|
APBRmetrics The statistical revolution will not be televised.
|
View previous topic :: View next topic |
Author |
Message |
Mike G
Joined: 14 Jan 2005 Posts: 3612 Location: Hendersonville, NC
|
Posted: Mon Oct 22, 2007 8:16 am Post subject: Page, Fellingham, and Reese: JQAS paper |
|
|
From the Journal of Quantitative Analysis in Sports:
http://www.bepress.com/jqas/vol3/iss4/1/
Using Box-Scores to Determine a Position's Contribution to Winning Basketball Games
Garritt L. Page, Gilbert W. Fellingham, C. Shane Reese
The study looks at boxscore totals for a game, by position, and determines correlations with winning. These are translated into value per item for some things -- FG, FT, 3FG, Ast, Stl, TO; and into value per .01 for other quantities -- OReb%, DReb%, FG%, FT%. The value is expressed as 'points per 100 possessions'.
This seems to place an entirely different scale on those stats considered as a % , so I'm listing them separately. (I realize I may be interpreting some of this wrong, and I have no idea about the methodology.)
Data were from USAToday boxscores, from the 1996-97 season. These boxscores don't list Blocks or 3FGA. 3FG were determined by (Pts - 2*FG - FT). Only starters positions are listed, so everyone else's position is lumped in under 'Bench'.
I've hand-entered the data, dropped some digits, and ignored the uncertainty ranges. Here, it seems, are the positions and the relative importance of the stats, arranged by position.
Code: | Center Power Fwd Small Fwd Shooting Gd Point Guard Bench
.379 Stl .309 Ast .405 Ast .374 3FG .353 Ast .132 Ast
.325 Ast .284 TO .381 3FG .326 Ast .283 3FG .090 Stl
.254 TO .150 3FG .345 TO .215 TO .269 TO .084 TO
.135 FG .146 Stl .257 Stl .144 Stl .199 Stl .067 3FG
.067 FT .086 FT .104 FT .099 FT .144 FT .053 FT
-.014 3FG -.008 FG -.003 FG -.069 FG -.007 FG -.005 FG
|
Again, these are total FG, FT, and 3FG (not percents). Turnovers are negative. _________________ `
36% of all statistics are wrong |
|
Back to top |
|
|
ziller
Joined: 30 Jun 2005 Posts: 126 Location: Sac Metro
|
Posted: Mon Oct 22, 2007 12:40 pm Post subject: |
|
|
The conclusions drawn seem a bit stretched to me. Particularly w/r/t assists: There are causality issues without doubt. The authors don't address (mention? sort-of, near the end) that assists rely on made baskets, which clearly have an impact on point spread. If PG #1 has 12 assists and PG #2 has 3 assists, and Team #1 wins by three... is it because PG #1 had nine more assists, or is it because Team #1 shot better (thus earning PG #1 more assists). We don't know based on these findings, I'd said.
This issue also screams out with steals from the center position. Is that indicative of a defensive center with good hands, or a poor offense on the other side? Has anyone ever looked to see if steal rates for centers (or even PFs) are consistent on a season-by-season basis? _________________ SactownRoyalty.com
tziller@gmail.com |
|
Back to top |
|
|
Mountain
Joined: 13 Mar 2007 Posts: 1527
|
Posted: Mon Oct 22, 2007 1:14 pm Post subject: |
|
|
I mentioned the original thesis that served as the basis for this slimmer, revised article 2-3 weeks ago (among others) in the New England Symposium thread.
http://tinyurl.com/235wy6
I speculated it deserved to appear in a journal, perhaps JQAS. The only comment at the time was from Mike G. joking about "the posterior means" I pulled out for ease of viewing for readers here. Now the article comes out and Mike has pulled the same set (with updated values) as I did. Kinda funny how that worked out. Good that the article is out there more clearly now to get whatever comments. |
|
Back to top |
|
|
John Hollinger
Joined: 14 Feb 2005 Posts: 175
|
Posted: Mon Oct 22, 2007 5:38 pm Post subject: |
|
|
I believe the rebounds from guards is another area where this is an issue. Guards' rebounds tend to be almost entirely defensive boards; defensive boards, in turn, tend to require a shot to be missed by the opponent. |
|
Back to top |
|
|
Mike G
Joined: 14 Jan 2005 Posts: 3612 Location: Hendersonville, NC
|
Posted: Tue Oct 23, 2007 4:45 am Post subject: |
|
|
Yeah, I thought these numbers looked familiar. Also, it's illuminating to sort and group them differently. Here they are grouped by event, sorted by position of 'importance':
Code: | Assists Turnovers 3-Pointers Steals FT made FG made
.405 SF -.345 SF .381 SF .379 C .144 PG .135 C
.353 PG -.284 PF .374 SG .257 SF .104 SF -.003 SF
.326 SG -.269 PG .283 PG .199 PG .099 SG -.005 Be
.325 C -.254 C .150 PF .146 PF .086 PF -.007 PG
.309 PF -.215 SG .067 Be .144 SG .067 C -.008 PF
.132 Be -.084 Be -.014 C .090 Be .053 Be -.069 SG
|
Other than turnovers, these negative correlations aren't strong; except for FG by the SG. I'm supposing that indicates that reliance on the SG for scoring isn't a good thing.
As I mentioned earlier, the following categories don't seem to be on a comparable scale (nor will they all fit into one table), so here they are separately. Points/100 possessions, for each .010 of difference: Code: | FG% OffReb% DefReb% FT%
.200 Be .113 PG .067 SG .011 C
.168 SG .112 PF .061 PG -.001 SF
.128 PG .076 Be .028 SF -.001 PG
.115 SF .070 SF -.005 PF -.002 PF
.078 PF .052 SG -.010 C -.004 SG
.053 C .052 C -.016 Be -.007 Be
|
And grouped by position, skills in order of importance: Code: | Center Power Fwd Small Fwd Shooting Gd Point Gd Bench
.053 FG% .112 OR% .115 FG% .168 FG% .128 FG% .200 FG%
.052 OR% .078 FG% .070 OR% .067 DR% .113 OR% .076 OR%
.011 FT% -.002 FT% .028 DR% .052 OR% .061 DR% -.007 FT%
-.010 DR% -.005 DR% -.001 FT% -.004 FT% -.001 FT% -.016 DR%
|
_________________ `
36% of all statistics are wrong |
|
Back to top |
|
|
Mountain
Joined: 13 Mar 2007 Posts: 1527
|
Posted: Tue Oct 23, 2007 11:10 am Post subject: |
|
|
Thanks for the different data sortings. There are clues in there, though I don't know how much weight to put on the results.
Looking at these results and how specific teams were designed and operate may spur additional thoughts, new questions. Using Jorje's hoopstats.com tools in conjunction with the values on these tables, teams can be analyzed by position instead of by player, You can compare skill emphasis strategies suggested as being more favorable on paper, in aggregrate to what happens on the court with particular teams and players and ponder what to change...
A strong SF seems valuable, perhaps because it is comparatively rarer because of quantity of players that get that label and stronger correlation of skill & opportunity at other positions? |
|
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
|