View previous topic :: View next topic |
Author |
Message |
back2newbelf
Joined: 21 Jun 2005 Posts: 276
|
Posted: Thu Feb 03, 2011 5:11 pm Post subject: Adjusted rebounding |
|
|
Using the same technique as in viewtopic.php?t=2669 I computed adjusted rebounding for the last 1 1/2 years.
Home teams grab more available offensive rebounds (30.4% vs 29.7%).
Lambdas were 1500 for offensive rebounding and 3000 for defensive rebounding.
Offensive and Defensive numbers are per 100 "rebound opportunities"
If someone wants the raw data in bbv matchup format I can upload the file.
http://www.stats-for-the-nba.appspot.com/rebounding _________________ http://stats-for-the-nba.appspot.com/ |
|
Back to top |
|
|
Crow
Joined: 20 Jan 2009 Posts: 825
|
Posted: Fri Feb 04, 2011 2:08 am Post subject: |
|
|
Thanks.
10 guys estimated at +2 better on total rebounds impact per 100 possessions at each end of the court. Less than 2%.
Only about 15% of the 538 players listed are estimated at +1 rebound or better on this and about 15% are -1 or worse. So it is estimated, by this regularized model, that about 70% of players are between +1 and -1. FWIW, it appears there were even fewer values beyond equivalent thresholds in Joe Sill's earlier estimates.
Food for thought about the relative importance of individual rebounding and maybe the degree of regularization. |
|
Back to top |
|
|
Statman
Joined: 20 Feb 2005 Posts: 242 Location: Arlington, Texas
|
Posted: Fri Feb 04, 2011 5:51 am Post subject: |
|
|
Crow wrote: | Thanks.
10 guys estimated at +2 better on total rebounds impact per 100 possessions at each end of the court. Less than 2%.
Only about 15% of the 538 players listed are estimated at +1 rebound or better on this and about 15% are -1 or worse. So it is estimated, by this regularized model, that about 70% of players are between +1 and -1. FWIW, it appears there were even fewer values beyond equivalent thresholds in Joe Sill's earlier estimates.
Food for thought about the relative importance of individual rebounding and maybe the degree of regularization. |
Which is why Berri's stuff often looks so absurd to us (well, at least me) - valuing player rebounds at such a high rate. _________________ Dan
My current national college player rankings (and other stuff):
http://www.pointguardu.com/f136/statmans-ratings-56243/index6.html#post355594 |
|
Back to top |
|
|
back2newbelf
Joined: 21 Jun 2005 Posts: 276
|
Posted: Fri Feb 04, 2011 7:45 am Post subject: |
|
|
Crow wrote: | 10 guys estimated at +2 better on total rebounds impact per 100 possessions at each end of the court. |
I should have been clearer. It's not per possession. It's per "rebound opportunity". A "rebound opportunity" is every time a player of your team misses a shot that can be rebounded by your team (basically every shot except technical- and "first of two"-freethrows).
Compared to APM with points a missed shot (that can be rebounded) by Team A is equivalent to a possession by Team A. Team A getting the subsequent offensive rebound would be equivalent to Team A scoring one point.
I've also thought about doing this without all the available rebounds from free throws(when the last free throw was missed). I would imagine those kind of rebounds get very rarely rebounded by the offensive team and the guards of the offensive team have almost no influence on those. Thus, guards from teams that shoot (and miss) a lot of free throws get unfairly punished _________________ http://stats-for-the-nba.appspot.com/ |
|
Back to top |
|
|
EvanZ
Joined: 22 Nov 2010 Posts: 307
|
Posted: Fri Feb 04, 2011 8:29 am Post subject: |
|
|
back2newbelf wrote: |
I've also thought about doing this without all the available rebounds from free throws(when the last free throw was missed). I would imagine those kind of rebounds get very rarely rebounded by the offensive team and the guards of the offensive team have almost no influence on those. Thus, guards from teams that shoot (and miss) a lot of free throws get unfairly punished |
FWIW, I don't count give credit for dreb off free throws for similar reasons. I do give credit for oreb, though. _________________ http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir |
|
Back to top |
|
|
Mike G
Joined: 14 Jan 2005 Posts: 3623 Location: Hendersonville, NC
|
Posted: Fri Feb 04, 2011 10:24 am Post subject: Re: Adjusted rebounding |
|
|
back2newbelf wrote: | ...
Home teams grab more available offensive rebounds (30.4% vs 29.7%).
|
Hmm, how is it that league avg ORb% is .263 ?
Quote: | I've also thought about doing this without all the available rebounds from free throws(when the last free throw was missed). I would imagine those kind of rebounds get very rarely rebounded by the offensive team |
Even then, off missed FG the league ORb% is around .280.
What is it I'm missing? _________________ `
36% of all statistics are wrong |
|
Back to top |
|
|
Ilardi
Joined: 15 May 2008 Posts: 265 Location: Lawrence, KS
|
Posted: Fri Feb 04, 2011 10:38 am Post subject: |
|
|
back2newbelf -
Nice work (as usual). Have you taken a look to see how your regularized adjusted rebound rates compare with simple ORR% and DRR%? I would guess the correlations must be pretty high, but would love to see the actual r's if you or anyone else have time to check. |
|
Back to top |
|
|
back2newbelf
Joined: 21 Jun 2005 Posts: 276
|
Posted: Fri Feb 04, 2011 10:50 am Post subject: Re: Adjusted rebounding |
|
|
Mike G wrote: | Quote: | I've also thought about doing this without all the available rebounds from free throws(when the last free throw was missed). I would imagine those kind of rebounds get very rarely rebounded by the offensive team |
Even then, off missed FG the league ORb% is around .280.
What is it I'm missing? |
I didn't do this yet
Mike G wrote: |
back2newbelf wrote: | ...
Home teams grab more available offensive rebounds (30.4% vs 29.7%).
| Hmm, how is it that league avg ORb% is .263 ?
|
Is it .263 of *all rebounds* or just of those that can actually be rebounded by the attacking team? I don't count it as rebounding opportunity if the first of two free throws is missed, but it's always listed as a team rebound for the defending team _________________ http://stats-for-the-nba.appspot.com/ |
|
Back to top |
|
|
mystic
Joined: 18 Dec 2006 Posts: 21
|
Posted: Fri Feb 04, 2011 11:05 am Post subject: |
|
|
Ilardi wrote: | back2newbelf -
Nice work (as usual). Have you taken a look to see how your regularized adjusted rebound rates compare with simple ORR% and DRR%? I would guess the correlations must be pretty high, but would love to see the actual r's if you or anyone else have time to check. |
I just ran a linear regression with ORB% obtained by boxscores and those values for offensive rebounds:
Code: | Model Summary
Model R R Square Adjusted R Square Std. Error of the Estimate
1 ,363a ,132 ,130 ,7704233
a. Predictors: (Constant), ORB%
|
Same for DRB%
Code: | Model Summary
Model R R Square Adjusted R Square Std. Error of the Estimate
1 ,318a ,101 ,100 ,4245693
a. Predictors: (Constant), DRB%
|
And for the combination:
Code: | Model Summary
Model R R Square Adjusted R Square Std. Error of the Estimate
1 ,366a ,134 ,132 ,8530314
a. Predictors: (Constant), TRB%
|
Doesn't look like a high correlation to me. Which even gives more fuel to the "why Berri overrates rebounds" conversation. _________________ http://bbmetrics.wordpress.com |
|
Back to top |
|
|
Ilardi
Joined: 15 May 2008 Posts: 265 Location: Lawrence, KS
|
Posted: Fri Feb 04, 2011 11:42 am Post subject: |
|
|
Thanks, mystic. Wow, r's down in the .32-.36 range. I would have guessed much higher.
So, if we were to fully trust these adjusted rebounding numbers, they would imply that simple ORR% and DRR% figures are misleading as often than not, i.e., that a player's rebounding performance is so heavily contextual - dependent on teammates and opponents - that simple unadjusted rebound rate is nearly uninterpretable on its own.
The implications are not trivial . . . |
|
Back to top |
|
|
EvanZ
Joined: 22 Nov 2010 Posts: 307
|
|
Back to top |
|
|
Ilardi
Joined: 15 May 2008 Posts: 265 Location: Lawrence, KS
|
Posted: Fri Feb 04, 2011 11:55 am Post subject: |
|
|
A few of the adjusted rebounding values are so surprising that I'm having trouble understanding how they could be valid. For example, Dwight Howard doesn't even place among the top 50, while he ranks #3 among NBA starters in overall rebounding percentage. Any thoughts? |
|
Back to top |
|
|
DSMok1
Joined: 05 Aug 2009 Posts: 611 Location: Where the wind comes sweeping down the plains
|
Posted: Fri Feb 04, 2011 11:57 am Post subject: |
|
|
Wow, those are low r-values!
I wonder if that's why, when I ran my ASPM regressions, rebounding was weighted so lightly? _________________ GodismyJudgeOK.com/DStats
Twitter.com/DSMok1 |
|
Back to top |
|
|
EvanZ
Joined: 22 Nov 2010 Posts: 307
|
Posted: Fri Feb 04, 2011 12:04 pm Post subject: |
|
|
This is probably germaine.
When I take the REB100 from ezPM and correlate it with ORR (which is the offensive rebounding rate for a player relative to his counterpart), the R^2 is 0.43. When I do the same thing for DRR, the R^2 is 0.0001.
It's sort of blowing my mind. Either I'm giving not enough credit for defensive rebounds, or this has something to do with diminishing returns. I guess. _________________ http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir |
|
Back to top |
|
|
back2newbelf
Joined: 21 Jun 2005 Posts: 276
|
Posted: Fri Feb 04, 2011 12:06 pm Post subject: |
|
|
Everyone, please don't jump to conclusions. There might be an error in my code as there seem to be discrepancies with this and average OReb%, it might still be a not-so-good approximation or maybe the methodology isn't perfect for that kind of problem.
All things considered, we can only know if this is useful data if we compare it to standard OReb/DReb% (averaged over all on-court-players) as an predictor for out of sample data, which isn't hard to do but I don't have much time _________________ http://stats-for-the-nba.appspot.com/ |
|
Back to top |
|
|
|