|
APBRmetrics The statistical revolution will not be televised.
|
View previous topic :: View next topic |
Author |
Message |
DLew
Joined: 13 Nov 2006 Posts: 224
|
Posted: Thu Feb 07, 2008 2:08 pm Post subject: |
|
|
davis,
I really appreciate you posting these numbers regularly and I hope you continue to do so, I consider it a valuable resource. I think Dean's numbers are some of the best box score based numbers available.
I think Dan and I have made it clear that adjusted plus-minus has a myriad of issues, just look at the standard errors at basketballvalue, only a handful of players have ratings that are statistically different than average. |
|
Back to top |
|
|
cherokee_ACB
Joined: 22 Mar 2006 Posts: 157
|
Posted: Fri Feb 08, 2008 3:16 am Post subject: |
|
|
DLew wrote: |
I really appreciate you posting these numbers regularly and I hope you continue to do so, I consider it a valuable resource. |
Me too. |
|
Back to top |
|
|
Neil Paine
Joined: 13 Oct 2005 Posts: 774 Location: Atlanta, GA
|
Posted: Sat Feb 09, 2008 11:08 am Post subject: |
|
|
---
Last edited by Neil Paine on Sat Feb 16, 2008 11:24 am; edited 1 time in total |
|
Back to top |
|
|
HoopStudies
Joined: 30 Dec 2004 Posts: 706 Location: Near Philadelphia, PA
|
Posted: Mon Feb 11, 2008 8:00 am Post subject: |
|
|
davis21wylie2121 wrote: | ...
So, what does everyone think, are Dean's numbers merely descriptors and not measures of value? In fact, if DeanO is lurking -- once and for all, did you intend these numbers to be value measures or descriptors? Does the "skill curve" conception of player performance not fit with reality? Should I just stop posting these numbers altogether? |
Measures of value ARE just descriptors. Dave Berri says that because interseason correlations are high with his descriptor that they are measures of future value. Adj +/- advocates say that interseason correlations are low and that that measure is a descriptor of current season value.
How you put players together to make a team is the really tough part. Linearity assumptions and constant parameters are simplest and really where we are as a group. Skill curves add a tweak to using constant parameters, but really still work within a linear framework. Bob C's simulator is another approach that isn't linear but assumes certain constant parameters. The work I did on teammate fit suggests that even a simple game like fresco ball is NOT linear, but that linearity isn't horrible often.
So most things are descriptors. I hope that means that you keep posting... _________________ Dean Oliver
Author, Basketball on Paper
The postings are my own & don't necess represent positions, strategies or opinions of employers. |
|
Back to top |
|
|
Neil Paine
Joined: 13 Oct 2005 Posts: 774 Location: Atlanta, GA
|
Posted: Mon Feb 11, 2008 10:19 am Post subject: |
|
|
Thanks, Dean.
I guess I was just disheartened, thinking that I was perhaps barking up the wrong tree in terms of using ORtg/%Pos/DRtg as the fundamental (boxscore-based) tools for assessing a player's worth. But you're right, anything we do is obviously going to be a very simplified model of reality (as is the nature of any statistical model), and we're going to miss out on a lot of teammate-interaction effects as a result.
Then again, adjusted +/- theoretically does a better job of handling these non-linear aspects of the game, but the results are so noisy that I question the value of in-season adj +/- updates (in-season updates kinda being the entire premise of this thread).
So I'm going to stick with what I've been doing. It's not going to paint a perfectly accurate picture, but then again, which boxscore stat does? I just didn't want to be wasting my time on stats that people were dismissing out of hand. But most people seem to be appreciating it, so I'll keep updating whenever possible. |
|
Back to top |
|
|
HoopStudies
Joined: 30 Dec 2004 Posts: 706 Location: Near Philadelphia, PA
|
Posted: Mon Feb 11, 2008 10:44 am Post subject: |
|
|
davis21wylie2121 wrote: |
Then again, adjusted +/- theoretically does a better job of handling these non-linear aspects of the game, but the results are so noisy that I question the value of in-season adj +/- updates (in-season updates kinda being the entire premise of this thread).
|
It measures those aspects, albeit noisily. It doesn't provide a model of how those interactions change. It averages over the interactions that may be different in different combinations for the same team.
All of these things are tools towards understanding. None is the end in itself. _________________ Dean Oliver
Author, Basketball on Paper
The postings are my own & don't necess represent positions, strategies or opinions of employers. |
|
Back to top |
|
|
Mountain
Joined: 13 Mar 2007 Posts: 1527
|
Posted: Mon Feb 11, 2008 3:59 pm Post subject: |
|
|
Using ORtg/%Pos/DRtg as the fundamental (boxscore-based) tools for assessing a player's worth captures their value on that team in the role(s) assigned by the coach and the particular mix of lineups used in. Change the team, the coach directed role or lineup mix and player performance may change in various ways.
The ORtg/%Pos/DRtg of a player in subgroups of 5 man lineups might offer additional insight into player performance patterns under different conditions and expectations and into underlying abilities and could help guide future usage.
Change the team and the rollup data on previous team offers some general guidance but changes may occur. Understanding of a player at the 5 man lineup level might make projection of performance on a different team a little sharper- match up the 5 man lineup conditions the player will see in the rotation on the new team to the data of 5 man lienups most similar to that on the old team. That may be more informative the old team rollup average.
As others have suggested here and previously counterpart defensive ability and performance ideally should supplement the team defensive performance captured by "defensive rating". |
|
Back to top |
|
|
Harold Almonte
Joined: 04 Aug 2006 Posts: 616
|
Posted: Tue Feb 12, 2008 11:27 am Post subject: |
|
|
Every possession play is constituted by the: a) gain/lost of the possession, b) production/allowance of the scoring points, and c) the direct and non direct help/ non help. In the attempt to weight he help is when the troubles come. +/- weight it the same than the other part of the action, and Oliver probably weight it very high at the defensive end.
Bigs receive statistical defensive help from teammates, specially at rebounds, since these teammates could have less access to the rebounding zone; and these bigs help other teammates in the termination of a lot of shot defense and could be statistically reflected in some defensive metrics, but how much is the value of that and the frecuency? I've heard about aproximatedly 60/40 relationship between off/def rebounding (1 on 1); 60/40 between the inside/outside the painted zone usage, shots, and rebounds; 80/20 between the inside/outside the 3 point line usage and shots, but I suspect that the relationship between the help and the final action of a play (scoring, allowing, gaining or losing) is neither 80/20, nor 50/50. |
|
Back to top |
|
|
Neil Paine
Joined: 13 Oct 2005 Posts: 774 Location: Atlanta, GA
|
|
Back to top |
|
|
Neil Paine
Joined: 13 Oct 2005 Posts: 774 Location: Atlanta, GA
|
Posted: Sat Feb 23, 2008 12:26 pm Post subject: |
|
|
--- |
|
Back to top |
|
|
Neil Paine
Joined: 13 Oct 2005 Posts: 774 Location: Atlanta, GA
|
Posted: Sun Feb 24, 2008 11:45 am Post subject: |
|
|
--- |
|
Back to top |
|
|
Neil Paine
Joined: 13 Oct 2005 Posts: 774 Location: Atlanta, GA
|
Posted: Thu Feb 28, 2008 12:09 pm Post subject: |
|
|
--- |
|
Back to top |
|
|
Neil Paine
Joined: 13 Oct 2005 Posts: 774 Location: Atlanta, GA
|
|
Back to top |
|
|
Neil Paine
Joined: 13 Oct 2005 Posts: 774 Location: Atlanta, GA
|
Posted: Tue Mar 04, 2008 2:31 pm Post subject: |
|
|
--- |
|
Back to top |
|
|
Neil Paine
Joined: 13 Oct 2005 Posts: 774 Location: Atlanta, GA
|
Posted: Wed Mar 05, 2008 9:41 am Post subject: |
|
|
--- |
|
Back to top |
|
|
|
|
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
|