than either EFF or PER in some important respects. Berri often emphasizes that his metric does not reward a player for taking lots of shots if he's a low% shooter. And that's true: WS will give less value than the other metrics to a high-volume/low% shooter, in comparison to a low-volume/low% shooter.
But what happens when we compare two players who both take a lot of shots, but differ in shooting efficiency? Let's take player A (8 FGs, 20 FGAs) and player B (12 FGs, 20 FGAs). In this case, Berri actually gives LESS advantage to the efficient shooter than do the other metrics. The Player B advantage over Player A for the 3 metrics is:
WS +8
PER +9.5
EFF +12
By deducting points for missed shots, rather than all shots, PER and EFF actually punish inefficient shooting more severely than WS, not less. These metrics punish inefficient shooting, while WS just punishes shooting in general.
Now, maybe it's possible to improve on PER or EFF such that a 42% guy who takes a lot of shots doesn't look better than a 42% with few shots. But don't we care more about comparisons among players who DO take shots, and ensuring that efficient shooting is properly valued in those comparisons? Ironically, Berri's system actually doesn't do enough to value efficiency in those comparisons.
Back to top
View user's profile Send private message
Statman
Joined: 20 Feb 2005
Posts: 237
Location: Arlington, Texas
PostPosted: Thu May 24, 2007 11:50 am Post subject: Reply with quote
Mike G wrote:
Since there's a lot of agreement that one's teammates may inflate or depress one's RebRt, why don't we agree we should adjust this by the ratio of TmReb/OppReb, rather than strictly by (TmRb + OppRb) ?
Utah's league-best rebounding advantage -- led by Boozer -- suggests he'd have had an even higher RbRt on a weaker rebounding team, doesn't it?
I agree completely. When I look at rebounding rate - I only consider opponents rebounds (so as to not punish players as much with great rebounding teamates, or reward players too much who have poor rebounding teamates). Off my head - I think I use (reb/min)/(OPPreb/OPPmin)*100. 100 would be an average rebounder, 200 would be twice as good, 50 half as good, etc. I usually try to do all my stuff off of 100.
Now - this is best done on a per game basis - and compiled. That way if a player misses games - those games aren't included at all in the data. For example - Boozer gets 12 rebounds in a regulation game while playing 36 minutes. The opposing team had 40 rebounds. The ESTIMATED rebound total of the opponent would be 40/240*36 = 6. Boozer's rate for that game would be 200. Compiling all the player's rebounds for the season & dividing by the compiled estimated opponent's rebounds and multiplying by 100 would give one the player's rate for the season.
That's just my take on the rebound rate idea Mike G.
_________________
Dan
My current ratings (go to the last page for the most recent stuff):
http://www.pointguardu.com/f136/statmans-ratings-56243/
Back to top
View user's profile Send private message Send e-mail
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Thu May 24, 2007 6:34 pm Post subject: Reply with quote
Quote:
By deducting points for missed shots, rather than all shots, PER and EFF actually punish inefficient shooting more severely than WS, not less. These metrics punish inefficient shooting, while WS just punishes shooting in general.
If a FGA is already a lost possession (and an automatic punishment for the scorer), even a FGM is a lost possession, of course a missed FG is also one (the WOW school of no continuation plays). Why scorers must share FGM with the assistant (or assistants) and not the missed FG, or maybe every FGA? What if we charge him (them) a 30%?
Back to top
View user's profile Send private message
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Thu May 24, 2007 9:01 pm Post subject: Reply with quote
Quote:
But what happens when we compare two players who both take a lot of shots, but differ in shooting efficiency? Let's take player A (8 FGs, 20 FGAs) and player B (12 FGs, 20 FGAs). In this case, Berri actually gives LESS advantage to the efficient shooter than do the other metrics. The Player B advantage over Player A for the 3 metrics is:
WS +8
PER +9.5
EFF +12
Can somebody explain me how 12 missed FG are weighted the same as 8? and the FGM difference is the only thing that counts?
Back to top
View user's profile Send private message
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Fri May 25, 2007 12:27 am Post subject: Reply with quote
Phew. Alright.
Statman - The link is:
http://dberri.wordpress.com/2007/03/08/ ... t-mutombo/
1. The numbers aren't wrong. There is a chart in the above link, outlining their performance through 61 games, which is the sampe in question.
2. The basic argument is that if the Rockets went 20-12 with him filling in for Yao, he must still be pretty good. I dont actually see you responding to this argument, or presenting an alternative theory of what happened. Perhaps Chuck Hayes picked up the slack for Yao? Give Berri credit, at least he presents a very straightforward analysis and allows the data to do the driving.
3. Thank you for the COMPLETE stats. I apologize to anyone who didn't realize that Yao scored much more than Mutombo.
3A. Mark Eaton vs. Michaal Jordan? This little nugget shows that you haven't read the Wages of Wins and don't understand how it works. Berri adjusts for position. You dont compare centers to guards, you compare centers to centers, and shooting guards to shooting guards. Michael Jordan was way above the average for shooting guards. Mark Eaton, not so much relative to other centers.
4. Lee and Mutombo are rated on a per-minute basis. They are very far from the most productive players in the league. That honor went to Jason Kidd this year. He isn't very good I imagine, its just the 8 rebounds per game he collected at the point guard position that makes him look good. That's almost a rebound per game more than Eddy Curry btw. Lee's rebounding stats came from Knickerblogger, but clearly that's wrong since there are two people who have told me so. I got the stats here.
http://www.knickerblogger.net/stats/2007/jh_ALL_REB.htm
At any rate, he is one of the very best rebounders in the league.
We will see about Lee. I think he is that good, having watched him closely all year. Really, he and Balkman brought me back to the Garden and the Knicks. But I wouldnt say we have enough data yet to be sure. Still, its certain that he is much better than Curry, much much better.
What doesl +/- say about Curry btw?
Admin - I would love to know what the accepted wisdom on compensation in the NBA is. Berri makes a big deal of the fact that there is no correlation between payroll and wins. But there is a salary cap, so I don't know what that means. Until someone convinces me otherwise, I am going to stick with the belief that a lot of NBA Gm's don't have any idea what they are doing. As a Knicks fan, you will have to forgive me for feeling very justified in that assessment.
Gordon - I prefer to think that there is a third option here. Existing metrics discussed so far overrate scoring, in order to pass the laugh test, and Berri has now produced a rating that correctly weighs rebounding and scoring. That metric is not Winscore, which you guys keep referring to, but Wins Produced. Your argument is correct as it relates to WS, which was designed to simply be close enough for gov't work. But in Wins Produced, the baseline is not zero, but the league average for the position. It just isnt true that 100% of players value derive from rebounding, or even from the gaining and maintaining side of the ledger. I made that point already. But there is no question that Berri believes that gaining and maintaining possesion is more important than scoring. I don't really have any problem believing that. Scoring is what catches the eye but its impact is overvalued.
If someone came along with a statistical method for analyzing offensive and defensive line play, and said that the battle at the line of scrimmage was by far the most important determinate of success in the NFL, I wouldn't find that surprising in the least. But my experience with baseball statistics has made it very easy for me to believe that conventional wisdom is wrong in sports.
Vis a vis whether the ability to take shots at a league average efficiency is a common skill, I would say basically yes. I dont think scoring is really that difficult. I mean, Vince Carter style scoring is difficult, and difficult to watch in the playoffs, but in general I think most NBA teams could still score points without high usage players around. Looked what happend in Philly. Iverson, super high-usage player, decamps for Denver. Miraculously Philadelphia manages to get almost exactly as many shots off as before and score as many points.
Mike G - I just dont follow your ideas. I dont understand this "default" business. You seem to be saying that you think that a 45% shooter has to take shots so that other players can achieve there higher shooting efficiencies, since they couldnt do it without him and his high usage. This has been a consistent theme for you I think. Its another version of the thought that scorers create opportunities by missing shots. Anyway, thats not how I view basketball. Its a team game. It requires scorers and role players, and great role players are every bit as useful as conventional scoring stars.
Eddy Curry is a scorer, no question, but what really is the value of that 60% ts%? Is it worth putting a very poor rebounder out there who was second in the league in turnovers and who had a new statistical category created in his honor, the Eddy Curry line (think mendoza line)? I also love your phrase, "go-to scorer." Go to for a turnover, or directions to the nearest burger king, but certainly not a go-to player if you want to win basketball games.
If you think Eddy Curry is a good, above average basketball player, you are definitely guilty of overvaluing scoring. And I will leave it at that.
Alright. My computer went down, which is why I didn't respond. But I can't keep up here. And I won't convince anyone. So this will be my last post on this thread, Looking up, its not even a very good post, its just too much to bounce back something at each of you. It's been fun.
Back to top
View user's profile Send private message
Statman
Flint wrote:
Phew. Alright.
Statman - The link is:
http://dberri.wordpress.com/2007/03/08/ ... t-mutombo/
1. The numbers aren't wrong. There is a chart in the above link, outlining their performance through 61 games, which is the sampe in question.
2. The basic argument is that if the Rockets went 20-12 with him filling in for Yao, he must still be pretty good. I dont actually see you responding to this argument, or presenting an alternative theory of what happened. Perhaps Chuck Hayes picked up the slack for Yao? Give Berri credit, at least he presents a very straightforward analysis and allows the data to do the driving.
3. Thank you for the COMPLETE stats. I apologize to anyone who didn't realize that Yao scored much more than Mutombo.
3A. Mark Eaton vs. Michaal Jordan? This little nugget shows that you haven't read the Wages of Wins and don't understand how it works. Berri adjusts for position. You dont compare centers to guards, you compare centers to centers, and shooting guards to shooting guards. Michael Jordan was way above the average for shooting guards. Mark Eaton, not so much relative to other centers.
4. Lee and Mutombo are rated on a per-minute basis. They are very far from the most productive players in the league. That honor went to Jason Kidd this year. He isn't very good I imagine, its just the 8 rebounds per game he collected at the point guard position that makes him look good. That's almost a rebound per game more than Eddy Curry btw. Lee's rebounding stats came from Knickerblogger, but clearly that's wrong since there are two people who have told me so. I got the stats here.
http://www.knickerblogger.net/stats/2007/jh_ALL_REB.htm
At any rate, he is one of the very best rebounders in the league.
We will see about Lee. I think he is that good, having watched him closely all year. Really, he and Balkman brought me back to the Garden and the Knicks. But I wouldnt say we have enough data yet to be sure. Still, its certain that he is much better than Curry, much much better.
What doesl +/- say about Curry btw?
Admin - I would love to know what the accepted wisdom on compensation in the NBA is. Berri makes a big deal of the fact that there is no correlation between payroll and wins. But there is a salary cap, so I don't know what that means. Until someone convinces me otherwise, I am going to stick with the belief that a lot of NBA Gm's don't have any idea what they are doing. As a Knicks fan, you will have to forgive me for feeling very justified in that assessment.
Gordon - I prefer to think that there is a third option here. Existing metrics discussed so far overrate scoring, in order to pass the laugh test, and Berri has now produced a rating that correctly weighs rebounding and scoring. That metric is not Winscore, which you guys keep referring to, but Wins Produced. Your argument is correct as it relates to WS, which was designed to simply be close enough for gov't work. But in Wins Produced, the baseline is not zero, but the league average for the position. It just isnt true that 100% of players value derive from rebounding, or even from the gaining and maintaining side of the ledger. I made that point already. But there is no question that Berri believes that gaining and maintaining possesion is more important than scoring. I don't really have any problem believing that. Scoring is what catches the eye but its impact is overvalued.
If someone came along with a statistical method for analyzing offensive and defensive line play, and said that the battle at the line of scrimmage was by far the most important determinate of success in the NFL, I wouldn't find that surprising in the least. But my experience with baseball statistics has made it very easy for me to believe that conventional wisdom is wrong in sports.
Vis a vis whether the ability to take shots at a league average efficiency is a common skill, I would say basically yes. I dont think scoring is really that difficult. I mean, Vince Carter style scoring is difficult, and difficult to watch in the playoffs, but in general I think most NBA teams could still score points without high usage players around. Looked what happend in Philly. Iverson, super high-usage player, decamps for Denver. Miraculously Philadelphia manages to get almost exactly as many shots off as before and score as many points.
Mike G - I just dont follow your ideas. I dont understand this "default" business. You seem to be saying that you think that a 45% shooter has to take shots so that other players can achieve there higher shooting efficiencies, since they couldnt do it without him and his high usage. This has been a consistent theme for you I think. Its another version of the thought that scorers create opportunities by missing shots. Anyway, thats not how I view basketball. Its a team game. It requires scorers and role players, and great role players are every bit as useful as conventional scoring stars.
Eddy Curry is a scorer, no question, but what really is the value of that 60% ts%? Is it worth putting a very poor rebounder out there who was second in the league in turnovers and who had a new statistical category created in his honor, the Eddy Curry line (think mendoza line)? I also love your phrase, "go-to scorer." Go to for a turnover, or directions to the nearest burger king, but certainly not a go-to player if you want to win basketball games.
If you think Eddy Curry is a good, above average basketball player, you are definitely guilty of overvaluing scoring. And I will leave it at that.
Alright. My computer went down, which is why I didn't respond. But I can't keep up here. And I won't convince anyone. So this will be my last post on this thread, Looking up, its not even a very good post, its just too much to bounce back something at each of you. It's been fun.
Well - I'm annoyed - just lost a large post I made. Oh well - good debate.
Suffice to say that I believe that Berri's system of adjusting scoring off of league norm in terms of efficiency while seemingly adjusting everything else off of mean zero comes out flawed. Scoring is way undervalued, while rebounding becomes WAY overvalued because of their prevelance.
when the data comes out - I'll BET that the most "underrated" players according to Berri will be guys that don't shoot much and are good to great rebounders for their positions - across the board. Just a guess Wink
_________________
Dan
page 8 missing
PostPosted: Sun May 20, 2007 3:33 pm
Flint wrote:
Nick - The problem here is that Win Score, as noted before, is just a quick and dirty metric. It omits team defense. That's why you observe an anomaly. The more complete Wins Produced metric, does incorporate team defense. And if you look at it through Wins Produced, both player's "teams" do come out equal.
How does that work?
Team B has a defensive rating that is 3% worse than team A, but their score is ~6% higher. That implies that the adjustment is greater than the difference in their defenses.
Back to top
View user's profile Send private message
NickS
PostPosted: Sun May 20, 2007 3:51 pm Post subject:
Based on all of this I did a quick and dirty player rating based on points generated above a baseline scoring efficiency and it's interesting. I just through this together, so I will double check, but here are the top 10 on a per/minute basis for a couple of values of a possession.
It's interesting to see how usage becomes more important, compared to keeping and getting possessions as you change the baseline rate.
Looking at the numbers, I may be undercounting the cost of a TO (I'll double check that), but the lists are plausible. FWIW, David Lee (the WoW MVP) clocks in at #11 on the first list.
Possession = 1 point:
1: nash,steve
2: wade,dwyane
3: duncan,tim
4: camby,marcus
5: nowitzki,dirk
6: ginobili,manu
7: ming,yao
8: stoudemire,amare
9: garnett,kevin
10: gasol,pau
Possession = .8 points (so credit for scoring at better than .8pt/possesion)
1: wade,dwyane
2: nash,steve
3: bryant,kobe
4: ming,yao
5: nowitzki,dirk
6: arenas,gilbert
7: duncan,tim
8: ginobili,manu
9: stoudemire,amare
10: garnett,kevin
Possession = .65 points
1: wade,dwyane
2: bryant,kobe
3: arenas,gilbert
4: ming,yao
5: nash,steve
6: nowitzki,dirk
7: anthony,carmelo
8: ginobili,manu
9: mcgrady,tracy
10: james,lebron
Back to top
View user's profile Send private message
NickS
PostPosted: Sun May 20, 2007 4:10 pm Post subject:
One last note (and I do hope that anyone who isn't interested has stopped reading at this point). The rankings above are not pace adjusted which helps Nash and Camby. Also, just for fun I calculated a top 10 if you value possessions at 1.15 points and the rebounders and shot blockers shoot to the top at that point.
1: camby,marcus
2: mutombo,dikembe
3: lee,david
4: nash,steve
5: duncan,tim
6: chandler,tyson
7: okafor,emeka
8: marion,shawn
9: kidd,jason
10: millsap,paul
Back to top
View user's profile Send private message
Flint
PostPosted: Sun May 20, 2007 4:26 pm Post subject:
You have to subtract the value of made defensive fgs from each team's totals. A made fg is worth . .33. A made 3 point fg is worth .066. (see page 103 of the WOW) If you do so the numbers work. It's different from Wins Score, and more accurate.
Back to top
View user's profile Send private message
NickS
PostPosted: Sun May 20, 2007 5:20 pm Post subject:
Thanks, so let me run those numbers:
Player A
39 3 pointers made: +2.574
56 2 pointers given up: -1.792
44 Defensive Rebounds: +1.496
15 Offensive Rebounds: +.51
76 missed Field Goals: -2.58
Total: .204
Player B
56 2 pointers made: +1.848
39 3 pointers given up: -2.574
61 Defensive Rebounds: +2.074
11 Offensive Rebounds: +.374
44 missed Field Goals: -1.496
Total: .226
I still have player B ranking 10% better than player A.
That's before team adjustment, but using the weights on page 103.
Back to top
View user's profile Send private message
Flint
PostPosted: Sun May 20, 2007 6:21 pm Post subject:
Your example is too complicated.
Let's assume each players has twelve possessions and there are no offensive rebounds. That is throwing things out of whack. I don't exactly know how he calculates wins produced, but if i were to take a crack at it it would look something like this.
Player A (two point shooter)
Two point fgm 6 x .33 = .198
D. Rebounds 8 x .034= .272
FGs Missed 6 x .034= -.204
3FGs allowed 4 x .066= -.264
Net .002
Player B (Three Point Shooter)
3 pt Fg made 4 x .066 = .264
DRB 6 x .034 = .204
FG missed 8 x .034 = -272
2 fg allowed 6 x .032 = .-192
Net .004
They dont quite add up to zero, I would assume this is some sort of rounding error. But basically, it definitely works out. The players performance is identical, as you would expect.
I dont really feel like tackling your example right now. Clearly, Wins Produced is going to come out clean any way you slice it. There may be fatal flaws in his approach, but the numbers adding up is not going to be it. I dont think he gets the book published or an economics professorship without being able to get that right.
Vis a vis rebounding being overvalued, Berri had a long post at this link
http://dberri.wordpress.com/2006/11/09/ ... -rebounds/
He finds that even if he reduces the value of a rebound to .7 of a point his results are almost exactly the same. This shows, he says, that the difference between his system and others is entirely related to how it accounts for scoring and shooting efficiency.
I saw Paul Millsap on one of your lists. Berri looked at his college numbers and predicted he would be one of the best rookies. Turned out to be very prescient, although he thinks that either Roy or Rondo are most deserving of the Rookie of the Year award.
Back to top
View user's profile Send private message
NickS
PostPosted: Mon May 21, 2007 12:14 am Post subject:
Flint wrote:
Vis a vis rebounding being overvalued, Berri had a long post at this link
http://dberri.wordpress.com/2006/11/09/ ... -rebounds/
He finds that even if he reduces the value of a rebound to .7 of a point his results are almost exactly the same.
This is an interesting point. I went back and re-read the Dave Berri post and realized that it says something slightly different than what you describe and that it actually supports the opposite interpretation.
You say that changing the value of a rebound from 1 to .7 produces "results [that] are almost exactly the same." Dave Berri writes "The correlation between Win Score and the Kaufman inspired formula is 0.99"
If you have a number that is made of a variety of components and you change the value of one of the components and get a new number that correlates exactly with the original number that means one of two things. Either the factor for which you have changed the weighting is insignificant, and the new number is the same as the original number, or the factor completely dominates the rating and the new number is scalled to the orginal number by the amount which you have changed the valuation of that factor.
Given that we know that the total value of scoring among all players is 0. It seems very likely that for many players rebounding is the major component of their win score.
So this suggests that the second scenario (rebounding dominates win score) is more likely than the first scenario (rebounding is insignificant to win score).
So the fact that the correlation is extremely close to 1 after changing the value of rebounding is probably actually a sign that rebounding completely dominates the metric, rather than a sign that it is insignificant.
I never noticed that before.
Back to top
View user's profile Send private message
Mike G
PostPosted: Mon May 21, 2007 7:28 am Post subject:
I confess that I haven't poured my whole soul into trying to follow the zeroes and ones that make up this accounting system. But it's intuitively bizarre that "... the total value of scoring among all players is 0."
And if so, it's doubly bizarre that 'the total value of rebounding' would not be zero.
So I refer back to my much earlier question: If missed shots are assigned a deficit in these formulas, why aren't 'ungotten' rebounds also considered to be a negative?
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Harold Almonte
PostPosted: Mon May 21, 2007 10:36 am Post subject: Reply with quote
Quote:
Player A (two point shooter)
Two point fgm 6 x .33 = .198
D. Rebounds 8 x .034= .272
FGs Missed 6 x .034= -.204
3FGs allowed 4 x .066= -.264
Net .002
Player B (Three Point Shooter)
3 pt Fg made 4 x .066 = .264
DRB 6 x .034 = .204
FG missed 8 x .034 = -272
2 fg allowed 6 x .032 = .-192
Net .004
The only thing I can see here is: scoring totals are more important than efficiency, rebounds and TOs. The player B tied the game having less efficiency, less rebounds and more missed FGs. You can say, It's three points that have an extra point prize. Well, bad luck scoring 3p is a more valuable skill.
The same if you have an "unskilled" but careful worker A who only produces 9 articles with just 1 mistake a whole day of labor; and a "skilled" and experienced worker B who produces 30 with 7 mistakes the same day. Who give more money to the company? The "no mistake" or the "productive"?
You can say a DReb earns the possession, but did the DReb avoid the scoring?. It's DReb (1 poss. prize) the counterpart of missFG (-1 poss. punish)?. Even if the offensive team score, you'll also have your possession. The problem with WS is that DReb. suposedly did all the defense...This is old stuff.
Back to top
View user's profile Send private message
NickS
PostPosted: Mon May 21, 2007 11:00 am Post subject:
Harold Almonte wrote:
The problem with WS is that DReb. suposedly did all the defense...This is old stuff.
Actually, I'm arguing that, separate from any dispute over how WS assigns credit, that it is generating the wrong amount of credit to assign.
Think about this example, if we imagine the game of 1-on-1, the same player is forcing the miss and getting the rebound, so there isn't any problem with the assignment of credit. The problem is that the worse player accumulates more credits than the better one, showing that there's a problem with how credits are generated.
Back to top
Flint
Posts: 112
PostPosted: Mon May 21, 2007 11:04 am
Nick -
NBA Efficiency and the WOW value points scored and rebounds exactly the same. In each system a point scored and a rebound collected have the same value. If rebounding "dominated win score," then it should dominate NBA Efficiency also, and the two ratings should be basically identical. However, they are not. The reason they are not is that they treat shooting efficiency much differently. This would indicate to me that how you account for scoring is a major part of the player rating, probably more important than rebounding.
If you have read the book, you know that the major point of it is that scoring is wildly overrated in the NBA relative to its contribution to wins. Players are paid huge salaries and given all star and post season honors and shoe contracts on the basis of how many points they score, with little regard for shooting efficiency or their other contributions, or lack therof in other areas of the box score.
Given that this is the case, I find it surprising, to say the least, that you think shooting efficiency doesnt matter at all in Berri's rating and that rebounding is the only thing that counts.
Mike - I think what you mean is, why aren't you deducted for allowing an offensive rebound? In the same way that a shooter costs his team a possession (unless his team recovers it), doesn't someone cost the defensive team a possession when an offensive rebounding is allowed? I would say to that perhaps it comes out in the wash in Wins Produced when you deduct for opponent's made field goals. The "cost" of allowing this offensive rebound is spread throughout the team through the mechanism of reduced defensive efficiency. The "cost" of not capturing that rebound is the points scored off the offensive rebound, and this cost is deducted from each player. Would that make it even out in your view?
Back to top
View user's profile Send private message
Harold Almonte
PostPosted: Mon May 21, 2007 11:33 am
Quote:
The "cost" of not capturing that rebound is the points scored off the offensive rebound
How can we assure an offensive rebound will be automatically converted in score?
The only thing a DReb avoids is the other team don't rebound back the ball (and does that only 70% of the time). It doesn't avoid scoring, "defense around" avoided scoring, but then you come and give the full possession value to the rebound and even you add a team defense adjust. You are overrating defense. And I'm not denying frontcout "might" have a higher defensive value then backcourt. And even this need to be wide proved.
One metric is offensive biased, the other is defensive (rebound) biased. The utility of NBAeff. is none, PER is the usage concept, and WP the team win fit, but all of them are just "almost credible"
Back to top
View user's profile Send private message
NickS
PostPosted: Mon May 21, 2007 12:11 pm
Flint wrote:
Given that this is the case, I find it surprising, to say the least, that you think shooting efficiency doesnt matter at all in Berri's rating and that rebounding is the only thing that counts.
That's not what I'm saying, exactly. I was arguing that the Rebounding component of the WinScore/WinsProduced metric is the major source of variability in player rankings.
(noted that this is the simpler metric)
Points + Rebounds + Steals + ½Assists + ½Blocked Shots – Field Goal Attempts – Turnovers - ½Free Throw Attempts - ½Personal Fouls
There are no steals, assists, blocked shots, turnovers, free throws or fouls.
So that is
Points + Rebounds - Field Goal Attempts
You can divide that into a scoring portion (Points - Field Goal Attempts) and a rebounding portion (rebounds).
Consider our example of player A and player B:
Player A: Scoring (117 - 115) Rebounding (49)
= 2 + 49
Player B: Scoring (112 - 111) rebounding (72)
= 1 + 72
What you can see is that the rebounding portion of the total is much larger than the scoring portion. This means that if you have or double the value of rebounding it will correlate extremely well with the original.
2+ (49 / 2) ~ (2+49)/2
dividing the value of rebounding in half correlates well with just dividing the total score in half.
I'm saying that the high level of correlation that Berri mentions in that post means that the scoring potion (including positive value for points scored and negative value for possessions used) is being swamped by the rebounding portion.
Back to top
View user's profile Send private message
Mark
PostPosted: Mon May 21, 2007 4:29 pm
Mike G wrote:
And if so, it's doubly bizarre that 'the total value of rebounding' would not be zero.
So I refer back to my much earlier question: If missed shots are assigned a deficit in these formulas, why aren't 'ungotten' rebounds also considered to be a negative?
If the rating system is intended to represent both offense and defense fully then I agree the double entry accounting for rebounding (credit/debit) to each team should sum to zero- in the end. A defensive rebounding departure from average should get dealt with in the team defense adjustment. Perhaps some of the value of those defensive rebounds gets backed out? Seems like it has to to get back to balance.
To some extent "ungotten" rebounds can have an impact on position adjusted standings- if you get less than average it will affect your score relative to so called peers.
Looking at individual rebounding relative to position expectation and/or actual counterpart data and plugging it directly into WP score could improve ratings and reduce size of team adjustrment. Individual shot defense is another step that could help get to full double entry accounting completeness at player level but the quality of the database gets raised by some as a stumbling block as discussed before.
Last edited by Mark on Mon May 21, 2007 7:46 pm; edited 6 times in total
Back to top
View user's profile Send private message
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Mon May 21, 2007 5:32 pm
The only thing that can be done with linear formulas is: to shortcut or to make the team adjust (but not to do both things at the same time, then you are cheating).
We're talking about give credits of 2/3 or .75 to FGA and DReb. Give similar credit to the OReb. But, how do you punish scoring teammates by not offensive rebounding (their 30%) if you fail to score?. How do you punish defense by being scored?. How do you spare the supposed 0.25 of defense around a DReb?. All those 1/3s must necessearily be adjusted at both end. That's the best justice.
I don't like any linear formula, I don't see the big importance of an overall season performance rating (the team-win-$ fit of WP only-they don't want to repair). I would prefer another kind of "scouting point of view" metric where you define some skill-block and you use the block of metrics (almost like PER but with defense) you thing best qualify those skills. How would you weight those skill-block? 4 factors would be the guide, I think you don't need more regression than that. But even with this metric, the boxscore wouldn't help with somethings like screen-assists, potential assists, getting free, counted blocks, counted deflections, counted double teams, charges, zones, etc.
I would add emotionals factors to this metric: command and leadership, self esteem, egocentrism.
page 6 missing
jeffpotts77
PostPosted: Wed May 23, 2007 1:11 pm
Flint wrote:
You referenced numbers, but I don't think they show convincingly that Yao is better. A poster at the WOW broke down their numbers in a comment. Here is what he said:
Quote:
on a per 48 minute basis:
blocks : tied at around 3
turnovers: 1.5 for dikembe, 5.4 for yao
assists: .5 for dikembe, 3 for yao
rebounds: 18.5 for dikembe, 13.3 for yao
fg% 56.6 for dikembe, 51.5 for yao
personal fouls: 5.5 for dikembe, 5.0 for yao
So it appears that dikembes main advantages are in turnovers, and rebounds, and field goal percentage.
Your point about TS% is well taken. Yao obviously is the better scorer since he scores more pointa with the same efficiency. But Mutombo is a MUCH better rebounder than Yao, and he commits many fewer turnovers. Over 48 minutes Deke was worth nine more possessions. That's a big difference. I understand your point that the reason Mutombo doesn't commit turnovers is that he is a low usage player, and that actually he has a higher TO rate. But I am not sure that really translates into an impact on wins. The implication of Berri's work is that there is always room for at least one very low usage player out there. You can't play with five of them, but having one of them out there is a great thing. For this genre of ball player, the defensive center, Mutombo is the absolute prototype (well Bill Russell maybe) and Chandler is the newest version.
For every low usage player included in a lineup, the other players must absorb/use extra possessions. The fact that Yao can use so many possessions so efficiently, with a lower turnover rate and a higher TS%, makes him much more valuable than Deke, even with the disparity in their rebound rates. You keep harping on the fact that the Rockets were better with Yao out of the lineup but this argument ignores the fact that this event did not occur in a vacuum. So many things could have occured when Yao went on the shelf that can't be found in the boxscores. Was Tracy McGrady playing with more fire during this stretch? Were various players (other than Deke) playing with more intensity on the defensive end? Was their schedule easier during this stretch? Was this just a case of too small a sample size? I'm sure there are more variables that I haven't thought of. You cannot draw the (seemingly absurd) conclusion that Deke is better than Yao without first addressing those other possibilities.
Back to top
View user's profile Send private message Yahoo Messenger
Flint
PostPosted: Wed May 23, 2007 1:32 pm
Justin - Well, I object to the fact, as a WOW fan, that some people, Mike G I am looking at you, think its a weaker metric than EFF.
I would say there are two very good reasons to discuss NBA Efficiency. First, it's the official metric of the league. Second, money.
Berri has some cool stuff about salaries in th NBA. The controversy over his Wins Produced metric has obscured his (at least to me interesting) findings about compensation. I know a lot of studies have already been done. But He found he "could explain 63% of player salary when we employed the NBA's performance metric," while Wins Produced only explains 47% of salary. I wonder what the number is for PER.
He finds that points scored works a little bit better than NBA efficiency, at 63.5% and that:
"
Quote:
if we add a player's accumulation in the past season of rebounds, steals, blocked shots, assists, turnovers, personal fouls, and shooting efficiency," our explanatory power rises to 64%. Of all these statistics, only blocked shots were found to significantly impact compensation. And like our study of the All-Rookie team, the number of points scored has the largest impact on player salary. Furthermore, increases in rebounds, steals, assists, turnovers, personal fouls, and shooting efficiency on both field goals and free throw attempts were not found to have significant impact on player salary."
The basic take away is that "player salary in the NBA is dominated by how many points a player scores. And it does not seem to matter much if the player scores efficiently, or if he rebounds, generates steals, or avoids turnovers. Basically if a player can score, he will score a major payday. One should note that this same pattern is also seen in the coaches’ voting for the All-Rookie team."
I know others said that before Berri, but it was shocking to me.
Back to top
View user's profile Send private message
Statman
PostPosted: Wed May 23, 2007 1:48 pm
Wow - alot to respond here - let me give it a go:
1.
Quote:
Ok. You dont think Berri's evaluation of Mutombo passes the laugh test.
[quote="Flint"]
Quote:
Seriously - not a single objective (or even casual) observer or APBRmetric stats freak would EVER dare to say Dikembe was more valuable than Yao. Yet - Berri's numbers seem to say this. HUGE red flag for me.
It certainly does go against the grain. But you have to give the guy credit. He doesn't turn tail when the data coughs up strange results, he embraces these quirks. He posted on Mutombo-Yao a while back, and in my mind it was one of his bests post.
http://dberri.wordpress.com/2007/03/08/ ... mutombo/re
If you dont think Deke was as good as Yao this year, the burden is on you to explain how the Rockets had a better record without Yao after his injury, than they had with him before. Really, your laugh test conclusion is actually difficult to defend , given this fact. I posed this question on a previous thread I think, but its a great little puzzle. How do you explain it? How did the Rockets win more games without Yao?
Couldn't find that article. Anyway - WHERE do you get this info about the Rockets having a better record without Yao? The Rockets were 20-14 without Yao Ming,. They also lost the only game he played less than 19 minutes (6 minutes AT HOME against the LA Clippers). Only half those 34 games were versus playoff teams - and TWENTY of those games were at home. YET - they still were MUCH WORSE record wise without Yao than with him (20-14 to 32-16) despite favorable conditions (more home games) without him.
So - the Rockets were much better per possession when Yao was on the court than when he when he wasn't - and were slightly WORSE when Dikembe was on the court than when he wasn't. The Rockets were much worse record wise w/o Yao than with him considering the conditions. Is that good enough for the "burden of proof"?
2.
Quote:
And all this doesn't mean Mutombo is the better player right now to have on your team. If I were a GM and I had the choice between Mutombo and Yao, of course I would choose Yao. But it would largely because of future potential and his ability to play more minutes. Yao also was much better last year than this year in terms of his WOW numbers. But I would certainly rather have Mutombo in his prime than Yao and the performance he has provided so far in his career.
Uh - what would Mutambo's WoW FOR THIS YEAR have ANYTHING to do with Mutambo in his prime? Mutambo isn't CLOSE to the player he was in his prime - and he's nowhere CLOSE to the player Yao is now. But - Berri has him "rated" better.
3.
Quote:
You referenced numbers, but I don't think they show convincingly that Yao is better. A poster at the WOW broke down their numbers in a comment. Here is what he said:
Quote:
on a per 48 minute basis:
blocks : tied at around 3
turnovers: 1.5 for dikembe, 5.4 for yao
assists: .5 for dikembe, 3 for yao
rebounds: 18.5 for dikembe, 13.3 for yao
fg% 56.6 for dikembe, 51.5 for yao
personal fouls: 5.5 for dikembe, 5.0 for yao
So it appears that dikembes main advantages are in turnovers, and rebounds, and field goal percentage.
Your point about TS% is well taken. Yao obviously is the better scorer since he scores more pointa with the same efficiency. But Mutombo is a MUCH better rebounder than Yao, and he commits many fewer turnovers. Over 48 minutes Deke was worth nine more possessions. That's a big difference. I understand your point that the reason Mutombo doesn't commit turnovers is that he is a low usage player, and that actually he has a higher TO rate. But I am not sure that really translates into an impact on wins. The implication of Berri's work is that there is always room for at least one very low usage player out there. You can't play with five of them, but having one of them out there is a great thing. For this genre of ball player, the defensive center, Mutombo is the absolute prototype (well Bill Russell maybe) and Chandler is the newest version.
INTERESTING. Let's see - you completely don't include points. You use FG% INSTEAD of TS% - which incorporates getting to the line and hitting free throws - something Yao does much better than Dikembe.
Here are the COMPLETE stats - per 48 minutes:
TS% - exactly the same 60.1%
Points: Yao 35.5 to Dikembe 8.6 (Yao is 4.16 TIMES better)
Rebounds: Yao 13.4 to Dikembe 18.1 (Dikembe is about 1.36 times better)
Assists:: Yao 2.8 to Dikembe 0.5 (Yao is 5.74 times better)
Steals: Yao 0.5 to Dikembe 0.8 (Dikembe is 1.63 times better)
Turnovers: Yao 4.9 to Dikembe 1.4 (Dikembe is 3.49 times better)
blocks: Yao 2.8 to Dikembe 2.8 (Dikembe is 1.02 times better)
You call Dikembe MUCH better a rebounder. Well - Yao is three times a better SCORER (btw - scoring points still wins games) than Mutambo is a better rebounder. Yao turns the ball over much more - but garners assists and points at a MUCH higher rate the he commits turnovers in relation to Dikembe (can't have all that good without some bad).
Look at those stats. Now - ALSO consider that Yao's +/- is MUCH better than Dikembe's - and Houston was MUCH better when Yao started than when Dikembe started - do you actually THINK it's even a comparison?
In 90-91 - Mark Eaton & Michael Jordan had almost identical TS%. Eaton was a MUCH better rebounder, INFINITELY better shot blocker, turned the ball over MUCH less. Would anyone even remotely consider Eaton better than Jordan? Well - considering Dikembe better than Yao is almost (but not quite) on par to this statistically.
4.
Quote:
Vis a vis Eddy Curry, I have a visceral hatred for the man's game. Having watched him play most of this year, I wouldn't hesitate to say that its far more logical to call David Lee one of the best performers in the league on a per minute basis, than to call Eddy Curry an above average player. Berri doesnt actually rate Curry as slightly below average. He rates him way below average. In Berri's system, .100 is the average player, so Curry's .002 suggests that Curry contributed almost nothing in terms of wins this year, despite playing 35 minutes per.
I love this btw:
Quote:
His own coach didn't see fit to play him more than Curry (who to you was OBVIOUSLY horrible in every facet of the game besides shooting).
His coach is Isaiah Thomas!
Curry was "obviously horrible" in every facet of the game other than scoring. He was actually historically bad. He assist to turnover ratio was ..24.And he collected less than twice as many rebounds as he committed turnovers. He was second in the entire league in turnovers! He had more turnovers than Steve Nash. And he is one of the worst rebounding centers in the league. For a center, those are fairly historic numbers. He blocked fewer shots than Renaldo Balkman and his off box score defense could charitably be described as lousy.
What the numbers say to me is that Eddy Curry is one of the worst centers playing more than 30 minutes in the NBA. I can think of 20 players at the position who I would rather see in a Knicks uniform. There is no way he even approaches average, unless you consider, as PER and EFF do, scoring to be much the most important thing that players do on a basketball court.
David Lee is the man. He wasn't "well above average" in rebounding, he was, with Tyson Chandler, the best rebounder in the league per minute.
OK - you hate Curry (and Iverson from previous posts) - so it's understandable you latch onto this as "proof" Berri's system is on point. I understand. However - PER, Eff., and pretty much ALL rating systems have David Lee as a bit to much better than Curry per minute. None of them would have Lee (playing less than 28 minutes a game on a BAD team) as the BEST in the league - like Berri does. I agree - I don't think Curry should rate above average - BUT to denounce completely other systems because they do yet latch on in support of a system that has DAVID LEE as the best in the league per minute and DIKEMBE MUTAMBO as better than Yao Ming seems to be a HUGE leap of faith just because the system better fits your view of players like Curry and Iverson.
BTW - David Lee was 5th in rebounding per minute (not 2nd) - behind Evans, Mutambo, Chandler, & Foster. I'll be willing to bet ALL those guys rate VERY highly in Berri's system per minute - because the system obviously has a huge bias towards rebounders who don't ATTEMPT much of anything else. If you rebound alot and never take (thus possibly miss) a shot and never touch the ball (thus never turning the ball over) because you are an offensive zero - you will rate in the top 30 in the league.
5.
Quote:
Important difference. His Ts% was second in the league to Steve Nash. True, he didnt score much, but that's irrelavant and his 14.4 per 40 really isnt all that bad. Curry, Marbury, and Crawford dominated the ball. I can barely remember any plays being called for Lee. But having watched him closely, he seems like the kind of player who is more than low usage. One thing is for sure. Before he went down with injury, the Knicks had actually posted a respectable number of wins. This had nothing to do with Curry, and everything to do with Lee. Curry's numbers are exactly what you would expect given his previous performance, he just played more minutes this year.
I dont think its a pipe dream to imagine Lee next year playing 37 minutes, averaging 19 and 12 on 55% fg shooting, committing less than 3 turnovers per game, and chipping in 2.5 assists, a steal and block per. Those are certainly all-star numbers, even on a goofy looking white guy, and to me if he achieves those number he will have to enter the discussion of who the best players in the league are. Sound crazy? Does that pass the laugh test?
Well - to think David Lee THIS SEASON was the best player in the league per minute is definitely a pipe dream. You love the kid - we all do (not as much as you though) - but a very good rebounder who shoots quite efficiencently (while taking few shots) does not an All NBAer make. I'm glad Berri helps you feel he's at or near the top of the league - but the truth is he still was not a completely a full time player on a BAD team. He was the best of the Knicks per minute without a doubt - but anything beyond that is highly suspect I think (especially near - say - top 10 in the entire NBA).
Since Berri overvalues rebounds, as well as somewhat overpenalizes turnovers and definitely overpenalizes missed shots - David Lee fits PERFECTLY in this WoW.
BTW - don't get TOO excited about Lee's future until you take a look at Danny Fortson's early years (much better scorer and rebounder per minute than Lee - and I am CERTAIN would have rated at or near the top of the league in WoW per minute). He hasn't really made any all star teams.
_________________
Dan
My current ratings (go to the last page for the most recent stuff):
http://www.pointguardu.com/f136/statmans-ratings-56243/
Back to top
View user's profile Send private message Send e-mail
Statman
s
PostPosted: Wed May 23, 2007 1:53 pm
jeffpotts77 wrote:
For every low usage player included in a lineup, the other players must absorb/use extra possessions. The fact that Yao can use so many possessions so efficiently, with a lower turnover rate and a higher TS%, makes him much more valuable than Deke, even with the disparity in their rebound rates. You keep harping on the fact that the Rockets were better with Yao out of the lineup but this argument ignores the fact that this event did not occur in a vacuum. So many things could have occured when Yao went on the shelf that can't be found in the boxscores. Was Tracy McGrady playing with more fire during this stretch? Were various players (other than Deke) playing with more intensity on the defensive end? Was their schedule easier during this stretch? Was this just a case of too small a sample size? I'm sure there are more variables that I haven't thought of. You cannot draw the (seemingly absurd) conclusion that Deke is better than Yao without first addressing those other possibilities.
Exactly - not to mention (as I just did the post before) that Houston was WORSE in EVERY way you look at it without Yao - against weaker competition and a higher percentage of home games. I don't know where this "better record without Yao" came from.
_________________
Dan
My current ratings (go to the last page for the most recent stuff):
http://www.pointguardu.com/f136/statmans-ratings-56243/
Back to top
View user's profile Send private message Send e-mail
Kevin Pelton
Site Admin
PostPosted: Wed May 23, 2007 2:27 pm
Flint wrote:
The basic take away is that "player salary in the NBA is dominated by how many points a player scores. ..."
This conclusion is not as strong, in my opinion, as Berri makes it out to be. Keep in mind that he is using total points instead of points per minute, so playing time plays a role in this correlation as well.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Gordon
PostPosted: Wed May 23, 2007 10:51 pm
Quote:
Nick - Its not rebounding that makes the difference. Its the difference in the way the WOW accounts for shooting efficiency. Rebounding is not overvalued. Its valued the same way as it is by Hollinger and NBA Efficiency.
There is no debate here -- you are both right (or wrong!). If a metric undervalues scoring, then it must overvalue rebounds (and vice-versa). What matters here is the RELATIVE weight each variable has in the metric -- how much influence does it have on the ranking of players? If scoring is rewarded less in one metric, but rebounds the same, then as a proportion of total player value points will be less important AND rebounds more.
Flint: consider a simple experiment: take PTS, FGA, and FTA entirely out of the WS formula. It would still be true, as Berri notes, that the formula differs only in its handling of points, not rebounds. But the new formula would indeed overvalue rebounds, in the sense that they would essentially determine player rankings all by themselves. So clearly, changing the values for scoring variables can impact the relative weight of rebounds in a metric.
Now, does WS overvalue rebounds/undervalue scoring? Clearly, Berri is right to value efficiency. But all metrics give more value to higher-efficiency shooters. The question is whether it is correct to use average NBA performance as the baseline for scoring (avg = 0), while using zero as the baseline for rebounding. Essentially, WS assumes that being an average NBA shooter and making roughly half your shots has no value at all to your team -- if you don't shoot it, someone else will -- but being an average NBA rebounder -- which means pulling down half the missed shots -- has huge value. The reason this is done is presumably because there are no "missed rebounds" in the box score. But that doesn't make it right to consider opportunities consumed (FGAs) in one case, but ignore the issue when it comes to rebounding.
Consider that for a statistically average player, in WS virtually none of his value derives from scoring, while almost 100% of his value derives from rebounding. Does anyone believe that the ability to take shots and score at league average efficiency is an ubiquitous skill that has no value? Does that seem plausible or reasonable?
Back to top
View user's profile Send private message
Mike G
PostPosted: Thu May 24, 2007 6:45 am
Gordon, thanks for elucidating the question I was asking on page 1.
WoW apparently supposes that if I don't take this 45% shot, someone else on my team will, by default, get a 50% shot. Then if I don't go for this rebound that I have a 50% chance to get, will a teammate automatically appear with a 70% chance to get it?
Both of those assumptions are absurd, but WoW seems to make one and exclude the other. Note that Flint, when asked about these absurdities, tends to reply along the lines of "But scoring is overvalued"; as if that justifies anything.
It may really be about the notion that average play = zero contribution. It makes more sense to me to say that the 20 RebRt guy is twice the rebounder as the 10 RebRt guy. Not that he's infinitely better (10 being average). Nor is he contributing 5x as much as the 12 RbRt player.
As long as scoring has no value (or negligible values both sides of zero), then Eddy Curry just makes turnovers and occasionally rebounds. If scoring had value, then he'd be a go-to scorer who rebounds and has too many TO. And that's what he is.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Thu May 24, 2007 8:22 am Post subject: Reply with quote
Quote:
if you don't shoot it, someone else will -- but being an average NBA rebounder -- which means pulling down half the missed shots -- has huge value. The reason this is done is presumably because there are no "missed rebounds" in the box score. But that doesn't make it right to consider opportunities consumed (FGAs) in one case, but ignore the issue when it comes to rebounding.
Quote:
WoW apparently supposes that if I don't take this 45% shot, someone else on my team will, by default, get a 50% shot. Then if I don't go for this rebound that I have a 50% chance to get, will a teammate automatically appear with a 70% chance to get it?
I can't understand why if you prize a DReb, punish the missed FG automatically, but when you get an OReb, who do you punish at the defensive end? Why do you rate scoring but no defense?
I think, but I don't have the skills to prove it, that WOW can be very good predicting (with luck help) no prolific rebounders players trades (Miller-A.I.), but have troubles when a lot of rebounds are in the mid (B. Wallace). After all a player's total rebounds depends on his teammates's rebounding ability.
O.K. you would say, there's no missed Rebs in boxscore, I will make up with a team defense adjust. But, why don't you cut also missed FG and make up with a team offense adjust, and let's see what you'll get? Don't you say Reb=1 or 0.75 poss. and missedFG = 1 or 0.75 poss?, well, then 50-50.
Scoring is a matter of 5 people, although just one shoots the ball, in the end is supossed to be a (70/30) relationship between the usager and his teammates. The same, rebounding is a matter of 5 people, although in the end just one can catch the ball.
P.D. what you will get is that in the team adjust, the missed FG will be shared among all scorers, like missed Reb are. Is it fair? I think so, Is it real? Of course not, like is not real there is no one guiltier rebounder for every missed Reb.
OK. You will say there are more missed FG than FGM, and there are less missed Reb than DRebs, and in the end unidimensional low usage scorers and not-center defenders will be hurted more. That's basketball as I understand it.
Back to top
View user's profile Send private message
Mike G
Joined: 14 Jan 2005
Posts: 3552
Location: Hendersonville, NC
PostPosted: Thu May 24, 2007 10:32 am Post subject: Reply with quote
Statman wrote:
... - David Lee was 5th in rebounding per minute (not 2nd) - behind Evans, Mutombo, Chandler, & Foster. ..
Lee was 4th in RebRt., three spots ahead of Foster. In 'my' Reb rate, Boozer is #1, followed by Chandler, DHoward, Mutombo, Lee, and Evans.
Since there's a lot of agreement that one's teammates may inflate or depress one's RebRt, why don't we agree we should adjust this by the ratio of TmReb/OppReb, rather than strictly by (TmRb + OppRb) ?
Utah's league-best rebounding advantage -- led by Boozer -- suggests he'd have had an even higher RbRt on a weaker rebounding team, doesn't it?
David Lee started 12 games this year. In those, he averaged 13.3 Reb/40. In his (46) games off the bench, he had 14.1 Reb/40. Curry (8.0 Reb/40) started every game. Is lee a 74% better rebounder than Curry, or just 66% better?
_________________
`
36% of all statistics are wrong
Last edited by Mike G on Thu May 24, 2007 10:53 am; edited 1 time in total
Back to top
View user's profile Send private message Send e-mail
Gordon
Joined: 02 May 2007
Posts: 3
PostPosted: Thu May 24, 2007 10:46 am Post subject: Reply with quote
Perhaps this has been noted elsewhere, but one of the ironies in the WoW debate is that WS actually does less to penalize low percentage shooters than either EFF or PER in some important respects. Berri often emphasizes that his metric does not reward a player for taking lots of shots if he's a low% shooter. And that's true: WS will give less value than the other metrics to a high-volume/low% shooter, in comparison to a low-volume/low% shooter.
But what happens when we compare two players who both take a lot of shots, but differ in shooting efficiency? Let's take player A (8 FGs, 20 FGAs) and player B (12 FGs, 20 FGAs). In this case, Berri actually gives LESS advantage to the efficient shooter than do the other metrics. The Player B advantage over Player A for the 3 metrics is:
WS +8
PER +9.5
EFF +12
By deducting points for missed shots, rather than all shots, PER and EFF actually punish inefficient shooting more severely than WS, not less. These metrics punish inefficient shooting, while WS just punishes shooting in general.
Now, maybe it's possible to improve on PER or EFF such that a 42% guy who takes a lot of shots doesn't look better than a 42% with few shots. But don't we care more about comparisons among players who DO take shots, and ensuring that efficient shooting is properly valued in those comparisons? Ironically, Berri's system actually doesn't do enough to value efficiency in those comparisons.
Back to top
View user's profile Send private message
Statman
Joined: 20 Feb 2005
Posts: 237
Location: Arlington, Texas
PostPosted: Thu May 24, 2007 11:50 am Post subject: Reply with quote
Mike G wrote:
Since there's a lot of agreement that one's teammates may inflate or depress one's RebRt, why don't we agree we should adjust this by the ratio of TmReb/OppReb, rather than strictly by (TmRb + OppRb) ?
Utah's league-best rebounding advantage -- led by Boozer -- suggests he'd have had an even higher RbRt on a weaker rebounding team, doesn't it?
I agree completely. When I look at rebounding rate - I only consider opponents rebounds (so as to not punish players as much with great rebounding teamates, or reward players too much who have poor rebounding teamates). Off my head - I think I use (reb/min)/(OPPreb/OPPmin)*100. 100 would be an average rebounder, 200 would be twice as good, 50 half as good, etc. I usually try to do all my stuff off of 100.
Now - this is best done on a per game basis - and compiled. That way if a player misses games - those games aren't included at all in the data. For example - Boozer gets 12 rebounds in a regulation game while playing 36 minutes. The opposing team had 40 rebounds. The ESTIMATED rebound total of the opponent would be 40/240*36 = 6. Boozer's rate for that game would be 200. Compiling all the player's rebounds for the season & dividing by the compiled estimated opponent's rebounds and multiplying by 100 would give one the player's rate for the season.
That's just my take on the rebound rate idea Mike G.
_________________
Dan
My current ratings (go to the last page for the most recent stuff):
http://www.pointguardu.com/f136/statmans-ratings-56243/
Back to top
View user's profile Send private message Send e-mail
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Thu May 24, 2007 6:34 pm Post subject: Reply with quote
Quote:
By deducting points for missed shots, rather than all shots, PER and EFF actually punish inefficient shooting more severely than WS, not less. These metrics punish inefficient shooting, while WS just punishes shooting in general.
If a FGA is already a lost possession (and an automatic punishment for the scorer), even a FGM is a lost possession, of course a missed FG is also one (the WOW school of no continuation plays). Why scorers must share FGM with the assistant (or assistants) and not the missed FG, or maybe every FGA? What if we charge him (them) a 30%?
Back to top
View user's profile Send private message
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
PostPosted: Thu May 24, 2007 9:01 pm Post subject: Reply with quote
Quote:
But what happens when we compare two players who both take a lot of shots, but differ in shooting efficiency? Let's take player A (8 FGs, 20 FGAs) and player B (12 FGs, 20 FGAs). In this case, Berri actually gives LESS advantage to the efficient shooter than do the other metrics. The Player B advantage over Player A for the 3 metrics is:
WS +8
PER +9.5
EFF +12
Can somebody explain me how 12 missed FG are weighted the same as 8? and the FGM difference is the only thing that counts?
Back to top
View user's profile Send private message
Flint
Joined: 25 Mar 2007
Posts: 112
PostPosted: Fri May 25, 2007 12:27 am Post subject: Reply with quote
Phew. Alright.
Statman - The link is:
http://dberri.wordpress.com/2007/03/08/ ... t-mutombo/
1. The numbers aren't wrong. There is a chart in the above link, outlining their performance through 61 games, which is the sampe in question.
2. The basic argument is that if the Rockets went 20-12 with him filling in for Yao, he must still be pretty good. I dont actually see you responding to this argument, or presenting an alternative theory of what happened. Perhaps Chuck Hayes picked up the slack for Yao? Give Berri credit, at least he presents a very straightforward analysis and allows the data to do the driving.
3. Thank you for the COMPLETE stats. I apologize to anyone who didn't realize that Yao scored much more than Mutombo.
3A. Mark Eaton vs. Michaal Jordan? This little nugget shows that you haven't read the Wages of Wins and don't understand how it works. Berri adjusts for position. You dont compare centers to guards, you compare centers to centers, and shooting guards to shooting guards. Michael Jordan was way above the average for shooting guards. Mark Eaton, not so much relative to other centers.
4. Lee and Mutombo are rated on a per-minute basis. They are very far from the most productive players in the league. That honor went to Jason Kidd this year. He isn't very good I imagine, its just the 8 rebounds per game he collected at the point guard position that makes him look good. That's almost a rebound per game more than Eddy Curry btw. Lee's rebounding stats came from Knickerblogger, but clearly that's wrong since there are two people who have told me so. I got the stats here.
http://www.knickerblogger.net/stats/2007/jh_ALL_REB.htm
At any rate, he is one of the very best rebounders in the league.
We will see about Lee. I think he is that good, having watched him closely all year. Really, he and Balkman brought me back to the Garden and the Knicks. But I wouldnt say we have enough data yet to be sure. Still, its certain that he is much better than Curry, much much better.
What doesl +/- say about Curry btw?
Admin - I would love to know what the accepted wisdom on compensation in the NBA is. Berri makes a big deal of the fact that there is no correlation between payroll and wins. But there is a salary cap, so I don't know what that means. Until someone convinces me otherwise, I am going to stick with the belief that a lot of NBA Gm's don't have any idea what they are doing. As a Knicks fan, you will have to forgive me for feeling very justified in that assessment.
Gordon - I prefer to think that there is a third option here. Existing metrics discussed so far overrate scoring, in order to pass the laugh test, and Berri has now produced a rating that correctly weighs rebounding and scoring. That metric is not Winscore, which you guys keep referring to, but Wins Produced. Your argument is correct as it relates to WS, which was designed to simply be close enough for gov't work. But in Wins Produced, the baseline is not zero, but the league average for the position. It just isnt true that 100% of players value derive from rebounding, or even from the gaining and maintaining side of the ledger. I made that point already. But there is no question that Berri believes that gaining and maintaining possesion is more important than scoring. I don't really have any problem believing that. Scoring is what catches the eye but its impact is overvalued.
If someone came along with a statistical method for analyzing offensive and defensive line play, and said that the battle at the line of scrimmage was by far the most important determinate of success in the NFL, I wouldn't find that surprising in the least. But my experience with baseball statistics has made it very easy for me to believe that conventional wisdom is wrong in sports.
Vis a vis whether the ability to take shots at a league average efficiency is a common skill, I would say basically yes. I dont think scoring is really that difficult. I mean, Vince Carter style scoring is difficult, and difficult to watch in the playoffs, but in general I think most NBA teams could still score points without high usage players around. Looked what happend in Philly. Iverson, super high-usage player, decamps for Denver. Miraculously Philadelphia manages to get almost exactly as many shots off as before and score as many points.
Mike G - I just dont follow your ideas. I dont understand this "default" business. You seem to be saying that you think that a 45% shooter has to take shots so that other players can achieve there higher shooting efficiencies, since they couldnt do it without him and his high usage. This has been a consistent theme for you I think. Its another version of the thought that scorers create opportunities by missing shots. Anyway, thats not how I view basketball. Its a team game. It requires scorers and role players, and great role players are every bit as useful as conventional scoring stars.
Eddy Curry is a scorer, no question, but what really is the value of that 60% ts%? Is it worth putting a very poor rebounder out there who was second in the league in turnovers and who had a new statistical category created in his honor, the Eddy Curry line (think mendoza line)? I also love your phrase, "go-to scorer." Go to for a turnover, or directions to the nearest burger king, but certainly not a go-to player if you want to win basketball games.
If you think Eddy Curry is a good, above average basketball player, you are definitely guilty of overvaluing scoring. And I will leave it at that.
Alright. My computer went down, which is why I didn't respond. But I can't keep up here. And I won't convince anyone. So this will be my last post on this thread, Looking up, its not even a very good post, its just too much to bounce back something at each of you. It's been fun.
Back to top
View user's profile Send private message
Statman
Flint wrote:
Phew. Alright.
Statman - The link is:
http://dberri.wordpress.com/2007/03/08/ ... t-mutombo/
1. The numbers aren't wrong. There is a chart in the above link, outlining their performance through 61 games, which is the sampe in question.
2. The basic argument is that if the Rockets went 20-12 with him filling in for Yao, he must still be pretty good. I dont actually see you responding to this argument, or presenting an alternative theory of what happened. Perhaps Chuck Hayes picked up the slack for Yao? Give Berri credit, at least he presents a very straightforward analysis and allows the data to do the driving.
3. Thank you for the COMPLETE stats. I apologize to anyone who didn't realize that Yao scored much more than Mutombo.
3A. Mark Eaton vs. Michaal Jordan? This little nugget shows that you haven't read the Wages of Wins and don't understand how it works. Berri adjusts for position. You dont compare centers to guards, you compare centers to centers, and shooting guards to shooting guards. Michael Jordan was way above the average for shooting guards. Mark Eaton, not so much relative to other centers.
4. Lee and Mutombo are rated on a per-minute basis. They are very far from the most productive players in the league. That honor went to Jason Kidd this year. He isn't very good I imagine, its just the 8 rebounds per game he collected at the point guard position that makes him look good. That's almost a rebound per game more than Eddy Curry btw. Lee's rebounding stats came from Knickerblogger, but clearly that's wrong since there are two people who have told me so. I got the stats here.
http://www.knickerblogger.net/stats/2007/jh_ALL_REB.htm
At any rate, he is one of the very best rebounders in the league.
We will see about Lee. I think he is that good, having watched him closely all year. Really, he and Balkman brought me back to the Garden and the Knicks. But I wouldnt say we have enough data yet to be sure. Still, its certain that he is much better than Curry, much much better.
What doesl +/- say about Curry btw?
Admin - I would love to know what the accepted wisdom on compensation in the NBA is. Berri makes a big deal of the fact that there is no correlation between payroll and wins. But there is a salary cap, so I don't know what that means. Until someone convinces me otherwise, I am going to stick with the belief that a lot of NBA Gm's don't have any idea what they are doing. As a Knicks fan, you will have to forgive me for feeling very justified in that assessment.
Gordon - I prefer to think that there is a third option here. Existing metrics discussed so far overrate scoring, in order to pass the laugh test, and Berri has now produced a rating that correctly weighs rebounding and scoring. That metric is not Winscore, which you guys keep referring to, but Wins Produced. Your argument is correct as it relates to WS, which was designed to simply be close enough for gov't work. But in Wins Produced, the baseline is not zero, but the league average for the position. It just isnt true that 100% of players value derive from rebounding, or even from the gaining and maintaining side of the ledger. I made that point already. But there is no question that Berri believes that gaining and maintaining possesion is more important than scoring. I don't really have any problem believing that. Scoring is what catches the eye but its impact is overvalued.
If someone came along with a statistical method for analyzing offensive and defensive line play, and said that the battle at the line of scrimmage was by far the most important determinate of success in the NFL, I wouldn't find that surprising in the least. But my experience with baseball statistics has made it very easy for me to believe that conventional wisdom is wrong in sports.
Vis a vis whether the ability to take shots at a league average efficiency is a common skill, I would say basically yes. I dont think scoring is really that difficult. I mean, Vince Carter style scoring is difficult, and difficult to watch in the playoffs, but in general I think most NBA teams could still score points without high usage players around. Looked what happend in Philly. Iverson, super high-usage player, decamps for Denver. Miraculously Philadelphia manages to get almost exactly as many shots off as before and score as many points.
Mike G - I just dont follow your ideas. I dont understand this "default" business. You seem to be saying that you think that a 45% shooter has to take shots so that other players can achieve there higher shooting efficiencies, since they couldnt do it without him and his high usage. This has been a consistent theme for you I think. Its another version of the thought that scorers create opportunities by missing shots. Anyway, thats not how I view basketball. Its a team game. It requires scorers and role players, and great role players are every bit as useful as conventional scoring stars.
Eddy Curry is a scorer, no question, but what really is the value of that 60% ts%? Is it worth putting a very poor rebounder out there who was second in the league in turnovers and who had a new statistical category created in his honor, the Eddy Curry line (think mendoza line)? I also love your phrase, "go-to scorer." Go to for a turnover, or directions to the nearest burger king, but certainly not a go-to player if you want to win basketball games.
If you think Eddy Curry is a good, above average basketball player, you are definitely guilty of overvaluing scoring. And I will leave it at that.
Alright. My computer went down, which is why I didn't respond. But I can't keep up here. And I won't convince anyone. So this will be my last post on this thread, Looking up, its not even a very good post, its just too much to bounce back something at each of you. It's been fun.
Well - I'm annoyed - just lost a large post I made. Oh well - good debate.
Suffice to say that I believe that Berri's system of adjusting scoring off of league norm in terms of efficiency while seemingly adjusting everything else off of mean zero comes out flawed. Scoring is way undervalued, while rebounding becomes WAY overvalued because of their prevelance.
when the data comes out - I'll BET that the most "underrated" players according to Berri will be guys that don't shoot much and are good to great rebounders for their positions - across the board. Just a guess Wink
_________________
Dan
page 8 missing