The Major-Conference Offensive Impact Leaders

Pretty self-explanatory. I've made a few important corrections to the formula recently.*
Must have played 40% of minutes to qualify.
Offensive rating and team SOS data from kenpom.com

	player	Conf	adjOI
1	Marcus Denmon (Missouri)	B12	11.40
2	Steven Pledger (Oklahoma)	B12	11.00
3	Henry Sims (Georgetown)	BigEast	9.41
4	Kenny Boynton (Florida)	SEC	9.27
5	Jared Sullinger (Ohio St.)	B10	9.21
6	Ricardo Ratliffe (Missouri)	B12	9.18
7	Tyrus McGee (Iowa St.)	B12	8.55
8	Travon Woodall (Pittsburgh)	BigEast	8.17
9	Cody Zeller (Indiana)	B10	8.10
10	Kevin Jones (West Virginia)	BigEast	8.01
11	Davante Gardner (Marquette)	BigEast	7.80
12	Jae Crowder (Marquette)	BigEast	7.79
13	Darius Johnson-Odom (Marquette)	BigEast	7.77
14	Mike Scott (Virginia)	ACC	7.77
15	Hollis Thompson (Georgetown)	BigEast	7.24
16	J'Covan Brown (Texas)	B12	7.22
17	Arnett Moultrie (Mississippi St.)	SEC	7.14
18	Terrell Stoglin (Maryland)	ACC	7.12
19	Dion Waiters (Syracuse)	BigEast	7.00
20	Pierre Jackson (Baylor)	B12	7.00
21	JaMychal Green (Alabama)	SEC	6.97
22	Doron Lamb (Kentucky)	SEC	6.97
23	Jeremy Lamb (Connecticut)	BigEast	6.76
24	Erick Green (Virginia Tech)	ACC	6.64
25	Erving Walker (Florida)	SEC	6.63
26	Tyler Zeller (North Carolina)	ACC	6.57
27	Robbie Hummel (Purdue)	B10	6.38
28	Andre Young (Clemson)	ACC	6.14
29	Shabazz Napier (Connecticut)	BigEast	5.99
30	Brandon Young (DePaul)	BigEast	5.98
31	Mike Rosario (Florida)	SEC	5.96
32	Nasir Robinson (Pittsburgh)	BigEast	5.77
33	John Jenkins (Vanderbilt)	SEC	5.73
34	Drew Crawford (Northwestern)	B10	5.71
35	Thomas Robinson (Kansas)	B12	5.70
36	Ashton Gibbs (Pittsburgh)	BigEast	5.61
37	Maalik Wayns (Villanova)	BigEast	5.53
38	Trae Golden (Tennessee)	SEC	5.47
39	Ryan Kelly (Duke)	ACC	5.45
40	Kris Joseph (Syracuse)	BigEast	5.41
41	Deshaun Thomas (Ohio St.)	B10	5.39
42	C.J. Harris (Wake Forest)	ACC	5.37
43	B.J. Young (Arkansas)	SEC	5.22
44	Scott Wood (North Carolina St.)	ACC	5.11
45	Jeronne Maymon (Tennessee)	SEC	5.10
46	Reggie Bullock (North Carolina)	ACC	5.07
47	Trevor Mbakwe (Minnesota)	B10	5.04
48	John Shurna (Northwestern)	B10	4.99
49	Victor Oladipo (Indiana)	B10	4.96
50	Mouphtaou Yarou (Villanova)	BigEast	4.95

* I adjusted the overall formula slightly** to more evenly adjust for competition, and to more accurately portray players (R^2 is about 5% better than my prior formula)

**
adjDsos = (% of games against D-I * Adj Def.SOS) + (% games against non D-I * 114)
PointsProducedPer100=ORTG*PercentPoss/100

adjOI = (PointsProducedPer100)/(100.3*0.2)*n1 + PercentPoss/20*n2 + ORTG/100.3*n3 + 0.2*(100.3-adjD.SOS)

where:
n1= 26.97

n2= -22.38

n3= -4.73

Indiana great. Gonzaga good. Cal mediocre.

This is my original method for rating teams after which my blog is named, although I've never done league-wide rankings.The normal distribution states that each team's chance of winning is dependent on their strength (efficiency or point differential) and the consistency of that strength (the standard deviation of efficiency differential).

I adjust standard deviation for competition by simply using Actual minus Expected point margin.*


Let's look at the ten teams who gain the most in terms of win% when we adjust for consistency.

	team	win% diff	pyth rnk (adj rnk)	adj win%
1	Gonzaga	8.7%	34 (20)	93.5%
2	Murray St.	8.7%	50 (31)	89.1%
3	Temple	8.6%	36 (21)	92.7%
4	Creighton	8.5%	55 (33)	88.2%
5	Indiana	8.1%	14 (5)	98.2%
6	Miami FL	7.7%	63 (55)	83.9%
7	Lehigh	7.6%	88 (67)	77.9%
8	Tulsa	7.3%	95 (75)	74.6%
9	Connecticut	7.2%	23 (12)	95.4%
10	Long Beach St.	7.0%	47 (35)	87.8%

Gonzaga's actual versus expected point margin only has a standard deviation of 6.1 - this is significantly low.

Next, the ten teams in Pomeroy's top-100++ who drop the most.
(++The list was mostly bottom-feeders anyways, so we'll just look at teams we care more about).

		win% diff	pyth rnk (adj rnk)	adj win%
1	California	-2.7%	20 (44)	86.4%
2	Purdue	-2.7%	15 (40)	87.0%
3	Virginia	-0.4%	24 (36)	87.8%
4	Florida St.	-0.1%	32 (48)	85.8%
5	Baylor	0.7%	13 (25)	90.9%
6	West Virginia	1.2%	33 (46)	86.3%
7	St. Louis	1.2%	22 (30)	89.6%
8	Missouri	1.2%	8 (17)	94.0%
9	Marquette	1.5%	6 (13)	94.7%
10	New Mexico	1.8%	37 (50)	85.3%

Cal plays very strongly on average, but has an adjusted standard deviation of 15.7. Quite inconsistent.


And here are the top 25 overall:

	team	adj win%	pyth	diff
1	Kentucky	0.9913	0.9686	0.0227
2	Ohio St.	0.9907	0.9647	0.0259
3	Syracuse	0.9890	0.9511	0.0379
4	North Carolina	0.9885	0.9519	0.0366
5	Indiana	0.9817	0.9011	0.0806
6	Kansas	0.9813	0.9310	0.0503
7	Wisconsin	0.9812	0.9733	0.0080
8	Duke	0.9789	0.9264	0.0525
9	Michigan St.	0.9740	0.9155	0.0586
10	Florida	0.9710	0.9158	0.0552
11	Louisville	0.9690	0.9237	0.0452
12	Connecticut	0.9539	0.8816	0.0723
13	Marquette	0.9471	0.9321	0.0149
14	Brigham Young	0.9430	0.8964	0.0467
15	Xavier	0.9406	0.8957	0.0448
16	Georgetown	0.9400	0.8933	0.0467
17	Missouri	0.9397	0.9272	0.0124
18	Wichita St.	0.9359	0.8794	0.0565
19	Stanford	0.9355	0.8967	0.0388
20	Gonzaga	0.9353	0.8485	0.0868
21	Temple	0.9269	0.8408	0.0861
22	Texas	0.9234	0.8812	0.0422
23	Nevada Las Vegas	0.9124	0.8742	0.0382
24	Alabama	0.9119	0.8851	0.0268
25	Baylor	0.9086	0.9012	0.0074

 
Yes, yes. The Hoosiers are doing quite well....





*- While explaining individual parts of a team or player in "tempo-free" language is beneficial, describing a team by "efficiency margin" rather than "point margin" is not particularly beneficial. In the NBA, point margin usually predicts win-percentage only slightly worse than efficiency margin (although last year, point margin did better). However, for numerous statistical and tactical reasons, pace is important to consider, in my opinion, when rating teams. It is much easier for a team to put away opponents if they are more efficient at a much quicker pace; statistically this makes sense: the more samples we have, the better chance the final output is near our expected output. Furthermore, I don't have access to game-by-game efficiency margin, so the only way I can measure standard deviations/variance/consistency is by measuring Pomeroy-Ratings's expected point margin versus the actual game point margin.

A Terrible Blog Post About Murray St.

Data from Statsheet.com and Kenpom.com


Don't have time to write more at the moment.
This was made using the formula from Holy Grail part II, except I took out the plus-minus component (not good enough data). This means that these rankings formula is biased towards offense (+/- improves the defensive component heavily).

Also, I fit the "impact per game" to Murray State's expected efficiency margin against an average team (13.2, according to Kenpom.com).

Short, but sweet. Tells us what we already know, I suppose:

1) Canaan rules on offense.
2) The Racers are pretty dern good at stealing the ball (#39 in Stl% as a team).

Top 10 Offensive Impact Players, as of 12/2

I've updated the formula slightly* since I wrote the essay for the College Basketball Prospectus preview this year. Offensive Ratings, usage, and strength of schedule all from kenpom.com.

Offensive impact estimates how much a player's contributions (by usage & efficiency) impact his team's offense above average (i.e. a +13 player would make an average team score 1.13 PPP, if the league average was 1 PPP).

	Player	Team	Year	ORTG	%USG	adjusted Offensive Impact
1	Marcus Denmon	Missouri	Sr	143.3	24.2	13.0
2	Damian Lillard	Weber St.	Jr	125.7	31.7	12.8
3	Steven Pledger	Oklahoma	Jr	146.5	21.7	11.9
4	Darius Johnson-Odom	Marquette	Sr	134.7	25.1	11.7
5	Ricardo Ratliffe	Missouri	Sr	142.9	22.5	11.6
6	Doug McDermott	Creighton	So	127.2	28.7	11.6
7	Anthony Drmic	Boise St.	Fr	135.4	24.2	11.0
8	Trae Golden	Tennessee	So	122.6	30.5	11.0
9	Tahj Tate	Delaware St.	Fr	120.8	31.2	10.9
10	Justin Edwards	Maine	Fr	130.2	25.9	10.8

I expect more Major-Conference heavy hitters to top this list soon. We've all taken the oath of small sample sizes here, right?

I don't feel comfortable sharing more than the top-10 because many players with Top-50 Offensive Impact numbers are cut out by not having high enough Offensive Ratings (for example: Jordan Taylor, at least at the moment) to be grouped on Pomeroy's site. But I don't have the time to input every NCAA player manually.



*Math:
The coefficients for ORTG, Usage, and ORTG*Usage are still the same, but I used a more theory-based adjustment for strength of schedule. Also I added a term that helps predict team ORTG better (the numbers were too small so each player gets boosted by 1ish).

Why North Carolina Will Win The National Championship

Because John Henson is magic.

"Whhhhhhhhh"
"OKAY! YOU CAN HAVE IT!"