|
APBRmetrics The statistical revolution will not be televised.
|
View previous topic :: View next topic |
Author |
Message |
Roland_Beech
Joined: 14 Jan 2005 Posts: 43
|
Posted: Fri Jan 14, 2005 2:37 pm Post subject: Player Assisted % on made shots |
|
|
so looking over the Ast'd % for players (that is the % of their own shots on which an assist was recorded), the player creating his own shot the least right now is (min 50 FGM)
98% Fred Hoiberg (52 of 53 made FG's were assisted)
90% Brent Barry
89% Greg Buckner
89% Bruce Bowen
...and at the opposite end
22% Nash
23% Francis
23% Iverson
25% Marbury
what prompted this was someone commenting on Kobe's 30% mark.
Anyway, I tried to paste in some excel data showing all players with breakouts by shot type but it doesn't format properly...can we do attachments on this forum?
I guess I'll dump it on the site:
http://www.82games.com/nbaplayerassistednumbers.xls |
|
Back to top |
|
|
Ed Küpfer
Joined: 30 Dec 2004 Posts: 785 Location: Toronto
|
Posted: Fri Jan 14, 2005 3:05 pm Post subject: Re: Player Assisted % on made shots |
|
|
Roland_Beech wrote: | Anyway, I tried to paste in some excel data showing all players with breakouts by shot type but it doesn't format properly... |
Did you try the CODE tag? It formats text in monospace Code: | f o n t s
l i k e
t h i s |
Roland_Beech wrote: | can we do attachments on this forum? |
No.
Thanks for that! _________________ ed |
|
Back to top |
|
|
KnickerBlogger
Joined: 30 Dec 2004 Posts: 180
|
Posted: Fri Jan 14, 2005 4:31 pm Post subject: |
|
|
Ed pointed me to tabs2spaces.
Thanks to this board I can actually search & find the original post,
and the link to the software. Very nice.
Just paste excel into notepad -> select All -> Copy (to clipboard) -> Open tabs2psaces -> choose clipboard --> then paste into here -> use CODE (see above).
Much quicker to do than typing or reading that. |
|
Back to top |
|
|
Kevin Pelton Site Admin
Joined: 30 Dec 2004 Posts: 979 Location: Seattle
|
Posted: Fri Jan 14, 2005 4:57 pm Post subject: |
|
|
Early last season, I typed in all the assisted field goal percentages and tried to come up with a relatively simple formula to predict it (Dean has his own, but I'm pretty sure it's a lot more complicated).
I came up with the following:
aFG% = .42 - .63*assist share - .89*(FTA/Min) + 2/3*(team aFG%)
(assist share is the estimated percentage of assists a player is responsible for when he's on the floor -- this is to reflect the fact that you can't assist yourself)
Here are my estimates for Roland's top and bottom players:
Code: | Player aFG est
------------------
Hoiberg .98 .72
Barry .90 .63
Buckner .89 .69
Bowen .89 .74
Nash .22 .32
Francis .23 .33
Iverson .23 .36
Marbury .25 .34 |
And here are my estimated top and bottom guys (minimum 250 minutes):
Code: | Player aFG est
--------------------
Ostertag .41 .78
Nesterovic .68 .78
Freije .85 .77
Olowokandi .59 .77
Knight .24 .31
Nash
Francis
Marbury |
Okay, so clearly I'm underestimating the effect of
a. Taking a lot of three-pointers
b. Putbacks
I'll have to see if I can do something better with Roland's file. |
|
Back to top |
|
|
Ed Küpfer
Joined: 30 Dec 2004 Posts: 785 Location: Toronto
|
Posted: Sat Jan 15, 2005 4:10 pm Post subject: |
|
|
Here's Roland's data, with players <20 FGM removed, shown in a boxplot.
_________________ ed |
|
Back to top |
|
|
Kevin Pelton Site Admin
Joined: 30 Dec 2004 Posts: 979 Location: Seattle
|
Posted: Mon Jan 17, 2005 9:07 pm Post subject: |
|
|
After poking around the data a bit, I've come up with a vastly improved formula:
aFG%/tmaFG% = 1.53 - 1.442*assist share -.041*Oreb/48 - .787*Ps% + .014*Spec
This now predicts the ratio of a player's assisted field goal percentage and his team's overall percentage with an r^2 of .566.
The key was coming up with the three-point specialist rating. I settled on (3A/FGA)^2 * (1/Ps%)^2. The less often you shoot, and the more often you shoot threes, the more of a three-point specialist you are. This fixes much of the problem I was having with underrated the assisted percentage of these guys:
Code: | Player aFG est
------------------
Hoiberg .98 .96
Barry .90 .79
Buckner .89 .76
Bowen .89 .75
Jacobsen .77 .91
McCarty .88 .90
Russell .90 .97
Nachbar .90 .94
Korver .85 .82
Bowen .89 .89 |
|
|
Back to top |
|
|
Ed Küpfer
Joined: 30 Dec 2004 Posts: 785 Location: Toronto
|
Posted: Mon Jan 17, 2005 9:21 pm Post subject: |
|
|
What's that? _________________ ed |
|
Back to top |
|
|
Kevin Pelton Site Admin
Joined: 30 Dec 2004 Posts: 979 Location: Seattle
|
Posted: Mon Jan 17, 2005 9:34 pm Post subject: |
|
|
Possession percentage (which in this case is actually represented as from zero to one, thus making it not a percentage, but whatever). |
|
Back to top |
|
|
Ed Küpfer
Joined: 30 Dec 2004 Posts: 785 Location: Toronto
|
Posted: Mon Jan 17, 2005 9:46 pm Post subject: |
|
|
If you want to try the 03-04 numbers, I have them here. Else, I'll give it a shot tomorrow. _________________ ed |
|
Back to top |
|
|
Mike G
Joined: 14 Jan 2005 Posts: 3590 Location: Hendersonville, NC
|
Posted: Mon Aug 07, 2006 6:26 am Post subject: |
|
|
admin wrote: |
aFG%/tmaFG% = 1.53 - 1.442*assist share -.041*Oreb/48 - .787*Ps% + .014*Spec
This now predicts the ratio of a player's assisted field goal percentage and his team's overall percentage with an r^2 of .566.
The key was coming up with the three-point specialist rating. I settled on (3A/FGA)^2 * (1/Ps%)^2. |
Hey, I found it!
Kevin, can you explain these terms?
-- assist share is a players % of team assists?
-- Ps% (possession %) is what? Usage Rate?
I would like to tinker with coefficients and the exponent, for best fit. Then I could do historical estimates of %AstFG, or actually %UnAstFG. This seems to nicely separate the values of Pts and Ast, within a team. _________________ `
36% of all statistics are wrong |
|
Back to top |
|
|
Kevin Pelton Site Admin
Joined: 30 Dec 2004 Posts: 979 Location: Seattle
|
Posted: Mon Aug 07, 2006 10:04 am Post subject: |
|
|
Mike G wrote: | -- assist share is a players % of team assists? |
Yes.
Quote: | -- Ps% (possession %) is what? Usage Rate? |
Yes, but my own flavor of it -- FGA + (.44*FTA) + TO + .25*AST - .25*(some other estimate of assisted field goals made) |
|
Back to top |
|
|
Mike G
Joined: 14 Jan 2005 Posts: 3590 Location: Hendersonville, NC
|
Posted: Wed Oct 04, 2006 9:44 am Post subject: |
|
|
OK, only a couple of months later I've got a maybe decent estimator of Unassisted-FG %. That is to say, the % of a player's FG (made) which were unassisted:
uAst% = FTA*.50 + OReb*.34 + Ast*1.64 + Stl*2.27 + TO*2.78 - (3FG/FG)*.127
All these terms are per-minute, excepting the final one, the % of a player's FG which were 3's.
I have no idea why steals and turnovers correlate positively with shot-creation; perhaps they're all indicators of 'independence' on the court.
Here's a comparison of Estimated uAst% and actual, for the top 15 FG-makers in the league last year; in order of 'error' (overestimation):
Code: | player team est% uAs% err
hamilton,richard det .42 .24 .18
anthony,carmelo den .52 .37 .15
wade,dwyane mia .81 .66 .15
pierce,paul bos .64 .52 .12
marion,shawn pho .34 .24 .10
carter,vince njn .56 .50 .06
arenas,gilbert was .68 .63 .05
bryant,kobe lal .60 .56 .04
brand,elton lac .44 .40 .04
jamison,antawn was .29 .29 .00
redd,michael mil .41 .41 .00
james,lebron cle .66 .68 -.02
iverson,allen phi .73 .76 -.03
richardson,jason gsw .42 .45 -.03
allen,ray sea .42 .50 -.08
nowitzki,dirk dal .39 .49 -.10
|
The average error is right at the league median of .08
With these elite scorers, their uAst% is mostly overestimated. For non-scoring types, it seems to be reversed. (total [+ and -] error = 0)
Here's the same bunch, sorted in order of 'actual' unassistedFG%:
Code: | Player team est% uAst err
iverson,allen phi .73 .76 -.03
james,lebron cle .66 .68 -.02
wade,dwyane mia .81 .66 .15
arenas,gilbert was .68 .63 .05
bryant,kobe lal .60 .56 .04
pierce,paul bos .64 .52 .12
carter,vince njn .56 .50 .06
allen,ray sea .42 .50 -.08
nowitzki,dirk dal .39 .49 -.10
richardson,jason gsw .42 .45 -.03
redd,michael mil .41 .41 .00
brand,elton lac .44 .40 .04
anthony,carmelo den .52 .37 .15
jamison,antawn was .29 .29 .00
hamilton,richard det .42 .24 .18
marion,shawn pho .34 .24 .10
|
The order isn't too far off, though I've greatly overestimated the unAst% of Wade, Melo, Pierce, and Hamilton; underestimated Dirk and Ray.
I didn't follow Kevin's lead of incorporating Ast/TmAst; I wanted to try getting decent estimates without that step. _________________ `
36% of all statistics are wrong |
|
Back to top |
|
|
Mike G
Joined: 14 Jan 2005 Posts: 3590 Location: Hendersonville, NC
|
Posted: Wed Oct 04, 2006 10:56 am Post subject: |
|
|
Using the formula explained above, here are comparisons with uAst% for the lowest 15 players in the league in that measure. (From among 285 players who made at least 100 FG): Code: | Player team est% uAs% err
collins,jason njn .24 .09 .15
lafrentz,raef bos .19 .11 .08
robinson,cliff njn .22 .11 .11
korver,kyle phi .22 .12 .10
thomas,kurt pho .27 .13 .14
posey,james mia .15 .14 .01
garrity,pat orl .17 .14 .03
buckner,greg den .28 .15 .13
vujacic,sasha lal .28 .15 .13
jones,james pho .14 .15 -.01
bowen,bruce san .17 .15 .02
jones,jumaine cha .22 .17 .05
ross,quinton lac .28 .17 .11
marshall,donyell cle .22 .18 .04
songaila,darius chi .41 .19 .22
|
The formula tends not to predict quite such low rates as we find in the league. Songaila's unprediction is among the 10 worst, while James Jones' is somehow right in line.
May as well see the predictability of the 15 shot-creatingest players: Code: | Player team est% uAs% err
paul,chris nor .75 .79 -.04
davis,baron gsw .77 .77 .00
nash,steve pho .83 .77 .06
iverson,allen phi .73 .76 -.03
arroyo,carlos orl .53 .75 -.22
ford,t.j. mil .68 .73 -.05
marbury,stephon nyk .61 .73 -.12
francis,steve orl .65 .73 -.08
tinsley,jamaal ind .71 .72 -.01
calderon,jose tor .60 .72 -.12
james,mike tor .54 .70 -.16
hudson,troy min .39 .69 -.30
telfair,sebastian por .56 .69 -.13
lue,tyronn atl .45 .68 -.23
knight,brevin cha .84 .68 .16 |
All point guards, and most are estimated not quite as high as they really are. It kinda seems the biggest deviants are members of sub-playoff teams.
Here's a slice out of the middle of the stack. Now limited to the 70 players who scored at least 1000 pts: Code: | err player team est% uAs%
.03 fisher,derek gsw .52 .49
.08 diaw,boris pho .57 .49
.02 howard,dwight orl .45 .47
.09 turkoglu,hedo orl .37 .46
.03 richardson,jason gsw .42 .45
.02 harrington,al atl .47 .45
.02 jackson,stephen ind .43 .45
.08 walker,antoine mia .36 .44
.06 bibby,mike sac .49 .43
.17 kirilenko,andrei uta .60 .43
.13 o'neal,shaquille mia .55 .42
.06 mobley,cuttino lac .36 .42
.00 redd,michael mil .41 .41
.04 lewis,rashard sea .37 .41
.00 ilgauskas,zydrun cle .41 .41
|
Without the extremes, we get an average error (column 1) of only .055 in this bunch. There are definitely differences between teams. For Dallas, virtually everyones uA% is 'overestimated'; for the Cavs, it's under. _________________ `
36% of all statistics are wrong |
|
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
|