OU-LA. Monroe Statistical Plus/Minus Review

[DSMok1]

OU vs. Louisiana-Monroe Statistical Plus/Minus

OU did not play particularly well this game--the winning margin was just 11 points against a fairly weak opponent. Who played well? This game, Crocker and Willie led the way for OU, but two players from LA-Monroe actually had the best games!

(Boxscore)
OU Statistical +/-

Player			SPM	Min	Contribution
Crocker, Tony		7.5	34	6.4
Warren, Willie		6.4	33	5.3
Wright, Ryan		1.7	22	1.0
Pledger, Steven		2.0	16	0.8
Davis, Cade		-1.0	28	-0.7
Gallon, Tiny		-3.0	27	-2.0
Mason-Griffin, Tommy	-3.2	32	-2.6
Fitzgerald, Andrew	-13.0	8	-2.6

ULM Statistical +/-

Player			SPM	Min	Contribution
Hooper, Tony		10.0	30	7.5
Forbes, Dynile		12.1	23	7.0
Gilbert, Lawrence	1.1	22	0.6
Burt, Rory		2.4	10	0.6
Sykes, Tommie		0.0	18	0.0
Carr, Colby		-6.7	5	-0.8
Turner, Rudy		-1.4	29	-1.0
Hill, Jarvis		-10.4	16	-4.2
Fuselier, Warren	-16.7	18	-7.5
Thomas, Malcolm		-10.6	29	-7.7

Crocker and Warren played well. Crocker played a clean, efficient game: no turnovers, no fouls, 13 points on 6 shots, 3 offensive rebounds, and 1 steal. Warren got 24 points on just 12 shots, but had only 2 rebounds and had 4 turnovers to balance his play out a bit. Ryan Wright played a good, solid game--making a few shots, getting a few rebounds, making a block. Pledger again played solidly, scoring 8 points, grabbing an offensive rebound and getting a steal in just 16 minutes. The rest of the team played mediocre at best--Gallon got rebounds, but 9 of them were defensive rebounds which are a lot easier to come by. Fitzgerald didn't do a whole lot well--no points or rebounds; just a block.

On the other side, Tony Hooper played an excellent game against superior opposition. Seventeen points on 13 shots; 5 rebounds, 2 assists, 4 turnovers (but 2 steals). Forbes had an efficient game as well, scoring nicely and playing well on the defensive end (3 steals in only 23 minutes!)

All in all, this was a weak outing for OU. The post men failed to do well; the guards again had to carry the team. At least Crocker stepped up and played a clean, senior-like game.

Statistical Plus/Minus (SPM) is a method of estimating each player's impact from the box score statistics. SPM is listed in points above the average player playing per 40 minutes--so if that player was replaced by an average player for 40 minutes, SPM is the difference in the final margin. The total of all player's contributions will sum to the actual scoring margin (each team's total will equal half of the overall margin). The original method was outlined by Dan Rosenbaum at 82games.com; recently additional factors were added by Neil Paine at Basketball-Reference.com. I previously compiled the complete 2008-2009 NCAA numbers on this forum.​

 

[BoulderSooner]

thanks for the post .. this is good info

 

[tomkar98]

I am becoming more and more of a believer in this +/- stuff. I'm big on statistics, so its cool to have an understanding of this stuff, now.

 

[DSMok1]

I am becoming more and more of a believer in this +/- stuff. I'm big on statistics, so its cool to have an understanding of this stuff, now.

I like adjusted plus/minus, as it is the only statistic that DIRECTLY measures a player's impact. That said, it suffers from difficulty in calculation--particularly if two players usually come into and exit the game at the same time. Thus there can be a very large margin for error (see the 1 year (~12 game) APM for the NBA this year at Basketball Value)

We can't do adjusted plus/minus for college because we would have to compile play-by-play data for all games played (impossible, except ESPN perhaps could pull it off). The matrix would be massive, with ~3000 players to estimate for. Statistical Plus/Minus estimates APM from the box score data--but it is imprecise, particularly with defense (the few defensive stats don't measure defensive impact very well). It's about the best we can do, however, for NCAA games.

 

[MNsooner]

Yeah, I am a fan of it in someways. The problem is that players +/- gets effected by whether they play on the first or second unit.

 

[OULad]

I am a numbers guy, and I like this stuff but it can be a bit misleading. I thought Cade played an excellent game, but it obviously didnt show in the box score.

 

[DSMok1]

Yeah, I am a fan of it in someways. The problem is that players +/- gets effected by whether they play on the first or second unit.

Adjusted Plus/Minus compensates for that completely. Statistical Plus/Minus does approximately.

APM accounts for each player on the court (both offense and defense), and solves the equations of who is actually contributing.

 

[DSMok1]

I am a numbers guy, and I like this stuff but it can be a bit misleading. I thought Cade played an excellent game, but it obviously didnt show in the box score.

Yes, the box score can only capture so much.

 

[MNsooner]

How does the adjusted plus minus work if you don't mind explaining?

 

[DSMok1]

How does the adjusted plus minus work if you don't mind explaining?

Hang on!

Adjusted Plus/Minus

Suppose you have two lineups face each other, for perhaps 5 possessions each way. Team A outscores Team B 12 to 7. An equation is created for this "session":

(12-7)
------ = TMA1 + TMA2 + TMA3 + TMA4 + TMA5 - TMB1 - TMB2 - TMB3 - TMB4 - TMB5
  5

(I'm simplifying things a bit, but you get the idea.) Basically, we take this session as a sum of all the players for one team minus all of the players for another. In this case, Team A's "margin" was 1 point per possession. Then, some players are substituted. Over 3 possessions, Team A scores 5 points and Team B scores 6 points. Another equation is created:

(5-6)
------ = TMA6 + TMA7 + TMA3 + TMA4 + TMA5 - TMB6 - TMB2 - TMB3 - TMB4 - TMB5
  3

Over the course of the season, thousands of equations are created. A computer can solve this system of equations, minimizing the error; the equations are weighted according to how many possessions were in each session. The result is a value for each player, relative to all other players--basically, the player's value in points-per-possession above average. To make the numbers easier to read, the results are shown in points per 100 possessions.

For instance, here are some results for the NBA last year:

					2 Year		1 Year	
					  Adj.		  Adj.
Teams 	Player 	Min 	  		  +/- 	SE 	  +/- 	SE 
MIA	Wade, Dwayne	3,048.32	 11.17	3.31	 22.15	5.16
CLE	James, LeBron	3,054.18	 14.87	2.96	 16.92	5.24
LAL	Odom, Lamar	2,203.02	 7.84	2.82	 16.64	4.43
DAL	Kidd, Jason	2,814.47	 7.9	2.73	 15.23	5.49
NOH	Paul, Chris	2,888.32	 11.69	4.02	 15.06	6.3
BOS	Allen, Ray	2,691.23	 5.36	3.15	 13.91	5.57
PHI	Iguodala, Andre	3,269.05	 7.78	3.48	 12.14	5.01
ORL	Lewis, Rashard	2,858.98	 6.82	3.09	 10.19	4.82

The problem is that sometimes 1 or even 2 years isn't enough to get a good grasp; the solver can sometimes have problems differentiating players that play together a lot. And players do change over the years.

A full discussion of Adjusted Plus/Minus, Statistical Plus/Minus, and ways of minimizing the errors is available at 82games.com. On the Association of Pro Basketball Research message boards, several better methods have been proposed and used to create less error in the estimate: 1-year stabilized estimates, 6-year average estimates, and others. Steve Ilardi created many of these new models; he's got articles at 82 games.com also.

 

[SoonerfanTU]

I'm not a huge numbers guy, because I typically think most stats only show one part of the overall analysis needed to actually get a good idea of what is happening. That, and they can be influenced.

That said, Adjusted Plus/Minus is a pretty good stat.