This is Google's cache of viewtopic.php?p=31051&sid=da9e0d9516063d5ea98a00b831ac0ba6. It is a snapshot of the page as it appeared on Apr 2, 2011 02:31:09 GMT. The current page could have changed in the meantime. Learn more

Text-only version
These search terms are highlighted: bayesian adjusted plus minus  
APBRmetrics :: View topic - Dwight Howard's Adjusted Plus/Minus
APBRmetrics Forum Index APBRmetrics
The statistical revolution will not be televised.
 
 FAQFAQ   SearchSearch   MemberlistMemberlist   UsergroupsUsergroups   RegisterRegister 
 ProfileProfile   Log in to check your private messagesLog in to check your private messages   Log inLog in 

Dwight Howard's Adjusted Plus/Minus
Goto page 1, 2  Next
 
Post new topic   Reply to topic    APBRmetrics Forum Index -> General discussion
View previous topic :: View next topic  
Author Message
erivera7



Joined: 19 Jan 2009
Posts: 182
Location: Chicago, IL

PostPosted: Sat Mar 27, 2010 4:44 pm    Post subject: Dwight Howard's Adjusted Plus/Minus Reply with quote

Not sure if anyone else has been keeping tabs with Dwight Howard's adjusted plus/minus this year but is it usually normal for someone to make that big of a leap in this metric? He's gone from +1.04 in 2009 to +20.35 (first in the NBA, coincidentally), standard errors notwithstanding. I know that Durant's APM has gotten pub this year and he's made a similar jump, too, but I was wondering what people thought of Howard's leap?

Does that say more that last year was a fluke, statistically, for him or does that say more to what he's accomplished this year?
_________________
@erivera7
I cover the Orlando Magic - Magic Basketball
Back to top
View user's profile Send private message
Joe



Joined: 27 Sep 2009
Posts: 94
Location: Long Island, NY

PostPosted: Sun Mar 28, 2010 2:20 am    Post subject: Reply with quote

BV has a lot of teams that are incredibly skewed towards frontcourt players. See the Nuggets for another example. Do you believe that Howard/Gortat are the only good players on the Magic and Nene/KMart/Bird are the only good players on the Nuggets or is this perhaps a case of the APM not properly adjusting for what happens when teams play small ball?
_________________
http://www.hoopdata.com
Back to top
View user's profile Send private message Visit poster's website
DSMok1



Joined: 05 Aug 2009
Posts: 611
Location: Where the wind comes sweeping down the plains

PostPosted: Sun Mar 28, 2010 8:21 am    Post subject: Reply with quote

Joe wrote:
BV has a lot of teams that are incredibly skewed towards frontcourt players. See the Nuggets for another example. Do you believe that Howard/Gortat are the only good players on the Magic and Nene/KMart/Bird are the only good players on the Nuggets or is this perhaps a case of the APM not properly adjusting for what happens when teams play small ball?


I've noticed this consistently. When teams have their front court and back court rotate independently (no overlap), there can be a disconnect between them in the APM.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
schtevie



Joined: 18 Apr 2005
Posts: 412

PostPosted: Sun Mar 28, 2010 10:00 am    Post subject: Reply with quote

Ummmmm...1 year APM is noisy, mostly because of multicollinearity issues.

Dwight Howard and Rashard Lewis play a lot together - 68% of DH's minutes this year and 76% last year. Last year RL had a 1 year APM of 10.19 to DH's 1.04. This year, the big switcheroo of -4.84 and 20.35. And the two year average APM is 5.74 and 6.70.

Checking Joe Sill's RAPM data, you get the picture that DH hasn't changed much at all between this year and last but that RH's performance has dipped. And the raw data hints that this relates to a decline in offensive performance.

The case of Kevin Durant is a bit different....
Back to top
View user's profile Send private message
erivera7



Joined: 19 Jan 2009
Posts: 182
Location: Chicago, IL

PostPosted: Sun Mar 28, 2010 3:02 pm    Post subject: Reply with quote

schtevie wrote:
Ummmmm...1 year APM is noisy, mostly because of multicollinearity issues.

Dwight Howard and Rashard Lewis play a lot together - 68% of DH's minutes this year and 76% last year. Last year RL had a 1 year APM of 10.19 to DH's 1.04. This year, the big switcheroo of -4.84 and 20.35. And the two year average APM is 5.74 and 6.70.

Checking Joe Sill's RAPM data, you get the picture that DH hasn't changed much at all between this year and last but that RH's performance has dipped. And the raw data hints that this relates to a decline in offensive performance.

The case of Kevin Durant is a bit different....

I know it's noisy, I just wanted to know what was causing the sudden jump in Howard's APM. Seems the consensus I've gotten, so far, [edited to avoid confusion] via a few PM's has been due to Lewis' decline.
_________________
@erivera7
I cover the Orlando Magic - Magic Basketball


Last edited by erivera7 on Sun Mar 28, 2010 4:48 pm; edited 1 time in total
Back to top
View user's profile Send private message
DLew



Joined: 13 Nov 2006
Posts: 224

PostPosted: Sun Mar 28, 2010 4:25 pm    Post subject: Reply with quote

Seems to me the consensus is random variation...
Back to top
View user's profile Send private message
erivera7



Joined: 19 Jan 2009
Posts: 182
Location: Chicago, IL

PostPosted: Sun Mar 28, 2010 4:46 pm    Post subject: Reply with quote

DLew wrote:
Seems to me the consensus is random variation...

I've had a few people PM me about Dwight Howard's APM and determine that Rashard Lewis' decline is the main factor (that and last year looks more like a fluke with Howard with regards to his low APM), sorry for not being specific in my previous reply. That was my fault. I apologize for the confusion.
_________________
@erivera7
I cover the Orlando Magic - Magic Basketball
Back to top
View user's profile Send private message
DSMok1



Joined: 05 Aug 2009
Posts: 611
Location: Where the wind comes sweeping down the plains

PostPosted: Mon Mar 29, 2010 8:16 am    Post subject: Reply with quote

schtevie wrote:
Ummmmm...1 year APM is noisy, mostly because of multicollinearity issues.

Dwight Howard and Rashard Lewis play a lot together - 68% of DH's minutes this year and 76% last year. Last year RL had a 1 year APM of 10.19 to DH's 1.04. This year, the big switcheroo of -4.84 and 20.35. And the two year average APM is 5.74 and 6.70.

Checking Joe Sill's RAPM data, you get the picture that DH hasn't changed much at all between this year and last but that RH's performance has dipped. And the raw data hints that this relates to a decline in offensive performance.

The case of Kevin Durant is a bit different....


It's a spin on multicollinearity: when positions do not overlap. Exactly 1 of Gortat and Howard always play. Exactly 1 of Lewis, Anderson, and Bass always play. And these positions do not overlap with the backcourt. What does this mean? The plus/minus of these positions is decoupled from the plus/minus of the other positions.

See?
Center: Gortat is +7.66 this year, was -8.06 last year.
Howard is +20.33 this year, 1.04 last year.

Forward: Lewis is -4.9 this year, was +10.19 last year. (The others don't have enough data.)

Note that Gortat both last year and this year is about ~16 points worse than Howard... but the 2 of them have both moved on the APM scale.

I don't think this decoupling aspect of APM has been addressed properly. There are a number of teams where certain players play 1 position only, or 2 positions only. It is possible for these groups to vary inversely from the rest of the team with essentially no impact on the APM error calculation. They are decoupled.

Is there any way to fix this decoupling problem?
Back to top
View user's profile Send private message Send e-mail Visit poster's website
gabefarkas



Joined: 31 Dec 2004
Posts: 1313
Location: Durham, NC

PostPosted: Thu Apr 01, 2010 2:44 pm    Post subject: Reply with quote

DSMok1 - I brought up the same issue a few months ago in another thread. I don't think there's a good way to completely handle it. Ridge Regression (a la Joe Sill) is an interesting and potentially viable approach.
Back to top
View user's profile Send private message Send e-mail AIM Address
DSMok1



Joined: 05 Aug 2009
Posts: 611
Location: Where the wind comes sweeping down the plains

PostPosted: Thu Apr 01, 2010 3:52 pm    Post subject: Reply with quote

gabefarkas wrote:
DSMok1 - I brought up the same issue a few months ago in another thread. I don't think there's a good way to completely handle it. Ridge Regression (a la Joe Sill) is an interesting and potentially viable approach.


I'm not completely sold on that; it is a Bayesian style approach, but it has the issue of bringing all players toward 0--including scrubs. If there were a way to pull toward a sliding scale based on, say, mpg or something...
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Ryan J. Parker



Joined: 23 Mar 2007
Posts: 711
Location: Raleigh, NC

PostPosted: Thu Apr 01, 2010 4:03 pm    Post subject: Reply with quote

DSMok1 wrote:
it is a Bayesian style approach, but it has the issue of bringing all players toward 0--including scrubs. If there were a way to pull toward a sliding scale based on, say, mpg or something...


this isn't mathematically possible. don't even try. it will never work.
_________________
I am a basketball geek.
Back to top
View user's profile Send private message Visit poster's website
DSMok1



Joined: 05 Aug 2009
Posts: 611
Location: Where the wind comes sweeping down the plains

PostPosted: Thu Apr 01, 2010 4:17 pm    Post subject: Reply with quote

Ryan J. Parker wrote:
DSMok1 wrote:
it is a Bayesian style approach, but it has the issue of bringing all players toward 0--including scrubs. If there were a way to pull toward a sliding scale based on, say, mpg or something...


this isn't mathematically possible. don't even try. it will never work.


Yeah, I know. However, here is one possibility: adjust the (regressed toward 0) ridge-regression model according to your (MPG perhaps?) Bayesian prior; we have error values from both the prior and the ridge-regressed APM. Would that work at all?
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Ryan J. Parker



Joined: 23 Mar 2007
Posts: 711
Location: Raleigh, NC

PostPosted: Thu Apr 01, 2010 4:19 pm    Post subject: Reply with quote

nope. bayesian methods were invented to handcuff the researcher. no flexibility at all.
_________________
I am a basketball geek.
Back to top
View user's profile Send private message Visit poster's website
Ilardi



Joined: 15 May 2008
Posts: 265
Location: Lawrence, KS

PostPosted: Thu Apr 01, 2010 4:25 pm    Post subject: Reply with quote

gabefarkas wrote:
DSMok1 - I brought up the same issue a few months ago in another thread. I don't think there's a good way to completely handle it. Ridge Regression (a la Joe Sill) is an interesting and potentially viable approach.


I've had some luck with simply using multiple seasons' worth of data to overcome the multicollinearity issue: once we have 5+ seasons in the model, the standard errors of estimate are quite small, and the Howard/Gortat problems tend to disappear. Of course, such an approach has its own inherent limitation: it assumes that player APM effects remain reasonably constant over long stretches of time, and we know this is not always the case (due to injury, maturation effects, age-related deterioration, etc.).

We can try, perhaps, to have the best of both worlds by using multiple seasons to disentangle player effects (i.e., address multicollinearity) and then weighting the model very heavily toward the most recent season, but there's no way to know a priori which of the infinitely many weighting schemes is optimal. Joe Sill's suggestion of using out-of-sample goodness of fit as the main arbiter of such choices strikes me as eminently reasonable, and one I hope the field will adopt.
Back to top
View user's profile Send private message
DSMok1



Joined: 05 Aug 2009
Posts: 611
Location: Where the wind comes sweeping down the plains

PostPosted: Fri Apr 02, 2010 7:48 am    Post subject: Reply with quote

Ryan J. Parker wrote:
nope. bayesian methods were invented to handcuff the researcher. no flexibility at all.


Confused

I thought them rather useful... but then, I'm not a theoretical statistics expert.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Display posts from previous:   
Post new topic   Reply to topic    APBRmetrics Forum Index -> General discussion All times are GMT - 5 Hours
Goto page 1, 2  Next
Page 1 of 2

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum


Powered by phpBB © 2001, 2005 phpBB Group