Alternate player efficiency ranking (Mike G, 2009)
Posted: Fri Apr 15, 2011 4:14 am
page 1
Author Message
Mike G
PostPosted: Mon Nov 16, 2009 7:56 am Post subject: Reply with quote
What could you do to PER such that a team's PER is commensurate with it's point differential?
I see that of GSW players, filtered to 15+ mpg and 40+ G, 9 of 10 have a PER >14.2, 4 are >16. Your adjustments don't really change the overall picture of a well-above-average team. Yet they were 29-53 last year.
Their offense may have been above league average, independent of the context of their opponents. And you could argue that on a given team, PER ranks players about right. But between defensive and defenseless teams, there's no correction.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
supersub15
PostPosted: Mon Nov 16, 2009 8:58 am Post subject: Reply with quote
Mike G wrote:
What could you do to PER such that a team's PER is commensurate with it's point differential?
I see that of GSW players, filtered to 15+ mpg and 40+ G, 9 of 10 have a PER >14.2, 4 are >16. Your adjustments don't really change the overall picture of a well-above-average team. Yet they were 29-53 last year.
Their offense may have been above league average, independent of the context of their opponents. And you could argue that on a given team, PER ranks players about right. But between defensive and defenseless teams, there's no correction.
It has the same shortcomings as Hollinger's PER. But your question is valid: What could you do to PER such that a team's PER is commensurate with it's point differential?
Back to top
View user's profile Send private message
kjb
PostPosted: Mon Nov 16, 2009 9:44 am Post subject: Reply with quote
A couple years back, in the privacy of my own spreadsheet, I did a simple defensive adjustment to team PER that I then applied to the players on that team. Players on good defensive teams got a bonus; players on bad defensive teams got a penalty.
The process is so simple, it's almost embarrassing Smile -- lg. avg. drtg/team drtg. Then multiply that percentage by the individual player PER and voila -- defense adjusted PER.
Gilbert Arenas has a PER of 19.2, the Wiz have a drtg of 107.2 and league average efficiency is 105.5.
So, it's 105.5/107.2 * 19.2 = 18.9.
At the team level, the results end up correlating extremely well with winning percentage.
Smart people here can probably come up with something much better. This was something simple I did for myself a couple years back.
Back to top
View user's profile Send private message AIM Address Yahoo Messenger
Joe
PostPosted: Mon Nov 16, 2009 10:28 am Post subject: Reply with quote
PER already correlates at about an 85% level with wins, higher than AWS, WS, and EFF (based on some quick excel work I did on the 2009 season). Hollinger has written in these forums before he doesn't believe assigning team credit for defense to players is a good idea because it's going to unfairly distribute credit based on the performance of teammates. Ironically, that's exactly what I got away from by fixing the assist adjustment, but that's only because it's actually possible to do now, while it wasn't before.
To answer the other questions, to get PER to correlate with team efficiency, you'd need to make it based on actual possessions and apply equal weights to offense and defense in the formula, two things that would basically ruin the formula because it's not possible to accurately measure defense through box score statistics.
And Mike G, we have Team PER listed on our advanced team page, which calculates PER at a team level. And of course the adjustments to the player PER won't change anything at a team level, because all it's doing is redistributing the credit for scoring and passing, while the ultimate output on a team level is still the same.
_________________
http://www.hoopdata.com
Back to top
View user's profile Send private message Visit poster's website
Mike G
PostPosted: Mon Nov 16, 2009 5:46 pm Post subject: Reply with quote
So you aren't looking to 'fix' PER, but just to 'tweak' it a bit?
Joe
PostPosted: Mon Nov 16, 2009 6:03 pm Post subject: Reply with quote
Until someone reinvents the way basic box score statistics are tracked, specifically to make the accuracy and complexity of defensive stats on the same level as offensive stats, it's impossible to 'fix' PER in the way you want it to be fixed. Giving equal weight to defensive stats that are far less accurate in explanatory power is not a solution that would 'fix' PER. Likewise, neither is distributing credit for team defense to individual players regardless of if they deserve it.
All I'm looking to do is modify PER to account for things that are trackable, in addition to the things it already tracks well. One idea is distinguishing between free throws that account for possessions and ones that don't, rather than using the 0.44 estimator, a percentage which varies from player to player (though not on the same level as the %Assist rate was varying). If you want me to modify PER to make it some kind of end-all, be-all stat that matches up to efficiency differential, it's not happening, because it's not possible to do it accurately with the data at public disposal.
Mike G
PostPosted: Mon Nov 16, 2009 6:18 pm Post subject: Reply with quote
One stat that is tracked is Opponent Pts/G (or Pts/Poss). Just as a player can have a Reb% that is scaled to Tm and Opp Reb (not necessarily in equal measure), we can scale a player's Scoring to Tm and/or Opp Scoring.
In other words, 20 points in a 115 PPG environment is not equal to 20 points in 90. This seems pretty obvious; just as getting 10 Reb out of 50 is not as good as 10 of 35.
You wouldn't have to scale a player's whole PER to the available points, just his points. There is a 'scoring' component, isn't there?
Crow
PostPosted: Mon Nov 16, 2009 6:30 pm Post subject: Reply with quote
If you don't think you can fix PER to weight defense equally and accurately then I'd vote for an adjusted, strictly offensive PER.
The slightly adjusted for assisted rate and charges is an improvement over the classic version but I'd probably be much more regular in using an offensive PER.
I'd mainly use defensive Adjusted +/- for the other half of the game.
Back to top
View user's profile Send private message
Joe
PostPosted: Mon Nov 16, 2009 7:00 pm Post subject: Reply with quote
Mike G wrote:
One stat that is tracked is Opponent Pts/G (or Pts/Poss). Just as a player can have a Reb% that is scaled to Tm and Opp Reb (not necessarily in equal measure), we can scale a player's Scoring to Tm and/or Opp Scoring.
In other words, 20 points in a 115 PPG environment is not equal to 20 points in 90. This seems pretty obvious; just as getting 10 Reb out of 50 is not as good as 10 of 35.
You wouldn't have to scale a player's whole PER to the available points, just his points. There is a 'scoring' component, isn't there?
Rebounds are opportunities from missed field goals (and free throws), not total rebounds grabbed in a game. The total rebounds is just used as a quick estimate because rebounds and missed field goals tend to correlate very highly. Likewise, points are opportunities from possessions used, not total points scored in a game. Using points as a percentage of total points scored makes no sense, because possessions used and points vary wildly, plus we already can accurately estimate possessions, so why use something that makes less sense instead? PER is already based highly in possessions, with one notable exception, which I (and probably Hollinger) doesn't think can change without skewing the balance of offensive to defensive weights, thus ruining its usefulness.
Crow, that's a good idea, and something I've looked into.
___________
deepak
PostPosted: Mon Nov 16, 2009 7:22 pm Post subject: Reply with quote
Any plans to factor in charges into your APER, since you are tracking them?
Back to top
View user's profile Send private message
Joe
PostPosted: Mon Nov 16, 2009 7:26 pm Post subject: Reply with quote
From first post in this thread:
Quote:
I also added charges into the formula with the same weight as steals, which makes sense given they have the same exact effect in regards to possessions, and there's no reason not to account for them since they are now recorded.
Very Happy
_
deepak
PostPosted: Mon Nov 16, 2009 9:27 pm Post subject: Reply with quote
Don't know how I missed that.
I wonder, though, if a steal is really the same value as a charge. It's true that they both result in a defensive stop, but steals also typically lead to transition opportunities (hence, better offensive possessions). On the other hand, one could argue that players who collect a lot of steals also tend to to gamble more, so maybe that cancels out the offensive benefits on average.
Should such things be taken into account in these linear weights ratings?
Joe
PostPosted: Mon Nov 16, 2009 9:38 pm Post subject: Reply with quote
One of our plans is to eventually make a stat that accounts for steals that lead to made baskets within 5 seconds or so, something we could possibly modify the APER formula with when we do it. Another thing is distinguishing from blocks recovered by offense and blocks recovered by defense, which could also be modified in the APER.
The problem with changing the value of steals based on what they lead to, however, is technically, if you're going to give a player more than the value of a possession for stealing a ball, you really should be subtracting that extra value from scorers on the team, because you're suggesting that the credit for the scores steals lead to belong in part to the thief. It doesn't make sense to distribute the credit twice, such as statistics like Win Score over-counting assists by not deducting them from scorers. Then again, PER doesn't follow a strict possession formula, so it is something that can be considered.
_________________
http://www.hoopdata.com
Last edited by Joe on Mon Nov 16, 2009 9:41 pm;
ecumenopolis0
PostPosted: Mon Nov 16, 2009 9:40 pm Post subject: Reply with quote
deepak_e wrote:
Don't know how I missed that.
I wonder, though, if a steal is really the same value as a charge. It's true that they both result in a defensive stop, but steals also typically lead to transition opportunities (hence, better offensive possessions). On the other hand, one could argue that players who collect a lot of steals also tend to to gamble more, so maybe that cancels out the offensive benefits on average.
Should such things be taken into account in these linear weights ratings?
And of course, obvious note, taking a charge also puts a foul on the opponent.
page 2 of 5
Author Message
Crow
PostPosted: Tue Nov 17, 2009 1:52 am Post subject: Reply with quote
"I looked into modifying the accounting of positive possessions used (changing how PER doesn't deduct for possessions used on makes after accounting for the added value in points), but the results were wildly out of whack with reality, and I now understand why Hollinger weighed the formula the way he did."
I assume this is the same thing as looking at the break even point for FG% efficiency? Correct me if I am wrong.
I'd be interested in a look at the results where the break-even point were shifted 1/3 to 1/2 way towards what Wins Produced does. Or at least some more discussion of it if you wanted to.
If you keep the defensive stats would you consider increasing the weight for defensive rebounds 1/3 to 1/2 way towards what Wins Produced does or towards what the regression you trust says, whatever that is? Defensive rebounding is an acknowledged imperfect hole-filler for defense but I wonder (without rerunning all the debate and data thru my mind) if the defensive rebounding weight in PER is on its own, apart from that rationale, too low.
Would you change the way fouls are treated to less negative?
Are you satisfied with the level of credit to players for assists?
ultimately do you want to stay within adjusting PER fairly narrowly or could you reach a point where you decided to try to make essentially a new metric, perhaps a compromise metric?
Back to top
View user's profile Send private message
Mike G
PostPosted: Tue Nov 17, 2009 9:41 am Post subject: Reply with quote
Joe wrote:
Rebounds are opportunities from missed field goals (and free throws), not total rebounds grabbed in a game. The total rebounds is just used as a quick estimate ...
Do you have this backwards? You can estimate rebounds from rebound-opportunities. Rebounds are actually counted.
Is it more informative to know that your team got 55% of all potential rebounds, than to know your team got 55% of all actual rebounds?
Quote:
.. Likewise, points are opportunities from possessions used, not total points scored in a game. Using points as a percentage of total points scored makes no sense, because possessions used and points vary wildly, plus we already can accurately estimate possessions,..
So if your team averages 20 points per position, what does the possession count mean? Isn't it more meaningful that the opponent averages more or less than 20 per position?
You might be getting outscored by 110-100, due to high pace and poor shooting, low pace and poor defense, or whatever. Either way, your average position is getting beat 22-20.
Pts/Poss * Poss/G = Pts/G
Scoring 100 is very good when you allow 90; it's very weak vs 110.
This makes sense to me.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Joe
PostPosted: Tue Nov 17, 2009 10:30 am Post subject: Reply with quote
Crow wrote:
"I looked into modifying the accounting of positive possessions used (changing how PER doesn't deduct for possessions used on makes after accounting for the added value in points), but the results were wildly out of whack with reality, and I now understand why Hollinger weighed the formula the way he did."
I assume this is the same thing as looking at the break even point for FG% efficiency? Correct me if I am wrong.
I'd be interested in a look at the results where the break-even point were shifted 1/3 to 1/2 way towards what Wins Produced does. Or at least some more discussion of it if you wanted to.
If you keep the defensive stats would you consider increasing the weight for defensive rebounds 1/3 to 1/2 way towards what Wins Produced does or towards what the regression you trust says, whatever that is? Defensive rebounding is an acknowledged imperfect hole-filler for defense but I wonder (without rerunning all the debate and data thru my mind) if the defensive rebounding weight in PER is on its own, apart from that rationale, too low.
Would you change the way fouls are treated to less negative?
Are you satisfied with the level of credit to players for assists?
ultimately do you want to stay within adjusting PER fairly narrowly or could you reach a point where you decided to try to make essentially a new metric, perhaps a compromise metric?
The weights Hollinger uses for about everything in his formula are sound and in line with logic, which likewise align with the weights Rosenbaum used (for the most part) in his alternate win score. The weights on rebounds are perfect, with offensive rebounds weighing the defensive rebound percentage and defensive rebounds weighing the offensive rebound percentage. Changing those arbitrarily to something else to make the numbers more like win score? I don't understand why one would want to do that.
I agree with the assist weights, and they are close to in line with the weights Oliver and Pelton use in their metrics.
Fouls are whatever. Offensive fouls are being double counted in the current equation, but changing the foul portion of the equation probably would have marginal value and I don't really agree or disagree strongly with how they are used.
Quote:
Do you have this backwards? You can estimate rebounds from rebound-opportunities. Rebounds are actually counted.
Is it more informative to know that your team got 55% of all potential rebounds, than to know your team got 55% of all actual rebounds?
You're agreeing with me.
Quote:
So if your team averages 20 points per position, what does the possession count mean? Isn't it more meaningful that the opponent averages more or less than 20 per position?
No. It matters how many points they average per possession, as possessions are the opportunities to score points. If a shooting guard scores 15 points on 5 possessions, he is providing incredible efficiency, but by your idea, he'd be below average, which is ludicrous.
Quote:
You might be getting outscored by 110-100, due to high pace and poor shooting, low pace and poor defense, or whatever. Either way, your average position is getting beat 22-20.
Pts/Poss * Poss/G = Pts/G
Scoring 100 is very good when you allow 90; it's very weak vs 110.
This makes sense to me.
This is the logic for using possession-based costs in metrics, and it makes sense. Measuring points as a percentage of total points doesn't make sense for the reason outlined above.
_________________
http://www.hoopdata.com
Back to top
View user's profile Send private message Visit poster's website
Mike G
PostPosted: Tue Nov 17, 2009 10:44 am Post subject: Reply with quote
Joe,
Perhaps you are confusing 'position' and 'possession'.
I apologize for introducing this possibility. You might re-read with this clarification.
A 'position' is 1/5 of a team's performance; generally 48 player-minutes.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Joe
PostPosted: Tue Nov 17, 2009 10:48 am Post subject: Reply with quote
Mike G wrote:
Joe,
Perhaps you are confusing 'position' and 'possession'.
I apologize for introducing this possibility. You might re-read with this clarification.
A 'position' is 1/5 of a team's performance; generally 48 player-minutes.
You said 'Poss' - that is the abbreviation for possession. Position doesn't have two consecutive S's at any point in it, so there's no way you could've been referring to that.
Regardless, virtually every noteworthy composite metric out there uses possession-based costs for scoring opportunities as opposed to measuring as a function of 1/5 points scored, and I've explained the reasoning for it. I'm not the one who is confused here.
_________________
Mike G
C
PostPosted: Tue Nov 17, 2009 11:19 am Post subject: Reply with quote
I see I wrote "position" 3 times, "possession" once, and it's abbreviation, "Poss", twice.
Again apologize if this was distracting.
The idea was to suggest that players sharing a position combine for 1/5 of a team's performance, and that this performance (whether or not you call it 'efficiency') really is relative to the opposition.
A team may score 160 points in 100 possessions, and allow 165. This is not as effective as scoring 80 while allowing 75. In any number of possessions.
Since # of possessions is generally assumed to be equal for both a team and its opponent, it seems reasonable that this quantity cancels in an 'effectiveness' formula. To win, you still have to score >50% of the points in a game, which may consist of 80 or 100 or 120 possessions.
If you've been outscored by 100-90, then whatever efficiency you feel you've achieved (on a per-possession basis) could be said to be only 90/100 (or .90) as effective as it would be if the score were even. This is equally true as a per-game average or a 3rd-quarter recap.
If you scale the scoring portion of your metric to (TmPts/OpPts), then you primarily affect the 'scoring specialist' types, while basically leaving the 'defensive specialist' alone.
_________________
Joe
PostPosted: Tue Nov 17, 2009 11:39 am Post subject: Reply with quote
Mike G wrote:
I see I wrote "position" 3 times, "possession" once, and it's abbreviation, "Poss", twice.
Again apologize if this was distracting.
The idea was to suggest that players sharing a position combine for 1/5 of a team's performance, and that this performance (whether or not you call it 'efficiency') really is relative to the opposition.
A team may score 160 points in 100 possessions, and allow 165. This is not as effective as scoring 80 while allowing 75. In any number of possessions.
Since # of possessions is generally assumed to be equal for both a team and its opponent, it seems reasonable that this quantity cancels in an 'effectiveness' formula. To win, you still have to score >50% of the points in a game, which may consist of 80 or 100 or 120 possessions.
If you've been outscored by 100-90, then whatever efficiency you feel you've achieved (on a per-possession basis) could be said to be only 90/100 (or .90) as effective as it would be if the score were even. This is equally true as a per-game average or a 3rd-quarter recap.
If you scale the scoring portion of your metric to (TmPts/OpPts), then you primarily affect the 'scoring specialist' types, while basically leaving the 'defensive specialist' alone.
Here's the problem with this line of thinking (and I've yet to see you illustrate the problem with the alternative, which is currently widely used):
A shooting guard and a small forward each play 48 minutes in an imaginary game where their team scores 100 points. The shooting guard scores 20 and the small forward scores 20, however the shooting guard gets his scoring by hitting 10-for-10 FGA and the small forward gets his scoring by hitting 10-of-30 FGA. One used three times as many possessions as the other, yet they still scored the same points. By your scaling, they'd be considered equal contributors, when in reality, one basically contributed 20 points + 10 free possessions (if you assume the EV of a possession to be 1), and the other contributed 20 points - 10 extra possessions used over the EV. Their scoring numbers are equal, but in terms of roundabout impact to scoring, assuming possessions have an EV of 1, they actually have a difference of about 20 expected points, making them basically the difference between an All-NBA first teamer and a player who doesn't deserve to be in the league. Yet your scaling would consider them equal.
This is why possession-based costs is the norm, and there is no disadvantage to it. I don't see what kind of value using your suggested scale would add, though I do see a ton of problems with it.
_________________
http://www.hoopdata.com
Back to top
View user's profile Send private message Visit poster's website
Mike G
PostPosted: Tue Nov 17, 2009 12:06 pm Post subject: Reply with quote
Joe wrote:
... By your scaling, they'd be considered equal contributors....
Not in any way, shape, or form.
Here, you've pulled the switcheroo between 'points scored' and "the scoring portion of your metric ".
You've also not considered what the opposition has done while these guys are shooting 10-10 and 10-30.
If the opposition shot 25% for the game, then 10-30 is pretty good. If they shot 40% or better, it's bad.
A couple of years ago, someone 'demonstrated' that Michael Jordan hadn't (contrary to popular belief) dominated the Finals in '97 and '98; and in fact he'd shot badly. Not badly enough to lose the games, but badly relative to league average, Jordan average, etc.
In fact, his opponents (the Jazz) shot much worse; and Jordan's teammates shot worse than he. So he did, in fact (and concurring with popular belief) dominate those Finals.
In another playoff series, hitting 10-20 shots is not nearly good enough to keep up with the opposition. Ignore this at your peril.
Joe
PostPosted: Tue Nov 17, 2009 12:34 pm Post subject: Reply with quote
Mike G wrote:
Joe wrote:
... By your scaling, they'd be considered equal contributors....
Not in any way, shape, or form.
Here, you've pulled the switcheroo between 'points scored' and "the scoring portion of your metric ".
You've also not considered what the opposition has done while these guys are shooting 10-10 and 10-30.
If the opposition shot 25% for the game, then 10-30 is pretty good. If they shot 40% or better, it's bad.
A couple of years ago, someone 'demonstrated' that Michael Jordan hadn't (contrary to popular belief) dominated the Finals in '97 and '98; and in fact he'd shot badly. Not badly enough to lose the games, but badly relative to league average, Jordan average, etc.
In fact, his opponents (the Jazz) shot much worse; and Jordan's teammates shot worse than he. So he did, in fact (and concurring with popular belief) dominate those Finals.
In another playoff series, hitting 10-20 shots is not nearly good enough to keep up with the opposition. Ignore this at your peril.
I'm not ignoring anything. I already addressed earlier in this thread that distributing team defense to individual players is something that Hollinger himself has stated on this forum to be something he doesn't want to do, as it is not necessarily accurate. You're giving credit where it isn't necessarily deserved.
_________________
http://www.hoopdata.com
Back to top
View user's profile Send private message Visit poster's website
DLew
PostPosted: Tue Nov 17, 2009 1:33 pm Post subject: Reply with quote
I think that the adjustments made to PER are a good use of the new data, but that going much further (i.e. crediting the offense created by steals to the person who got the steal, etc) probably runs the risk of creating a new metric that shouldn't really be called adjusted PER.
Back to top
View user's profile Send private message
Joe
PostPosted: Tue Nov 17, 2009 1:39 pm Post subject: Reply with quote
DLew wrote:
I think that the adjustments made to PER are a good use of the new data, but that going much further (i.e. crediting the offense created by steals to the person who got the steal, etc) probably runs the risk of creating a new metric that shouldn't really be called adjusted PER.
I agree.
_________________
http://www.hoopdata.com
Back to top
View user's profile Send private message Visit poster's website
Mike G
Joined: 14 Jan 2005
Posts: 3564
Location: Hendersonville, NC
PostPosted: Tue Nov 17, 2009 1:41 pm Post subject: Reply with quote
Giving team-average defensive credit to everyone on a team (as kjb described) wouldn't change players' ranking on a team, relative to one another; it would only change their rankings relative to those on other teams.
Because it would also adjust the team relative to their actual strength, that would be an improvement, which also appears with the individuals' ratings. Some accuracy is gained, and nothing is lost, in this blanket adjustment.
Trying for more, one could suggest docking scoring specialists more heavily on poor defensive teams, relative to low scorers on good defensive teams. Again, all gain and no loss, in ranking player effectiveness.
None of this affects a player's per-possession effectiveness. The average possession itself has a value adjustment. Specifically, it's relative to TmPts/OppPts.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Crow
Joined: 20 Jan 2009
Posts: 806
PostPosted: Tue Nov 17, 2009 2:21 pm Post subject: Reply with quote
I don't think the FG% breakpoint is quite 'best" or the weight on defensive rebounds is "perfect" from all angles and I also disagree with the large size of the assist weight and fouls remain an area that could possibly be improved or dropped given the difficulty but I am not going to argue them now. I was asking you to clarify if you were sure what you were going to do and you did.
.
Last edited by Crow on Tue Nov 17, 2009 3:02 pm; edited 2 times in total
Back to top
View user's profile Send private message
kjb
Joined: 03 Jan 2005
Posts: 864
Location: Washington, DC
PostPosted: Tue Nov 17, 2009 2:29 pm Post subject: Reply with quote
There is an argument in favor of simply using team defensive rating as a blanket adjustment -- namely that defense is primarily about teamwork.
I can see the other side of the argument too -- that some players inevitably end up with too much or too little credit for the team's defensive performance. Brendan Haywood, who has been a good defender on a horrible defensive team, immediately springs to mind as someone who would be "unfairly" penalized because of his teammates' poor performance. I'm pretty ambivalent on the issue. The simple adjustment I used sufficed for my purposes, even recognizing all of its limitations.
Another option might be to use the on/off defensive data at 82games or one of the other pbp sites. Still not perfect, but perhaps an advancement over the "simple" technique I described earlier. I don't have the programming skills to rip that data from those sites; nor do I have time to do all the data entry by hand.
Adding in the defensive adjustment I described boosted correlation with efficiency differential from .85 to .95ish (I don't have that spreadsheet with me today to check the exact correlation).
Back to top
View user's profile Send private message AIM Address Yahoo Messenger
Joe
Joined: 27 Sep 2009
Posts: 94
Location: Long Island, NY
PostPosted: Tue Nov 17, 2009 3:14 pm Post subject: Reply with quote
Crow wrote:
I don't think the FG% breakpoint is quite 'best" or the weight on defensive rebounds is "perfect" from all angles and I also disagree with the large size of the assist weight and fouls remain an area that could possibly be improved or dropped given the difficulty but I am not going to argue them now. I was asking you to clarify if you were sure what you were going to do and you did.
.
Assists are subjective, always have been, always will be. Hollinger, Oliver, and Pelton all are between 2/3 to 3/4 a point for each field goal, however, and I have nothing substantive to argue against that.
The FG% breakpoint (and all scoring efficiency breakpoints) in PER are not ideal, because like I said, there is no deduction for possessions used on positive possessions. A two point field goal yields two points, while it should yield two points minus the VOP. The reason for this is the accuracy and complexity of defensive statistics compared to offensive statistics. By not deducting for possessions on made field goal attempts, it increases the relative weight of offense (namely scoring) in the equation, which is necessary because the defensive statistics shouldn't weigh so much given their poor explanatory power. This is why your suggestion of making an offense-only PER is a good idea, because this issue would no longer exist, and then you could just measure offense, which could be done with 99%+ accuracy.
Fouls are a tricky subject. Don't really know the best way to approach those.
kjb, it's not a bad idea, and I get its merits, but if I want to keep calling it APER, I don't think doing something that Hollinger has clearly stated on these forums that he wouldn't do is a good idea. If ESPN tracked assisted field goals and charges, and every other website using PER also did, I'm fairly certain Hollinger would've made these adjustments I did a long time ago, but it's kind of hard to change a statistic that is so widely used. I don't think anything I did changed the spirit of PER, but your adjustment would. Not that it's a bad idea, just not what PER is.
_________________
http://www.hoopdata.com
page 3
Author Message
Crow
Joined: 20 Jan 2009
Posts: 775
PostPosted: Tue Nov 17, 2009 4:15 pm Post subject: Reply with quote
Thanks for your attention to making replies to comments and questions generally. A good policy.
I'd support some form of defensive adjustment to PER or just offensive PER. Slight modified PER mostly the way it has been is ok too but I'd prefer one of the former choices over it.
On assists I 'll add that I did see Neil Paine's regression results where an A per 40 minutes was worth about 1/10th value of a turnover/40 (with the opposite sign of course)
http://www.basketball-reference.com/blog/?p=2191
Fouls are indicated as a moderate positive.
But his results even further diminished the value of defensive rebounds so I don't know if I can reasonably just take the parts I agree with.
Back to top
View user's profile Send private message
DLew
Joined: 13 Nov 2006
Posts: 222
PostPosted: Tue Nov 17, 2009 5:57 pm Post subject: Reply with quote
You really can't interpret those results that way because of the interaction term in that regression (pts*reb*ast). Some of the value of points, rebounds, and assists is being represented in that coefficient.
Back to top
View user's profile Send private message
Crow
Joined: 20 Jan 2009
Posts: 775
PostPosted: Tue Nov 17, 2009 6:47 pm Post subject: Reply with quote
True.
And that term can capture some of the value of defensive rebounds too.
I shouldn't have claimed simple support from this reference.
I was previously aware of the impact of the versatility index and minutes too.
But if your re-ran the regression without the versatility index or minutes would the assist weight be as high relatively as in PER? I'd like to know. I'd guess a decent to good chance it wouldn't be. Assists is only about 15% of the input to the versatility index on average.
Back to top
View user's profile Send private message
jkubatko
Joined: 05 Jan 2005
Posts: 702
Location: Columbus, OH
PostPosted: Wed Nov 18, 2009 8:08 am Post subject: Reply with quote
Joe wrote:
PER already correlates at about an 85% level with wins, higher than AWS, WS, and EFF (based on some quick excel work I did on the 2009 season).
That's not true. Yes, the correlation between team wins and team PER in 2008-09 was 0.855, but the correlation between team wins and team Win Shares was 0.988.
_________________
Regards,
Justin Kubatko
Basketball-Reference.com
Back to top
View user's profile Send private message Send e-mail Visit poster's website
DLew
Joined: 13 Nov 2006
Posts: 222
PostPosted: Wed Nov 18, 2009 11:12 am Post subject: Reply with quote
Justin,
I'm pretty sure that when he wrote "WS" he was referring to Win Score.
Back to top
View user's profile Send private message
Joe
Joined: 27 Sep 2009
Posts: 94
Location: Long Island, NY
PostPosted: Wed Nov 18, 2009 11:43 am Post subject: Reply with quote
DLew wrote:
Justin,
I'm pretty sure that when he wrote "WS" he was referring to Win Score.
Correct.
_________________
http://www.hoopdata.com
Back to top
View user's profile Send private message Visit poster's website
Mike G
Joined: 14 Jan 2005
Posts: 3504
Location: Hendersonville, NC
PostPosted: Wed Nov 18, 2009 4:03 pm Post subject: Reply with quote
Joe wrote:
Quote:
So if your team averages 20 points per position, what does the possession count mean? Isn't it more meaningful that the opponent averages more or less than 20 per position?
No. It matters how many points they average per possession....
Last year, the Mavs met the Spurs in the opening playoff round. Dal averaged 1.11 points per possession, and they won the series 4-1. Avg pt-diff +7.0 .
In round 2, they met the Nuggets and again averaged 1.11 Pts/Poss. This series they lost 4-1, by -7.4 PPG.
It doesn't matter what the league average Pts/Poss was in 2008-09, over 1230 NBA games. They won or lost due to their Pts/Poss being significantly more/less than that of their playoff opponents', for 5 games each.
In the Dal-SA series, an average possession netted 1.075 points; in the Dal-Den series, it was 1.155 . These were the marks they needed to beat, to likely win these series.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Joe
Joined: 27 Sep 2009
Posts: 94
Location: Long Island, NY
PostPosted: Wed Nov 18, 2009 5:43 pm Post subject: Reply with quote
Mike G wrote:
Joe wrote:
Quote:
So if your team averages 20 points per position, what does the possession count mean? Isn't it more meaningful that the opponent averages more or less than 20 per position?
No. It matters how many points they average per possession....
Last year, the Mavs met the Spurs in the opening playoff round. Dal averaged 1.11 points per possession, and they won the series 4-1. Avg pt-diff +7.0 .
In round 2, they met the Nuggets and again averaged 1.11 Pts/Poss. This series they lost 4-1, by -7.4 PPG.
It doesn't matter what the league average Pts/Poss was in 2008-09, over 1230 NBA games. They won or lost due to their Pts/Poss being significantly more/less than that of their playoff opponents', for 5 games each.
In the Dal-SA series, an average possession netted 1.075 points; in the Dal-Den series, it was 1.155 . These were the marks they needed to beat, to likely win these series.
So let's say Chris Paul is on an awful team that allows 1.1 points per possession on defense, which would be good for worst in the league most seasons. You're telling me that just because Chris Paul's teammates are awful on defense, he should be expected to do more per possession than Lebron James on a defense that is allowing 1.0 points per possession, even if Paul and James are about equal in defensive effectiveness, something that can't be measured objectively through statistics with accuracy, or at least not through box score statistics. You really think this is a good idea and would improve PER? Holding people to different standards of expected value based on things that are completely out of their control?
If you do think that is a good idea, I think we'll just have to agree to disagree there.
_________________
http://www.hoopdata.com
Back to top
View user's profile Send private message Visit poster's website
Mike G
Joined: 14 Jan 2005
Posts: 3504
Location: Hendersonville, NC
PostPosted: Thu Nov 19, 2009 10:46 am Post subject: Reply with quote
You've adjusted Chris Paul's PER from a league-leading and unprecedented 33.8 to an astronomical 37.6 . Wilt, LeBron, and Jordan topped out around 31.7 .
If you were to factor his team's DRtg to the league average, you might multiply by 1055/1129 (.934), and then his PER would be 31.6; his APER converts to merely 35.1 .
You tell us: Has Paul been merely as good as anyone who's ever played? -- or 11% better, or 19% better?
I don't 'expect' Paul, or Jarret Jack, or anyone else to do more than LeBron. If I ranked them higher than LeBron, then I would.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Joe
Joined: 27 Sep 2009
Posts: 94
Location: Long Island, NY
PostPosted: Thu Nov 19, 2009 11:05 am Post subject: Reply with quote
Mike G wrote:
You've adjusted Chris Paul's PER from a league-leading and unprecedented 33.8 to an astronomical 37.6 . Wilt, LeBron, and Jordan topped out around 31.7 .
If you were to factor his team's DRtg to the league average, you might multiply by 1055/1129 (.934), and then his PER would be 31.6; his APER converts to merely 35.1 .
You tell us: Has Paul been merely as good as anyone who's ever played? -- or 11% better, or 19% better?
I don't 'expect' Paul, or Jarret Jack, or anyone else to do more than LeBron. If I ranked them higher than LeBron, then I would.
Wow, you really like using misleading statistics, don't you? Lebron had a higher APER than Paul last season. You think this season's 10 game sample size is a more reliable indicator than that? Variation of course could not be a factor in a 10 game sample size, could it?
We also don't have access to Jordan's %Assisted number, so there's no telling how high he would've been, but his number was probably in line with Kobe's most years, which would likely bump his 31.7 to 33+. You think comparing a 10 game sample size of Paul, using a different formula, to Jordan's old PER number, is a fair comparison? Really?
Once again, I'm simply modifying the formula in ways that likely would've been done a long time ago, were the statistics available. The whole point of the part of the formula I modified was to estimate how many shots a player created on his own. No longer does it need to be estimated. I'm sorry the numbers don't jive with your expectations or subjective analysis of Paul's game, but quite frankly I don't care about either of those things. And once again, Hollinger has stated on these very forums he has never intended to apply team defensive numbers to individual defensive performance in PER, due to accuracy concerns. Why do you keep bringing this up?
And for the record, Paul was assisted on only 14.3% of his shots last season. Only two other players in the league were below 21.5% (Nash and Roko Ukic, though the latter's scoring efficiency suggests it probably wasn't a good idea for him -- and neither were below 17%). For a player to use as high a usage as Paul did, score with a monstrous 60% TS%, and only be assisted on 14.3% of his shots could very well be one of the greatest statistical accomplishments in NBA history. We quite frankly have no access to data going back far enough to say for sure, but the three seasons of data we do have access to suggest that his level of shot creation in combination with efficiency puts him in a very, very special class of player.
_________________
http://www.hoopdata.com
Back to top
View user's profile Send private message Visit poster's website
Mike G
Joined: 14 Jan 2005
Posts: 3504
Location: Hendersonville, NC
PostPosted: Thu Nov 19, 2009 11:40 am Post subject: Reply with quote
Joe wrote:
... you really like using misleading statistics, don't you? Lebron had a higher APER than Paul last season. You think this season's 10 game sample size is a more reliable indicator than that? ..
No, and no. I'm only trying to help.
You wrote: "... let's say Chris Paul is on an awful team that allows 1.1 points per possession", and that looks a lot like the Hornets this season. Last year, they were an above-average team defensively.
Quote:
.. to estimate how many shots a player created on his own. No longer does it need to be estimated. I'm sorry the numbers don't jive with your expectations or subjective analysis of Paul's game
No need to be sorry. I've been using %Unassisted as a factor for several years now in my ratings. Yet it's a small adjustment, even in Paul's case, compared to team per-possession differences.
Without adjusting for variations in possession value, the worst case is, as you try to describe, that a good defender among some bad ones is unfairly (dis-) credited. But the counter-argument is that without an adjustment, virtually everyone on a team is unfairly credited.
Almost every player in the league is not playing in a league-average possession value. Why credit every player's average possession relative to something so arbitrary? It changes year to year, and this becomes the standard applied to players; so why not team by team?
In 1991, the Nuggets gave up 1.15 Pts/Poss, while the Pistons gave up 1.046. Michael Adams' PER was 22.3, while Joe Dumars had a career peak 18. Would Adams be expected to substantially outplay Dumars? Would a team correction basically adjust their relative PER's to their relative effectiveness?
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Joe
Joined: 27 Sep 2009
Posts: 94
Location: Long Island, NY
PostPosted: Thu Nov 19, 2009 11:52 am Post subject: Reply with quote
Quote:
You wrote: "... let's say Chris Paul is on an awful team that allows 1.1 points per possession", and that looks a lot like the Hornets this season. Last year, they were an above-average team defensively.
I was using a hypothetical comparing two of the best players in the league. I don't know the Hornets' defensive efficiency numbers for the past two seasons off hand.
Quote:
Without adjusting for variations in possession value, the worst case is, as you try to describe, that a good defender among some bad ones is unfairly (dis-) credited. But the counter-argument is that without an adjustment, virtually everyone on a team is unfairly credited.
Actually, they're not credited at all. PER is an offensively biased statistic, something everyone knows. Anyone looking towards PER to explain team defense is looking at the wrong statistic. That is not its purpose. It is outside its scope.
Quote:
Almost every player in the league is not playing in a league-average possession value. Why credit every player's average possession relative to something so arbitrary? It changes year to year, so why not team to team?
Because the point of PER is to compare players across the league and put them on equal footing. If players' VOP change by up to 10% across the league, it ruins PER as a comparison tool for individual players. It may improve the explanatory power of team performance when adding up the sum of the players, but that's not what PER is meant for. There are other statistics made to do that. Further, like I said earlier, it's not fair to hold players to different standards based on things that are out of their control. It's not an arbitrary number either. For comparing players in the league as a whole, using the league average is far from arbitrary.
Quote:
In 1991, the Nuggets gave up 1.15 Pts/Poss, while the Pistons gave up 1.046. Michael Adams' PER was 22.3, while Joe Dumars' was 18. Would Adams be expected to outplay Dumars? Would a team adjustment basically correct their relative PER's?
Without getting into PER's predictive power for future performance, something I'm not entirely familiar with, I'd say that yes, PER would predict Adams to outplay Dumars, but only in the things that PER actually measures, which don't include team defense. I don't know if a team adjustment would "correct" their relative PER's, because I don't know if each contributed exactly 20% to their team defensive performance when on the floor. There's also a good chance this comparison is hurt by the problems with positive possession usage in PER, something that doesn't properly dock players for poor efficiency, though I don't know Adams' numbers off hand, nor am I familiar with him as a player, as I was five years old at the time of this season.
All of this discussion is lending more credence to Crow's suggestion of creating an offensive-only PER. No matter what method you apply to defensive adjustments in PER, be it using box score statistics with little explanatory power or applying a blanket adjustment assuming everyone is responsible for 20% of team defense, accuracy suffers, and it diminishes the explanatory power and accuracy of the offensive side of the equation, which actually is quite valuable when isolated and adjusted for proper possession usage.
_________________
http://www.hoopdata.com
Back to top
View user's profile Send private message Visit poster's website
DLew
Joined: 13 Nov 2006
Posts: 222
PostPosted: Thu Nov 19, 2009 11:59 am Post subject: Reply with quote
MikeG,
I think you are being a bit unfair here. Your criticisms of PER are on target, but Joe did not invent PER nor is he advocating that it be used as a one number player evaluation. I know when you do your EWins that you want those numbers to completely and totally represent player value, and that's admirable, but PER is not trying to do that, and more importantly Joe is not PER's keeper.
Back to top
View user's profile Send private message
Mike G
Joined: 14 Jan 2005
Posts: 3504
Location: Hendersonville, NC
PostPosted: Thu Nov 19, 2009 12:02 pm Post subject: Reply with quote
Joe wrote:
... If players' VOP change by up to 10% across the league, it ruins PER as a comparison tool for individual players. .
Alright then. We know that VOP varies by this much. Ignoring this doesn't make it go away. Factoring it in fixes it.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Joe
Joined: 27 Sep 2009
Posts: 94
Location: Long Island, NY
PostPosted: Thu Nov 19, 2009 12:03 pm Post subject: Reply with quote
Mike G wrote:
Joe wrote:
... If players' VOP change by up to 10% across the league, it ruins PER as a comparison tool for individual players. .
Alright then. We know that VOP varies by this much. Ignoring this doesn't make it go away. Factoring it in fixes it.
It "fixes" explanatory power at a team level. It "breaks" comparative value across teams.
_________________
http://www.hoopdata.com
Author Message
Mike G
PostPosted: Mon Nov 16, 2009 7:56 am Post subject: Reply with quote
What could you do to PER such that a team's PER is commensurate with it's point differential?
I see that of GSW players, filtered to 15+ mpg and 40+ G, 9 of 10 have a PER >14.2, 4 are >16. Your adjustments don't really change the overall picture of a well-above-average team. Yet they were 29-53 last year.
Their offense may have been above league average, independent of the context of their opponents. And you could argue that on a given team, PER ranks players about right. But between defensive and defenseless teams, there's no correction.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
supersub15
PostPosted: Mon Nov 16, 2009 8:58 am Post subject: Reply with quote
Mike G wrote:
What could you do to PER such that a team's PER is commensurate with it's point differential?
I see that of GSW players, filtered to 15+ mpg and 40+ G, 9 of 10 have a PER >14.2, 4 are >16. Your adjustments don't really change the overall picture of a well-above-average team. Yet they were 29-53 last year.
Their offense may have been above league average, independent of the context of their opponents. And you could argue that on a given team, PER ranks players about right. But between defensive and defenseless teams, there's no correction.
It has the same shortcomings as Hollinger's PER. But your question is valid: What could you do to PER such that a team's PER is commensurate with it's point differential?
Back to top
View user's profile Send private message
kjb
PostPosted: Mon Nov 16, 2009 9:44 am Post subject: Reply with quote
A couple years back, in the privacy of my own spreadsheet, I did a simple defensive adjustment to team PER that I then applied to the players on that team. Players on good defensive teams got a bonus; players on bad defensive teams got a penalty.
The process is so simple, it's almost embarrassing Smile -- lg. avg. drtg/team drtg. Then multiply that percentage by the individual player PER and voila -- defense adjusted PER.
Gilbert Arenas has a PER of 19.2, the Wiz have a drtg of 107.2 and league average efficiency is 105.5.
So, it's 105.5/107.2 * 19.2 = 18.9.
At the team level, the results end up correlating extremely well with winning percentage.
Smart people here can probably come up with something much better. This was something simple I did for myself a couple years back.
Back to top
View user's profile Send private message AIM Address Yahoo Messenger
Joe
PostPosted: Mon Nov 16, 2009 10:28 am Post subject: Reply with quote
PER already correlates at about an 85% level with wins, higher than AWS, WS, and EFF (based on some quick excel work I did on the 2009 season). Hollinger has written in these forums before he doesn't believe assigning team credit for defense to players is a good idea because it's going to unfairly distribute credit based on the performance of teammates. Ironically, that's exactly what I got away from by fixing the assist adjustment, but that's only because it's actually possible to do now, while it wasn't before.
To answer the other questions, to get PER to correlate with team efficiency, you'd need to make it based on actual possessions and apply equal weights to offense and defense in the formula, two things that would basically ruin the formula because it's not possible to accurately measure defense through box score statistics.
And Mike G, we have Team PER listed on our advanced team page, which calculates PER at a team level. And of course the adjustments to the player PER won't change anything at a team level, because all it's doing is redistributing the credit for scoring and passing, while the ultimate output on a team level is still the same.
_________________
http://www.hoopdata.com
Back to top
View user's profile Send private message Visit poster's website
Mike G
PostPosted: Mon Nov 16, 2009 5:46 pm Post subject: Reply with quote
So you aren't looking to 'fix' PER, but just to 'tweak' it a bit?
Joe
PostPosted: Mon Nov 16, 2009 6:03 pm Post subject: Reply with quote
Until someone reinvents the way basic box score statistics are tracked, specifically to make the accuracy and complexity of defensive stats on the same level as offensive stats, it's impossible to 'fix' PER in the way you want it to be fixed. Giving equal weight to defensive stats that are far less accurate in explanatory power is not a solution that would 'fix' PER. Likewise, neither is distributing credit for team defense to individual players regardless of if they deserve it.
All I'm looking to do is modify PER to account for things that are trackable, in addition to the things it already tracks well. One idea is distinguishing between free throws that account for possessions and ones that don't, rather than using the 0.44 estimator, a percentage which varies from player to player (though not on the same level as the %Assist rate was varying). If you want me to modify PER to make it some kind of end-all, be-all stat that matches up to efficiency differential, it's not happening, because it's not possible to do it accurately with the data at public disposal.
Mike G
PostPosted: Mon Nov 16, 2009 6:18 pm Post subject: Reply with quote
One stat that is tracked is Opponent Pts/G (or Pts/Poss). Just as a player can have a Reb% that is scaled to Tm and Opp Reb (not necessarily in equal measure), we can scale a player's Scoring to Tm and/or Opp Scoring.
In other words, 20 points in a 115 PPG environment is not equal to 20 points in 90. This seems pretty obvious; just as getting 10 Reb out of 50 is not as good as 10 of 35.
You wouldn't have to scale a player's whole PER to the available points, just his points. There is a 'scoring' component, isn't there?
Crow
PostPosted: Mon Nov 16, 2009 6:30 pm Post subject: Reply with quote
If you don't think you can fix PER to weight defense equally and accurately then I'd vote for an adjusted, strictly offensive PER.
The slightly adjusted for assisted rate and charges is an improvement over the classic version but I'd probably be much more regular in using an offensive PER.
I'd mainly use defensive Adjusted +/- for the other half of the game.
Back to top
View user's profile Send private message
Joe
PostPosted: Mon Nov 16, 2009 7:00 pm Post subject: Reply with quote
Mike G wrote:
One stat that is tracked is Opponent Pts/G (or Pts/Poss). Just as a player can have a Reb% that is scaled to Tm and Opp Reb (not necessarily in equal measure), we can scale a player's Scoring to Tm and/or Opp Scoring.
In other words, 20 points in a 115 PPG environment is not equal to 20 points in 90. This seems pretty obvious; just as getting 10 Reb out of 50 is not as good as 10 of 35.
You wouldn't have to scale a player's whole PER to the available points, just his points. There is a 'scoring' component, isn't there?
Rebounds are opportunities from missed field goals (and free throws), not total rebounds grabbed in a game. The total rebounds is just used as a quick estimate because rebounds and missed field goals tend to correlate very highly. Likewise, points are opportunities from possessions used, not total points scored in a game. Using points as a percentage of total points scored makes no sense, because possessions used and points vary wildly, plus we already can accurately estimate possessions, so why use something that makes less sense instead? PER is already based highly in possessions, with one notable exception, which I (and probably Hollinger) doesn't think can change without skewing the balance of offensive to defensive weights, thus ruining its usefulness.
Crow, that's a good idea, and something I've looked into.
___________
deepak
PostPosted: Mon Nov 16, 2009 7:22 pm Post subject: Reply with quote
Any plans to factor in charges into your APER, since you are tracking them?
Back to top
View user's profile Send private message
Joe
PostPosted: Mon Nov 16, 2009 7:26 pm Post subject: Reply with quote
From first post in this thread:
Quote:
I also added charges into the formula with the same weight as steals, which makes sense given they have the same exact effect in regards to possessions, and there's no reason not to account for them since they are now recorded.
Very Happy
_
deepak
PostPosted: Mon Nov 16, 2009 9:27 pm Post subject: Reply with quote
Don't know how I missed that.
I wonder, though, if a steal is really the same value as a charge. It's true that they both result in a defensive stop, but steals also typically lead to transition opportunities (hence, better offensive possessions). On the other hand, one could argue that players who collect a lot of steals also tend to to gamble more, so maybe that cancels out the offensive benefits on average.
Should such things be taken into account in these linear weights ratings?
Joe
PostPosted: Mon Nov 16, 2009 9:38 pm Post subject: Reply with quote
One of our plans is to eventually make a stat that accounts for steals that lead to made baskets within 5 seconds or so, something we could possibly modify the APER formula with when we do it. Another thing is distinguishing from blocks recovered by offense and blocks recovered by defense, which could also be modified in the APER.
The problem with changing the value of steals based on what they lead to, however, is technically, if you're going to give a player more than the value of a possession for stealing a ball, you really should be subtracting that extra value from scorers on the team, because you're suggesting that the credit for the scores steals lead to belong in part to the thief. It doesn't make sense to distribute the credit twice, such as statistics like Win Score over-counting assists by not deducting them from scorers. Then again, PER doesn't follow a strict possession formula, so it is something that can be considered.
_________________
http://www.hoopdata.com
Last edited by Joe on Mon Nov 16, 2009 9:41 pm;
ecumenopolis0
PostPosted: Mon Nov 16, 2009 9:40 pm Post subject: Reply with quote
deepak_e wrote:
Don't know how I missed that.
I wonder, though, if a steal is really the same value as a charge. It's true that they both result in a defensive stop, but steals also typically lead to transition opportunities (hence, better offensive possessions). On the other hand, one could argue that players who collect a lot of steals also tend to to gamble more, so maybe that cancels out the offensive benefits on average.
Should such things be taken into account in these linear weights ratings?
And of course, obvious note, taking a charge also puts a foul on the opponent.
page 2 of 5
Author Message
Crow
PostPosted: Tue Nov 17, 2009 1:52 am Post subject: Reply with quote
"I looked into modifying the accounting of positive possessions used (changing how PER doesn't deduct for possessions used on makes after accounting for the added value in points), but the results were wildly out of whack with reality, and I now understand why Hollinger weighed the formula the way he did."
I assume this is the same thing as looking at the break even point for FG% efficiency? Correct me if I am wrong.
I'd be interested in a look at the results where the break-even point were shifted 1/3 to 1/2 way towards what Wins Produced does. Or at least some more discussion of it if you wanted to.
If you keep the defensive stats would you consider increasing the weight for defensive rebounds 1/3 to 1/2 way towards what Wins Produced does or towards what the regression you trust says, whatever that is? Defensive rebounding is an acknowledged imperfect hole-filler for defense but I wonder (without rerunning all the debate and data thru my mind) if the defensive rebounding weight in PER is on its own, apart from that rationale, too low.
Would you change the way fouls are treated to less negative?
Are you satisfied with the level of credit to players for assists?
ultimately do you want to stay within adjusting PER fairly narrowly or could you reach a point where you decided to try to make essentially a new metric, perhaps a compromise metric?
Back to top
View user's profile Send private message
Mike G
PostPosted: Tue Nov 17, 2009 9:41 am Post subject: Reply with quote
Joe wrote:
Rebounds are opportunities from missed field goals (and free throws), not total rebounds grabbed in a game. The total rebounds is just used as a quick estimate ...
Do you have this backwards? You can estimate rebounds from rebound-opportunities. Rebounds are actually counted.
Is it more informative to know that your team got 55% of all potential rebounds, than to know your team got 55% of all actual rebounds?
Quote:
.. Likewise, points are opportunities from possessions used, not total points scored in a game. Using points as a percentage of total points scored makes no sense, because possessions used and points vary wildly, plus we already can accurately estimate possessions,..
So if your team averages 20 points per position, what does the possession count mean? Isn't it more meaningful that the opponent averages more or less than 20 per position?
You might be getting outscored by 110-100, due to high pace and poor shooting, low pace and poor defense, or whatever. Either way, your average position is getting beat 22-20.
Pts/Poss * Poss/G = Pts/G
Scoring 100 is very good when you allow 90; it's very weak vs 110.
This makes sense to me.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Joe
PostPosted: Tue Nov 17, 2009 10:30 am Post subject: Reply with quote
Crow wrote:
"I looked into modifying the accounting of positive possessions used (changing how PER doesn't deduct for possessions used on makes after accounting for the added value in points), but the results were wildly out of whack with reality, and I now understand why Hollinger weighed the formula the way he did."
I assume this is the same thing as looking at the break even point for FG% efficiency? Correct me if I am wrong.
I'd be interested in a look at the results where the break-even point were shifted 1/3 to 1/2 way towards what Wins Produced does. Or at least some more discussion of it if you wanted to.
If you keep the defensive stats would you consider increasing the weight for defensive rebounds 1/3 to 1/2 way towards what Wins Produced does or towards what the regression you trust says, whatever that is? Defensive rebounding is an acknowledged imperfect hole-filler for defense but I wonder (without rerunning all the debate and data thru my mind) if the defensive rebounding weight in PER is on its own, apart from that rationale, too low.
Would you change the way fouls are treated to less negative?
Are you satisfied with the level of credit to players for assists?
ultimately do you want to stay within adjusting PER fairly narrowly or could you reach a point where you decided to try to make essentially a new metric, perhaps a compromise metric?
The weights Hollinger uses for about everything in his formula are sound and in line with logic, which likewise align with the weights Rosenbaum used (for the most part) in his alternate win score. The weights on rebounds are perfect, with offensive rebounds weighing the defensive rebound percentage and defensive rebounds weighing the offensive rebound percentage. Changing those arbitrarily to something else to make the numbers more like win score? I don't understand why one would want to do that.
I agree with the assist weights, and they are close to in line with the weights Oliver and Pelton use in their metrics.
Fouls are whatever. Offensive fouls are being double counted in the current equation, but changing the foul portion of the equation probably would have marginal value and I don't really agree or disagree strongly with how they are used.
Quote:
Do you have this backwards? You can estimate rebounds from rebound-opportunities. Rebounds are actually counted.
Is it more informative to know that your team got 55% of all potential rebounds, than to know your team got 55% of all actual rebounds?
You're agreeing with me.
Quote:
So if your team averages 20 points per position, what does the possession count mean? Isn't it more meaningful that the opponent averages more or less than 20 per position?
No. It matters how many points they average per possession, as possessions are the opportunities to score points. If a shooting guard scores 15 points on 5 possessions, he is providing incredible efficiency, but by your idea, he'd be below average, which is ludicrous.
Quote:
You might be getting outscored by 110-100, due to high pace and poor shooting, low pace and poor defense, or whatever. Either way, your average position is getting beat 22-20.
Pts/Poss * Poss/G = Pts/G
Scoring 100 is very good when you allow 90; it's very weak vs 110.
This makes sense to me.
This is the logic for using possession-based costs in metrics, and it makes sense. Measuring points as a percentage of total points doesn't make sense for the reason outlined above.
_________________
http://www.hoopdata.com
Back to top
View user's profile Send private message Visit poster's website
Mike G
PostPosted: Tue Nov 17, 2009 10:44 am Post subject: Reply with quote
Joe,
Perhaps you are confusing 'position' and 'possession'.
I apologize for introducing this possibility. You might re-read with this clarification.
A 'position' is 1/5 of a team's performance; generally 48 player-minutes.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Joe
PostPosted: Tue Nov 17, 2009 10:48 am Post subject: Reply with quote
Mike G wrote:
Joe,
Perhaps you are confusing 'position' and 'possession'.
I apologize for introducing this possibility. You might re-read with this clarification.
A 'position' is 1/5 of a team's performance; generally 48 player-minutes.
You said 'Poss' - that is the abbreviation for possession. Position doesn't have two consecutive S's at any point in it, so there's no way you could've been referring to that.
Regardless, virtually every noteworthy composite metric out there uses possession-based costs for scoring opportunities as opposed to measuring as a function of 1/5 points scored, and I've explained the reasoning for it. I'm not the one who is confused here.
_________________
Mike G
C
PostPosted: Tue Nov 17, 2009 11:19 am Post subject: Reply with quote
I see I wrote "position" 3 times, "possession" once, and it's abbreviation, "Poss", twice.
Again apologize if this was distracting.
The idea was to suggest that players sharing a position combine for 1/5 of a team's performance, and that this performance (whether or not you call it 'efficiency') really is relative to the opposition.
A team may score 160 points in 100 possessions, and allow 165. This is not as effective as scoring 80 while allowing 75. In any number of possessions.
Since # of possessions is generally assumed to be equal for both a team and its opponent, it seems reasonable that this quantity cancels in an 'effectiveness' formula. To win, you still have to score >50% of the points in a game, which may consist of 80 or 100 or 120 possessions.
If you've been outscored by 100-90, then whatever efficiency you feel you've achieved (on a per-possession basis) could be said to be only 90/100 (or .90) as effective as it would be if the score were even. This is equally true as a per-game average or a 3rd-quarter recap.
If you scale the scoring portion of your metric to (TmPts/OpPts), then you primarily affect the 'scoring specialist' types, while basically leaving the 'defensive specialist' alone.
_________________
Joe
PostPosted: Tue Nov 17, 2009 11:39 am Post subject: Reply with quote
Mike G wrote:
I see I wrote "position" 3 times, "possession" once, and it's abbreviation, "Poss", twice.
Again apologize if this was distracting.
The idea was to suggest that players sharing a position combine for 1/5 of a team's performance, and that this performance (whether or not you call it 'efficiency') really is relative to the opposition.
A team may score 160 points in 100 possessions, and allow 165. This is not as effective as scoring 80 while allowing 75. In any number of possessions.
Since # of possessions is generally assumed to be equal for both a team and its opponent, it seems reasonable that this quantity cancels in an 'effectiveness' formula. To win, you still have to score >50% of the points in a game, which may consist of 80 or 100 or 120 possessions.
If you've been outscored by 100-90, then whatever efficiency you feel you've achieved (on a per-possession basis) could be said to be only 90/100 (or .90) as effective as it would be if the score were even. This is equally true as a per-game average or a 3rd-quarter recap.
If you scale the scoring portion of your metric to (TmPts/OpPts), then you primarily affect the 'scoring specialist' types, while basically leaving the 'defensive specialist' alone.
Here's the problem with this line of thinking (and I've yet to see you illustrate the problem with the alternative, which is currently widely used):
A shooting guard and a small forward each play 48 minutes in an imaginary game where their team scores 100 points. The shooting guard scores 20 and the small forward scores 20, however the shooting guard gets his scoring by hitting 10-for-10 FGA and the small forward gets his scoring by hitting 10-of-30 FGA. One used three times as many possessions as the other, yet they still scored the same points. By your scaling, they'd be considered equal contributors, when in reality, one basically contributed 20 points + 10 free possessions (if you assume the EV of a possession to be 1), and the other contributed 20 points - 10 extra possessions used over the EV. Their scoring numbers are equal, but in terms of roundabout impact to scoring, assuming possessions have an EV of 1, they actually have a difference of about 20 expected points, making them basically the difference between an All-NBA first teamer and a player who doesn't deserve to be in the league. Yet your scaling would consider them equal.
This is why possession-based costs is the norm, and there is no disadvantage to it. I don't see what kind of value using your suggested scale would add, though I do see a ton of problems with it.
_________________
http://www.hoopdata.com
Back to top
View user's profile Send private message Visit poster's website
Mike G
PostPosted: Tue Nov 17, 2009 12:06 pm Post subject: Reply with quote
Joe wrote:
... By your scaling, they'd be considered equal contributors....
Not in any way, shape, or form.
Here, you've pulled the switcheroo between 'points scored' and "the scoring portion of your metric ".
You've also not considered what the opposition has done while these guys are shooting 10-10 and 10-30.
If the opposition shot 25% for the game, then 10-30 is pretty good. If they shot 40% or better, it's bad.
A couple of years ago, someone 'demonstrated' that Michael Jordan hadn't (contrary to popular belief) dominated the Finals in '97 and '98; and in fact he'd shot badly. Not badly enough to lose the games, but badly relative to league average, Jordan average, etc.
In fact, his opponents (the Jazz) shot much worse; and Jordan's teammates shot worse than he. So he did, in fact (and concurring with popular belief) dominate those Finals.
In another playoff series, hitting 10-20 shots is not nearly good enough to keep up with the opposition. Ignore this at your peril.
Joe
PostPosted: Tue Nov 17, 2009 12:34 pm Post subject: Reply with quote
Mike G wrote:
Joe wrote:
... By your scaling, they'd be considered equal contributors....
Not in any way, shape, or form.
Here, you've pulled the switcheroo between 'points scored' and "the scoring portion of your metric ".
You've also not considered what the opposition has done while these guys are shooting 10-10 and 10-30.
If the opposition shot 25% for the game, then 10-30 is pretty good. If they shot 40% or better, it's bad.
A couple of years ago, someone 'demonstrated' that Michael Jordan hadn't (contrary to popular belief) dominated the Finals in '97 and '98; and in fact he'd shot badly. Not badly enough to lose the games, but badly relative to league average, Jordan average, etc.
In fact, his opponents (the Jazz) shot much worse; and Jordan's teammates shot worse than he. So he did, in fact (and concurring with popular belief) dominate those Finals.
In another playoff series, hitting 10-20 shots is not nearly good enough to keep up with the opposition. Ignore this at your peril.
I'm not ignoring anything. I already addressed earlier in this thread that distributing team defense to individual players is something that Hollinger himself has stated on this forum to be something he doesn't want to do, as it is not necessarily accurate. You're giving credit where it isn't necessarily deserved.
_________________
http://www.hoopdata.com
Back to top
View user's profile Send private message Visit poster's website
DLew
PostPosted: Tue Nov 17, 2009 1:33 pm Post subject: Reply with quote
I think that the adjustments made to PER are a good use of the new data, but that going much further (i.e. crediting the offense created by steals to the person who got the steal, etc) probably runs the risk of creating a new metric that shouldn't really be called adjusted PER.
Back to top
View user's profile Send private message
Joe
PostPosted: Tue Nov 17, 2009 1:39 pm Post subject: Reply with quote
DLew wrote:
I think that the adjustments made to PER are a good use of the new data, but that going much further (i.e. crediting the offense created by steals to the person who got the steal, etc) probably runs the risk of creating a new metric that shouldn't really be called adjusted PER.
I agree.
_________________
http://www.hoopdata.com
Back to top
View user's profile Send private message Visit poster's website
Mike G
Joined: 14 Jan 2005
Posts: 3564
Location: Hendersonville, NC
PostPosted: Tue Nov 17, 2009 1:41 pm Post subject: Reply with quote
Giving team-average defensive credit to everyone on a team (as kjb described) wouldn't change players' ranking on a team, relative to one another; it would only change their rankings relative to those on other teams.
Because it would also adjust the team relative to their actual strength, that would be an improvement, which also appears with the individuals' ratings. Some accuracy is gained, and nothing is lost, in this blanket adjustment.
Trying for more, one could suggest docking scoring specialists more heavily on poor defensive teams, relative to low scorers on good defensive teams. Again, all gain and no loss, in ranking player effectiveness.
None of this affects a player's per-possession effectiveness. The average possession itself has a value adjustment. Specifically, it's relative to TmPts/OppPts.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Crow
Joined: 20 Jan 2009
Posts: 806
PostPosted: Tue Nov 17, 2009 2:21 pm Post subject: Reply with quote
I don't think the FG% breakpoint is quite 'best" or the weight on defensive rebounds is "perfect" from all angles and I also disagree with the large size of the assist weight and fouls remain an area that could possibly be improved or dropped given the difficulty but I am not going to argue them now. I was asking you to clarify if you were sure what you were going to do and you did.
.
Last edited by Crow on Tue Nov 17, 2009 3:02 pm; edited 2 times in total
Back to top
View user's profile Send private message
kjb
Joined: 03 Jan 2005
Posts: 864
Location: Washington, DC
PostPosted: Tue Nov 17, 2009 2:29 pm Post subject: Reply with quote
There is an argument in favor of simply using team defensive rating as a blanket adjustment -- namely that defense is primarily about teamwork.
I can see the other side of the argument too -- that some players inevitably end up with too much or too little credit for the team's defensive performance. Brendan Haywood, who has been a good defender on a horrible defensive team, immediately springs to mind as someone who would be "unfairly" penalized because of his teammates' poor performance. I'm pretty ambivalent on the issue. The simple adjustment I used sufficed for my purposes, even recognizing all of its limitations.
Another option might be to use the on/off defensive data at 82games or one of the other pbp sites. Still not perfect, but perhaps an advancement over the "simple" technique I described earlier. I don't have the programming skills to rip that data from those sites; nor do I have time to do all the data entry by hand.
Adding in the defensive adjustment I described boosted correlation with efficiency differential from .85 to .95ish (I don't have that spreadsheet with me today to check the exact correlation).
Back to top
View user's profile Send private message AIM Address Yahoo Messenger
Joe
Joined: 27 Sep 2009
Posts: 94
Location: Long Island, NY
PostPosted: Tue Nov 17, 2009 3:14 pm Post subject: Reply with quote
Crow wrote:
I don't think the FG% breakpoint is quite 'best" or the weight on defensive rebounds is "perfect" from all angles and I also disagree with the large size of the assist weight and fouls remain an area that could possibly be improved or dropped given the difficulty but I am not going to argue them now. I was asking you to clarify if you were sure what you were going to do and you did.
.
Assists are subjective, always have been, always will be. Hollinger, Oliver, and Pelton all are between 2/3 to 3/4 a point for each field goal, however, and I have nothing substantive to argue against that.
The FG% breakpoint (and all scoring efficiency breakpoints) in PER are not ideal, because like I said, there is no deduction for possessions used on positive possessions. A two point field goal yields two points, while it should yield two points minus the VOP. The reason for this is the accuracy and complexity of defensive statistics compared to offensive statistics. By not deducting for possessions on made field goal attempts, it increases the relative weight of offense (namely scoring) in the equation, which is necessary because the defensive statistics shouldn't weigh so much given their poor explanatory power. This is why your suggestion of making an offense-only PER is a good idea, because this issue would no longer exist, and then you could just measure offense, which could be done with 99%+ accuracy.
Fouls are a tricky subject. Don't really know the best way to approach those.
kjb, it's not a bad idea, and I get its merits, but if I want to keep calling it APER, I don't think doing something that Hollinger has clearly stated on these forums that he wouldn't do is a good idea. If ESPN tracked assisted field goals and charges, and every other website using PER also did, I'm fairly certain Hollinger would've made these adjustments I did a long time ago, but it's kind of hard to change a statistic that is so widely used. I don't think anything I did changed the spirit of PER, but your adjustment would. Not that it's a bad idea, just not what PER is.
_________________
http://www.hoopdata.com
page 3
Author Message
Crow
Joined: 20 Jan 2009
Posts: 775
PostPosted: Tue Nov 17, 2009 4:15 pm Post subject: Reply with quote
Thanks for your attention to making replies to comments and questions generally. A good policy.
I'd support some form of defensive adjustment to PER or just offensive PER. Slight modified PER mostly the way it has been is ok too but I'd prefer one of the former choices over it.
On assists I 'll add that I did see Neil Paine's regression results where an A per 40 minutes was worth about 1/10th value of a turnover/40 (with the opposite sign of course)
http://www.basketball-reference.com/blog/?p=2191
Fouls are indicated as a moderate positive.
But his results even further diminished the value of defensive rebounds so I don't know if I can reasonably just take the parts I agree with.
Back to top
View user's profile Send private message
DLew
Joined: 13 Nov 2006
Posts: 222
PostPosted: Tue Nov 17, 2009 5:57 pm Post subject: Reply with quote
You really can't interpret those results that way because of the interaction term in that regression (pts*reb*ast). Some of the value of points, rebounds, and assists is being represented in that coefficient.
Back to top
View user's profile Send private message
Crow
Joined: 20 Jan 2009
Posts: 775
PostPosted: Tue Nov 17, 2009 6:47 pm Post subject: Reply with quote
True.
And that term can capture some of the value of defensive rebounds too.
I shouldn't have claimed simple support from this reference.
I was previously aware of the impact of the versatility index and minutes too.
But if your re-ran the regression without the versatility index or minutes would the assist weight be as high relatively as in PER? I'd like to know. I'd guess a decent to good chance it wouldn't be. Assists is only about 15% of the input to the versatility index on average.
Back to top
View user's profile Send private message
jkubatko
Joined: 05 Jan 2005
Posts: 702
Location: Columbus, OH
PostPosted: Wed Nov 18, 2009 8:08 am Post subject: Reply with quote
Joe wrote:
PER already correlates at about an 85% level with wins, higher than AWS, WS, and EFF (based on some quick excel work I did on the 2009 season).
That's not true. Yes, the correlation between team wins and team PER in 2008-09 was 0.855, but the correlation between team wins and team Win Shares was 0.988.
_________________
Regards,
Justin Kubatko
Basketball-Reference.com
Back to top
View user's profile Send private message Send e-mail Visit poster's website
DLew
Joined: 13 Nov 2006
Posts: 222
PostPosted: Wed Nov 18, 2009 11:12 am Post subject: Reply with quote
Justin,
I'm pretty sure that when he wrote "WS" he was referring to Win Score.
Back to top
View user's profile Send private message
Joe
Joined: 27 Sep 2009
Posts: 94
Location: Long Island, NY
PostPosted: Wed Nov 18, 2009 11:43 am Post subject: Reply with quote
DLew wrote:
Justin,
I'm pretty sure that when he wrote "WS" he was referring to Win Score.
Correct.
_________________
http://www.hoopdata.com
Back to top
View user's profile Send private message Visit poster's website
Mike G
Joined: 14 Jan 2005
Posts: 3504
Location: Hendersonville, NC
PostPosted: Wed Nov 18, 2009 4:03 pm Post subject: Reply with quote
Joe wrote:
Quote:
So if your team averages 20 points per position, what does the possession count mean? Isn't it more meaningful that the opponent averages more or less than 20 per position?
No. It matters how many points they average per possession....
Last year, the Mavs met the Spurs in the opening playoff round. Dal averaged 1.11 points per possession, and they won the series 4-1. Avg pt-diff +7.0 .
In round 2, they met the Nuggets and again averaged 1.11 Pts/Poss. This series they lost 4-1, by -7.4 PPG.
It doesn't matter what the league average Pts/Poss was in 2008-09, over 1230 NBA games. They won or lost due to their Pts/Poss being significantly more/less than that of their playoff opponents', for 5 games each.
In the Dal-SA series, an average possession netted 1.075 points; in the Dal-Den series, it was 1.155 . These were the marks they needed to beat, to likely win these series.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Joe
Joined: 27 Sep 2009
Posts: 94
Location: Long Island, NY
PostPosted: Wed Nov 18, 2009 5:43 pm Post subject: Reply with quote
Mike G wrote:
Joe wrote:
Quote:
So if your team averages 20 points per position, what does the possession count mean? Isn't it more meaningful that the opponent averages more or less than 20 per position?
No. It matters how many points they average per possession....
Last year, the Mavs met the Spurs in the opening playoff round. Dal averaged 1.11 points per possession, and they won the series 4-1. Avg pt-diff +7.0 .
In round 2, they met the Nuggets and again averaged 1.11 Pts/Poss. This series they lost 4-1, by -7.4 PPG.
It doesn't matter what the league average Pts/Poss was in 2008-09, over 1230 NBA games. They won or lost due to their Pts/Poss being significantly more/less than that of their playoff opponents', for 5 games each.
In the Dal-SA series, an average possession netted 1.075 points; in the Dal-Den series, it was 1.155 . These were the marks they needed to beat, to likely win these series.
So let's say Chris Paul is on an awful team that allows 1.1 points per possession on defense, which would be good for worst in the league most seasons. You're telling me that just because Chris Paul's teammates are awful on defense, he should be expected to do more per possession than Lebron James on a defense that is allowing 1.0 points per possession, even if Paul and James are about equal in defensive effectiveness, something that can't be measured objectively through statistics with accuracy, or at least not through box score statistics. You really think this is a good idea and would improve PER? Holding people to different standards of expected value based on things that are completely out of their control?
If you do think that is a good idea, I think we'll just have to agree to disagree there.
_________________
http://www.hoopdata.com
Back to top
View user's profile Send private message Visit poster's website
Mike G
Joined: 14 Jan 2005
Posts: 3504
Location: Hendersonville, NC
PostPosted: Thu Nov 19, 2009 10:46 am Post subject: Reply with quote
You've adjusted Chris Paul's PER from a league-leading and unprecedented 33.8 to an astronomical 37.6 . Wilt, LeBron, and Jordan topped out around 31.7 .
If you were to factor his team's DRtg to the league average, you might multiply by 1055/1129 (.934), and then his PER would be 31.6; his APER converts to merely 35.1 .
You tell us: Has Paul been merely as good as anyone who's ever played? -- or 11% better, or 19% better?
I don't 'expect' Paul, or Jarret Jack, or anyone else to do more than LeBron. If I ranked them higher than LeBron, then I would.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Joe
Joined: 27 Sep 2009
Posts: 94
Location: Long Island, NY
PostPosted: Thu Nov 19, 2009 11:05 am Post subject: Reply with quote
Mike G wrote:
You've adjusted Chris Paul's PER from a league-leading and unprecedented 33.8 to an astronomical 37.6 . Wilt, LeBron, and Jordan topped out around 31.7 .
If you were to factor his team's DRtg to the league average, you might multiply by 1055/1129 (.934), and then his PER would be 31.6; his APER converts to merely 35.1 .
You tell us: Has Paul been merely as good as anyone who's ever played? -- or 11% better, or 19% better?
I don't 'expect' Paul, or Jarret Jack, or anyone else to do more than LeBron. If I ranked them higher than LeBron, then I would.
Wow, you really like using misleading statistics, don't you? Lebron had a higher APER than Paul last season. You think this season's 10 game sample size is a more reliable indicator than that? Variation of course could not be a factor in a 10 game sample size, could it?
We also don't have access to Jordan's %Assisted number, so there's no telling how high he would've been, but his number was probably in line with Kobe's most years, which would likely bump his 31.7 to 33+. You think comparing a 10 game sample size of Paul, using a different formula, to Jordan's old PER number, is a fair comparison? Really?
Once again, I'm simply modifying the formula in ways that likely would've been done a long time ago, were the statistics available. The whole point of the part of the formula I modified was to estimate how many shots a player created on his own. No longer does it need to be estimated. I'm sorry the numbers don't jive with your expectations or subjective analysis of Paul's game, but quite frankly I don't care about either of those things. And once again, Hollinger has stated on these very forums he has never intended to apply team defensive numbers to individual defensive performance in PER, due to accuracy concerns. Why do you keep bringing this up?
And for the record, Paul was assisted on only 14.3% of his shots last season. Only two other players in the league were below 21.5% (Nash and Roko Ukic, though the latter's scoring efficiency suggests it probably wasn't a good idea for him -- and neither were below 17%). For a player to use as high a usage as Paul did, score with a monstrous 60% TS%, and only be assisted on 14.3% of his shots could very well be one of the greatest statistical accomplishments in NBA history. We quite frankly have no access to data going back far enough to say for sure, but the three seasons of data we do have access to suggest that his level of shot creation in combination with efficiency puts him in a very, very special class of player.
_________________
http://www.hoopdata.com
Back to top
View user's profile Send private message Visit poster's website
Mike G
Joined: 14 Jan 2005
Posts: 3504
Location: Hendersonville, NC
PostPosted: Thu Nov 19, 2009 11:40 am Post subject: Reply with quote
Joe wrote:
... you really like using misleading statistics, don't you? Lebron had a higher APER than Paul last season. You think this season's 10 game sample size is a more reliable indicator than that? ..
No, and no. I'm only trying to help.
You wrote: "... let's say Chris Paul is on an awful team that allows 1.1 points per possession", and that looks a lot like the Hornets this season. Last year, they were an above-average team defensively.
Quote:
.. to estimate how many shots a player created on his own. No longer does it need to be estimated. I'm sorry the numbers don't jive with your expectations or subjective analysis of Paul's game
No need to be sorry. I've been using %Unassisted as a factor for several years now in my ratings. Yet it's a small adjustment, even in Paul's case, compared to team per-possession differences.
Without adjusting for variations in possession value, the worst case is, as you try to describe, that a good defender among some bad ones is unfairly (dis-) credited. But the counter-argument is that without an adjustment, virtually everyone on a team is unfairly credited.
Almost every player in the league is not playing in a league-average possession value. Why credit every player's average possession relative to something so arbitrary? It changes year to year, and this becomes the standard applied to players; so why not team by team?
In 1991, the Nuggets gave up 1.15 Pts/Poss, while the Pistons gave up 1.046. Michael Adams' PER was 22.3, while Joe Dumars had a career peak 18. Would Adams be expected to substantially outplay Dumars? Would a team correction basically adjust their relative PER's to their relative effectiveness?
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Joe
Joined: 27 Sep 2009
Posts: 94
Location: Long Island, NY
PostPosted: Thu Nov 19, 2009 11:52 am Post subject: Reply with quote
Quote:
You wrote: "... let's say Chris Paul is on an awful team that allows 1.1 points per possession", and that looks a lot like the Hornets this season. Last year, they were an above-average team defensively.
I was using a hypothetical comparing two of the best players in the league. I don't know the Hornets' defensive efficiency numbers for the past two seasons off hand.
Quote:
Without adjusting for variations in possession value, the worst case is, as you try to describe, that a good defender among some bad ones is unfairly (dis-) credited. But the counter-argument is that without an adjustment, virtually everyone on a team is unfairly credited.
Actually, they're not credited at all. PER is an offensively biased statistic, something everyone knows. Anyone looking towards PER to explain team defense is looking at the wrong statistic. That is not its purpose. It is outside its scope.
Quote:
Almost every player in the league is not playing in a league-average possession value. Why credit every player's average possession relative to something so arbitrary? It changes year to year, so why not team to team?
Because the point of PER is to compare players across the league and put them on equal footing. If players' VOP change by up to 10% across the league, it ruins PER as a comparison tool for individual players. It may improve the explanatory power of team performance when adding up the sum of the players, but that's not what PER is meant for. There are other statistics made to do that. Further, like I said earlier, it's not fair to hold players to different standards based on things that are out of their control. It's not an arbitrary number either. For comparing players in the league as a whole, using the league average is far from arbitrary.
Quote:
In 1991, the Nuggets gave up 1.15 Pts/Poss, while the Pistons gave up 1.046. Michael Adams' PER was 22.3, while Joe Dumars' was 18. Would Adams be expected to outplay Dumars? Would a team adjustment basically correct their relative PER's?
Without getting into PER's predictive power for future performance, something I'm not entirely familiar with, I'd say that yes, PER would predict Adams to outplay Dumars, but only in the things that PER actually measures, which don't include team defense. I don't know if a team adjustment would "correct" their relative PER's, because I don't know if each contributed exactly 20% to their team defensive performance when on the floor. There's also a good chance this comparison is hurt by the problems with positive possession usage in PER, something that doesn't properly dock players for poor efficiency, though I don't know Adams' numbers off hand, nor am I familiar with him as a player, as I was five years old at the time of this season.
All of this discussion is lending more credence to Crow's suggestion of creating an offensive-only PER. No matter what method you apply to defensive adjustments in PER, be it using box score statistics with little explanatory power or applying a blanket adjustment assuming everyone is responsible for 20% of team defense, accuracy suffers, and it diminishes the explanatory power and accuracy of the offensive side of the equation, which actually is quite valuable when isolated and adjusted for proper possession usage.
_________________
http://www.hoopdata.com
Back to top
View user's profile Send private message Visit poster's website
DLew
Joined: 13 Nov 2006
Posts: 222
PostPosted: Thu Nov 19, 2009 11:59 am Post subject: Reply with quote
MikeG,
I think you are being a bit unfair here. Your criticisms of PER are on target, but Joe did not invent PER nor is he advocating that it be used as a one number player evaluation. I know when you do your EWins that you want those numbers to completely and totally represent player value, and that's admirable, but PER is not trying to do that, and more importantly Joe is not PER's keeper.
Back to top
View user's profile Send private message
Mike G
Joined: 14 Jan 2005
Posts: 3504
Location: Hendersonville, NC
PostPosted: Thu Nov 19, 2009 12:02 pm Post subject: Reply with quote
Joe wrote:
... If players' VOP change by up to 10% across the league, it ruins PER as a comparison tool for individual players. .
Alright then. We know that VOP varies by this much. Ignoring this doesn't make it go away. Factoring it in fixes it.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Joe
Joined: 27 Sep 2009
Posts: 94
Location: Long Island, NY
PostPosted: Thu Nov 19, 2009 12:03 pm Post subject: Reply with quote
Mike G wrote:
Joe wrote:
... If players' VOP change by up to 10% across the league, it ruins PER as a comparison tool for individual players. .
Alright then. We know that VOP varies by this much. Ignoring this doesn't make it go away. Factoring it in fixes it.
It "fixes" explanatory power at a team level. It "breaks" comparative value across teams.
_________________
http://www.hoopdata.com