page 1 of 2
Author Message
DSMok1
Joined: 05 Aug 2009
Posts: 611
Location: Where the wind comes sweeping down the plains
PostPosted: Mon Jan 11, 2010 12:34 pm Post subject: NCAA Statistical Plus/Minus 2009-2010 Reply with quote
NCAA Statistical Plus/Minus 2009-2010
Statistical plus/minus has been an underused metric for evaluating college players--it is the perfect place to use it! SPM can be readily corrected for strength of schedule and is calculated directly from the stat line.
So I have adapted the latest version of the SPM formula (see "Building an Updated Statistical Plus/Minus Model for 2009" and the actual regression) to the NCAA, similar to what I did for last year (only corrected, improved and updated).
The theory is the same: compile NCAA stats, convert to tempo-free rates, run the SPM regression, and sum the team SPM's to their Pomeroy Efficiency Margin. Several fixes: I realized the SPM regression runs on per 40 minute NBA rates--which would be per 76 possessions. I converted the NCAA stats to that baseline. Also, the output of the SPM is in points/100 possessions--a fact which had me confused previously.
Also, I added in a Bayesian "True Talent" estimation based on the system I outlined on this forum. This was a regression based on team margin, MPG, age, and position, to estimate the SPM, creating a prior from inputs not included in SPM (except MPG). A standard error of the prior was also estimated (along with the standard error of each player's SPM). Basically, I tried to replicate Rosenbaum's method for the SPM standard errors.
The result is multifaceted. We have the pure SPM, which is per 100 player-possessions. We can look at the actual player contribution, which is per 100 TEAM-possessions (probably the best actual rating). We can look at the true talent estimate--which regresses too hard due to the high errors on the individual SPM's, but with more data and perhaps more years would be very useful. I created an overall rank based on a weighted average of the player's contribution to the team (x2) and their true talent estimate.
Access the Spreadsheet HERE. It is editable so you can sort it (let me know if that works)--please don't delete or modify the data. (Don't worry--I have backups!)
The top 30 Players:
Code:
Player Team Pos Yr SPM Contrib True Talent
Jon Scheyer Duke G Sr 14.86 13.20 11.16
Jackson Emery Brigham Young G Jr 17.25 10.83 8.37
Jarvis Varnado Mississippi St. F Sr 15.81 11.04 7.09
Trevon Hughes Wisconsin G Sr 12.21 9.54 8.01
Jimmy Butler Marquette G Jr 12.32 9.82 7.28
Jeremy Lin Harvard G Sr 14.69 11.57 6.36
Ekpe Udoh Baylor C Jr 10.72 9.14 7.60
James Anderson Oklahoma St. G Jr 13.21 10.98 6.40
Dominique Jones South Florida G Jr 12.73 11.26 6.29
Wes Johnson Syracuse F Jr 10.35 8.43 8.55
Robbie Hummel Purdue F Jr 12.19 8.80 7.58
Draymond Green Michigan St. F So 16.41 9.97 6.45
Damion James Texas F Sr 12.09 8.16 7.69
Jacob Pullen Kansas St. G Jr 12.20 8.88 6.74
Gordon Hayward Butler F So 10.59 8.77 6.91
Aubrey Coleman Houston G Sr 10.83 9.91 6.28
Jon Leuer Wisconsin F Jr 11.49 7.95 7.15
Damian Saunders Duquesne F Jr 13.00 11.35 5.89
Jimmer Fredette Brigham Young G Jr 11.03 7.53 7.70
Jerome Dyson Connecticut G Sr 9.59 8.12 6.67
Roman Martinez New Mexico F Sr 10.91 8.39 6.43
Kevin Jones West Virginia F So 9.46 7.51 7.51
Manny Harris Michigan G Jr 11.84 10.47 5.88
Ben Uzoh Tulsa G Sr 9.94 7.92 6.27
Greivis Vasquez Maryland G Sr 9.61 7.65 6.29
Kyle Singler Duke F Jr 8.02 6.80 9.24
David Schneider William & Mary G Sr 12.17 9.44 5.80
Lavoy Allen Temple F Jr 8.47 7.10 6.75
Greg Monroe Georgetown C So 8.39 7.07 7.06
Marqus Blakely Vermont F Sr 15.35 13.05 5.31
Lazar Hayward Marquette F Sr 9.93 7.32 6.37
Cole Aldrich Kansas C Jr 10.30 6.66 8.23
Damian Johnson Minnesota F Sr 11.35 7.29 6.38
Trevor Booker Clemson F Sr 9.76 7.08 6.58
John Wall Kentucky G Fr 8.96 6.99 6.82
Patrick Christopher California G Sr 8.23 6.69 7.30
Jeremy Hazell Seton Hall G Jr 11.72 9.24 5.69
Xavier Henry Kansas G Fr 9.43 6.63 7.69
Landry Fields Stanford G Sr 12.15 10.76 5.28
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Crow
Joined: 20 Jan 2009
Posts: 818
PostPosted: Mon Jan 11, 2010 4:02 pm Post subject: Reply with quote
So only about ten of the guys in the first round of major mock drafts are in the top 30 on overall rating. I will be looking more at who in the mock draft is way down the list on SPM or talent level or overall rating than who is not on the mock draft and yet high on these lists, though a few of such cases may be worth more attention. This is adjusted for strength of schedule yet the gap between top college performance and perceived top NBA potential remains. Probably mostly right but something to consider further.
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 Jan 11, 2010 5:06 pm Post subject: Reply with quote
Crow wrote:
So only about ten of the guys in the first round of major mock drafts are in the top 30 on this SPM. I haven't waded thru the file yet but will be looking more at who in the mock draft is way down the list on SPM than who is not on the mock draft and yet high on this SPM list, though a few of such cases may be worth more attention. This metric is adjusted for strength of schedule yet the gap between top college performance and perceived top NBA potential remains. Probably mostly right but something to consider further.
Do look at what year each player is--if they are in the top 30 as a freshman, then they're probably NBA material. Senior, not so much so. I haven't done the aging curves, but there is a significant climb. Also, remember that some players already are highly skilled, and some are rather raw still--while still posting good numbers. Some have a lot of untapped potential still; others are already nearly as good as they can get.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Crow
Joined: 20 Jan 2009
Posts: 818
PostPosted: Mon Jan 11, 2010 5:28 pm Post subject: Reply with quote
This isn't comprehensive but Scheyer, Udoh, Varnado, James Anderson, Damion James, Vasquez are among those who do quite well here. S. Collins, Alabi and Favors among those not producing here as well as their mock draft rank.
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: Tue Jan 12, 2010 10:24 am Post subject: Reply with quote
I decided to change the spreadsheet to READ-ONLY after it got messed up last night--download the spreadsheet if you want to sort/edit it.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
tawtaw
Joined: 25 Jun 2008
Posts: 28
Location: Oregon
PostPosted: Tue Jan 12, 2010 8:05 pm Post subject: Reply with quote
Great stuff DSMok. Thanks a bunch.
Back to top
View user's profile Send private message
Statman
Joined: 20 Feb 2005
Posts: 242
Location: Arlington, Texas
PostPosted: Thu Feb 25, 2010 1:19 pm Post subject: Reply with quote
Are you getting your data from the NCAA boxscores also?
I am still trying to find a RELIABLE play by play data with proper substitutions.....
_________________
Dan
My current national college player rankings (and other stuff):
http://www.pointguardu.com/f136/statman ... post355594
Back to top
View user's profile Send private message Send e-mail
DSMok1
Joined: 05 Aug 2009
Posts: 611
Location: Where the wind comes sweeping down the plains
PostPosted: Fri Feb 26, 2010 9:18 am Post subject: Reply with quote
Statman wrote:
Are you getting your data from the NCAA boxscores also?
I am still trying to find a RELIABLE play by play data with proper substitutions.....
Yes, I get the data from NCAA.org, from pages like this: http://stats.ncaa.org/team/stats?org_id ... l_id=10260. I think NCAA.org has reliable play-by-play data, though I haven't worked with it myself. As usual, all substitution data is present except who started the halves. I haven't figured out the game ID# system... An example play-by-play log: http://stats.ncaa.org/game/play_by_play/114002.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Statman
Joined: 20 Feb 2005
Posts: 242
Location: Arlington, Texas
PostPosted: Fri Feb 26, 2010 9:30 am Post subject: Reply with quote
DSMok1 wrote:
Statman wrote:
Are you getting your data from the NCAA boxscores also?
I am still trying to find a RELIABLE play by play data with proper substitutions.....
Yes, I get the data from NCAA.org, from pages like this: http://stats.ncaa.org/team/stats?org_id ... l_id=10260. I think NCAA.org has reliable play-by-play data, though I haven't worked with it myself. As usual, all substitution data is present except who started the halves. I haven't figured out the game ID# system... An example play-by-play log: http://stats.ncaa.org/game/play_by_play/114002.
At statsheet.com, he appears to be using the NCAA PbP data also. I looked just at Arizona, and 10 of the 26 games appeared to have errors (the +/- totals didn't sum correctly). The last game Arizona played WSU specifically (18 point Arizona loss) - the sub out info was so bad (especially 12 minutes left in the game) that the results were completely flawed.
I went ahead and dropped my +/- project I was doing with Arizona because I couldn't find accuarate enough PbPs.
_________________
Dan
My current national college player rankings (and other stuff):
http://www.pointguardu.com/f136/statman ... post355594
Back to top
View user's profile Send private message Send e-mail
DSMok1
Joined: 05 Aug 2009
Posts: 611
Location: Where the wind comes sweeping down the plains
PostPosted: Fri Feb 26, 2010 9:45 am Post subject: Reply with quote
Statman wrote:
At statsheet.com, he appears to be using the NCAA PbP data also. I looked just at Arizona, and 10 of the 26 games appeared to have errors (the +/- totals didn't sum correctly). The last game Arizona played WSU specifically (18 point Arizona loss) - the sub out info was so bad (especially 12 minutes left in the game) that the results were completely flawed.
I went ahead and dropped my +/- project I was doing with Arizona because I couldn't find accuarate enough PbPs.
Ken Pomeroy parsed PbP's for last year's games (see the examples here: http://www.kenpom.com/team.php?team=Arizona&y=2009), but I note that he skipped many games. I think those are the games with flawed PbP data.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
DSMok1
Joined: 05 Aug 2009
Posts: 611
Location: Where the wind comes sweeping down the plains
PostPosted: Fri Feb 26, 2010 12:51 pm Post subject: Reply with quote
Spreadsheet updated:
http://www.editgrid.com/user/dsmok1/NCA ... -10_Feb_26
Back to top
View user's profile Send private message Send e-mail Visit poster's website
tawtaw
Joined: 25 Jun 2008
Posts: 28
Location: Oregon
PostPosted: Mon Apr 05, 2010 4:46 pm Post subject: Reply with quote
Great stuff DSMok1.
Looking forward to the season final totals if you post.
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 Apr 12, 2010 11:55 am Post subject: Reply with quote
tawtaw wrote:
Great stuff DSMok1.
Looking forward to the season final totals if you post.
Final Season Totals:
http://www.editgrid.com/user/dsmok1/NCA ... 9-10_Final
The issues obvious: steals and blocks dominate to a large extent; also volume shooters skew because of the TSA^2 term.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
tawtaw
Joined: 25 Jun 2008
Posts: 28
Location: Oregon
PostPosted: Tue Apr 13, 2010 10:15 am Post subject: Reply with quote
Great stuff, thanks.
Back to top
View user's profile Send private message
tpryan
Joined: 11 Feb 2005
Posts: 100
PostPosted: Sun Apr 18, 2010 1:12 am Post subject: Reply with quote
Well, I haven't been following this thread but I see that you have some interesting numbers, such as John Wall at #49 and Derrick Favors at #126, with both expected to be lottery picks and Wall perhaps going first in the draft. As a Ga. Tech fan, I was also interested to see Gani Lawal at #561. Personally, I can't see Lawal, a junior who has declared, as an NBA player who will contribute to any degree because he has very limited range on his shot and he is a poor free throw shooter. That might be okay if he were 7-3, but he is 6-9. Favors, also of Ga. Tech, has considerably more potential, in my opinion. It would be interesting to see how these rankings correlate with future NBA achievements.
page 2 of 2
Author Message
DSMok1
Joined: 05 Aug 2009
Posts: 608
Location: Where the wind comes sweeping down the plains
PostPosted: Mon Apr 19, 2010 8:15 am Post subject: Reply with quote
tpryan wrote:
Well, I haven't been following this thread but I see that you have some interesting numbers, such as John Wall at #49 and Derrick Favors at #126, with both expected to be lottery picks and Wall perhaps going first in the draft. As a Ga. Tech fan, I was also interested to see Gani Lawal at #561. Personally, I can't see Lawal, a junior who has declared, as an NBA player who will contribute to any degree because he has very limited range on his shot and he is a poor free throw shooter. That might be okay if he were 7-3, but he is 6-9. Favors, also of Ga. Tech, has considerably more potential, in my opinion. It would be interesting to see how these rankings correlate with future NBA achievements.
They probably don't correlate well... first, they don't measure NCAA achievement too well, because the SPM weights were derived for the NBA. Second, these are retrodictive, not predictive. A 18 year old freshman with the same numbers as a 24 year old senior is a far better prospect! Also, Favors (for instance) had poor point guards, so he was rarely given the ball in position to score. Note how much better Cousins' stats were!