APBRmetrics

Hey guys, I know most of you despise WP with all your hearts, but I still think you might find my latest post interesting.

http://jbrocato.wordpress.com/2012/08/0 ... -produced/


James

I've got some recipes that make dog doo less terrible.

Interesting stuff, James.

Thanks, Daniel.

Mike, there's really no need for snide remarks. I understand your position, but you don't have to be condescending about it.

James

Thanks for the new data using new assumptions. I like to have new angles on performance to consider alongside what is already out there.

I generally agree with your approach I think (may need to think more and more carefully about the efficiency break point) but maybe the offensive rebounding split could be even more nuanced.

With defensive rebounds, you reduced the credit to the defensive rebounder to give credit in the end back to the shot defenders. (In WP it is at team level, equally. In play by play metrics some can go to the individual defender.) What do you think about revising the split so 50% of the credit goes to the offensive rebounder, 20% to the shooter (who might, if they are wise, be looking at whether offensive rebounders are ready or are expected to be able to battle) and either 10% to each of the other 3 players for simplicity for team challenge on the offensive glass creating the possibility that someone on the offense gets the rebound or divide the remaining credit to the remaining 3 players based on average level of position involvement in offensive rebounding (in general or for the specific case of when the shot came from a certain position). Your method seems to me to give too much credit to the shooter.


"Unfortunately, I’m not at liberty to reveal the historical Wins Produced results, but I can compare the numbers of Wins Produced since 2000 to my adjusted productivity numbers."

I am not sure I am interpreting this passage correctly. Did you get denied use of some original formula WP numbers? Did you get denied use of the wages of wins website for this article?

James, have you read my ezPM articles? You appear to have arrived independently on some similar logic. Here are the references, if you're interested (read in this order):

http://d3coder.com/thecity/ezpm-yet-ano ... valuation/

http://d3coder.com/thecity/ezpm-1-0-now ... play-data/

http://d3coder.com/thecity/ezpm-v-2-0-i ... he-biggie/

Also, I would point you to ElGee's EV (Expected Value) model:

http://www.backpicks.com/expected-value/

Crow,

Thanks for the feedback. Interesting take on allocating some of the value to the other offensive players. I'll take a look at what might happen if we allocate the numbers the way you laid out. What about the other 20% though? 50+20+10 = 80. Either way, it's pretty arbitrary, but I'm still interested to see what the results look like. As for the passage, Dave is working on another book and has kindly asked us to not publicly share the WP results pre-2000, as he hopes to highlight them heavily in his book. And no, I wasn't denied any access or anything like that, I just felt this would be more appropriate on my blog since I did it completely independently.

Evan,

I'm vaguely familiar with most of the logic behind ezPM, though admittedly I haven't read all the details behind the logic and calculation. I'll check it out though, thanks.


James

James, how about changing the approach. Berri prompted out an hypothesis (even though he has never shown that this is true) that all positions would be equal. Now, you can use that hypothesis to find appropiate weights for the boxscore entries to fit the hypothesis.
The other thing you can do is: Use the appropiate equations for player possessions on offense and defense. Dean Oliver's book would be the starting point. For sure, that makes it more complicated than the current used equations, but it should rather solve the issues with WP48.

Did you do a retrodiction test on your "new" WP48? How is the explanatory factor?

Alright James, thanks for the replies and the interest in my suggestion.

To clarify, I meant 50+20+10 +10 +10 = 100 (or divide the 30% the other way I mentioned).

Player "X" is a C that only takes shots at the rim because he lacks the skill to shoot mid range and further jumpers. Player X hits 65% of his 7 shots at the rim.

Player "Y" is a SF that shoots from all over the court (inside, outside, 3 pointers etc..). Player "Y" hits 65% of hits 7 shot at the rim, but 43% of his 7 mid range and other long 2s, and 35% of his occasional 3s.

A 50% break even point or anything even close will suggest that "X" is contributing more to the offense because his efficiency is so high relative to "Y". That's preposterous when "Y" is actually an above average player from outside too and "X" has no skill from out there at all.

The bottom line is that "someone" has to take some of the jumpers and other below 50% shots that inevitably come up because the defense has done a good job. You can't punish the GREATER SKILLED players for taking them. The break even point should probably be around 40% because that's about average from mid range. Then those jumpers would be of little to no value to the player but also not punish them because 40% is about average.

A separate matter that is critical next step is finding as way to measure which players are taking those shots before they are needed (too early in the shot clock or when there were clearly better options). Those are the lower efficiency bad shot selection players that are actually hurting their teams with the extra shots. As far as I know, there is no stat like that. You actually have to watch the game and say "That was a horrible shot that Monta just took" etc.. "That was a tough shot by Westbrook, but Durant couldn't get around the pick and he had to take it" etc...

Italian Stallion wrote:Player "X" is a C that only takes shots at the rim because he lacks the skill to shoot mid range and further jumpers. Player X hits 65% of his 7 shots at the rim.

Player "Y" is a SF that shoots from all over the court (inside, outside, 3 pointers etc..). Player "Y" hits 65% of hits 7 shot at the rim, but 43% of his 7 mid range and other long 2s, and 35% of his occasional 3s.

A 50% break even point or anything even close will suggest that "X" is contributing more to the offense because his efficiency is so high relative to "Y". That's preposterous when "Y" is actually an above average player from outside too and "X" has no skill from out there at all.

The bottom line is that "someone" has to take some of the jumpers and other below 50% shots that inevitably come up because the defense has done a good job. You can't punish the GREATER SKILLED players for taking them.

Once you account for WP's position adjustment, players X and Y will be equally rated, assuming those numbers are representative/average for centers and small forwards. Even then it should be remembered that the rating is supposed to measure value brought to a basketball team, not overall quality as a basketball player. Small forwards as a class may be better at basketball than centers, but the assumption is that you need to have players that can fill both roles. People can also disagree with that, but that's how the metric works.

xkonk wrote:

Italian Stallion wrote:Player "X" is a C that only takes shots at the rim because he lacks the skill to shoot mid range and further jumpers. Player X hits 65% of his 7 shots at the rim.

Player "Y" is a SF that shoots from all over the court (inside, outside, 3 pointers etc..). Player "Y" hits 65% of hits 7 shot at the rim, but 43% of his 7 mid range and other long 2s, and 35% of his occasional 3s.

A 50% break even point or anything even close will suggest that "X" is contributing more to the offense because his efficiency is so high relative to "Y". That's preposterous when "Y" is actually an above average player from outside too and "X" has no skill from out there at all.

The bottom line is that "someone" has to take some of the jumpers and other below 50% shots that inevitably come up because the defense has done a good job. You can't punish the GREATER SKILLED players for taking them.

Once you account for WP's position adjustment, players X and Y will be equally rated, assuming those numbers are representative/average for centers and small forwards. Even then it should be remembered that the rating is supposed to measure value brought to a basketball team, not overall quality as a basketball player. Small forwards as a class may be better at basketball than centers, but the assumption is that you need to have players that can fill both roles. People can also disagree with that, but that's how the metric works.

But there are centers who are not like the player X above but who are more like player Y; centers whose role it is to draw the defensive center away from the basket and spread out the offense (and defense) in general. The Sonics made it to the Finals in 1996 with Sam Perkins bombing in 3-pointers. Bill Laimbeer and Wes Unseld were capable enough down low but did their best contributions as high post centers (and thus letting Dantley, Elvin Hayes, et al get the low post, close-in shots).

The average center may indeed resemble player X, but it appears that the centers who contributed by having outside games will tend to be underestimated by this technique.

If Perkins, Laimbeer, Unseld, and others like them were tabbed to shoot from further away while others got more of the close shots, perhaps these 'others' (like Dantley and so on) were better at the close shots? In which case it seems somewhat fair to 'underestimate' those centers for not being as proficient in that area.

xkonk wrote:If Perkins, Laimbeer, Unseld, and others like them were tabbed to shoot from further away while others got more of the close shots, perhaps these 'others' (like Dantley and so on) were better at the close shots? In which case it seems somewhat fair to 'underestimate' those centers for not being as proficient in that area.

Why should that be the case? If that would be reasonable, Dantley and Co. should also get rated down for their lack of interior defense, because if they were more proficient at that, the need for Perkins, Laimbeer, etc. wouldn't have been there. Or maybe, that is all complete nonsense and the overall goal of the game is to win a 5on5 competition, in which 5 players with different skills are trying to score more than the other 5 players on the court? And at that, it is completely irrelevant who is providing what and there are no special needs in terms of accumulating boxscore numbers from certain positions? Oh well ...

Btw, Dantley is a great example for a player who has a non-fitting skillset, which will lead to his numbers looking incredible efficient while the overall team result suffers. Dantley was a ball-stopper, used up a lot of time during a possession and only took a shot when it was an easy shot. Unfortunately, that resulted into him passing the ball up late in the possession and forcing his teammates to take bad shots. Obviously, that is not seen in the boxscore stats.

It seems to be a prevalent opinion that NBA coaches know what they're doing (although that may or may not be true). So if the coaches asked Dantley to shoot at the rim instead of Laimbeer, we would assume that Dantley is better at it.

It wouldn't make sense to mark Dantley down for lack of interior defense unless he played without a center and was thus asked to guard the paint. Even then he would have to do so poorly and cause the team to suffer, since WP treats defense as more of a team activity. But in the eyes of WP, not shooting as well as an average center is bad for centers. So if Laimbeer shoots from all over the court and lowers his efficiency, that would be a knock on him. If you think that being a proficient shooter from any location has value, then obviously no boxscore measure is going to make you happy because none that I know of make any distinctions between 2 point shots.

Speaking of Dantley, every boxscore measure I can find including Daniel's ASPM agrees that he was far above average as a player. Seems like such a consensus should mean something.