What is a running back worth? A better way to evaluate RBs

One hallmark of the sabermetric movement is the search for statistics that better reflect meaningful production. For example, rushing yards are positively correlated with winning, but they do not cause a team to win. Touchdowns do directly cause a team to win, but an individual player's touchdown count does not accurately reflect their meaningful contribution to that win.

Along these lines, I have calculated a point impact estimate on running plays for each non-QB college football player in 2011. Essentially, a point value can be attached to any game situation based on the probability that a team scores from that position. For example, 1st and goal at the 1 would be given a high point value and 4th and 20 at your own 20 would have a very low point value. The point impact of any play is that point value of the situation after the play minus the point value before the play.

The real challenge of this approach is accurately valuating the situation on the field. I have used two approaches - the first (PIA) a more technical approach using multinomial logistics, and the second (PIB) a more direct approach using Bayesian adjustments. The results from the two approaches are highly correlated but not identical. Averaging these results together I get the PI (point impact). (The data set I use for these calculations is not without gaps, so I have reported the number of carries (C) on which these results are based.)

Number 1 on the list by a good margin is Montee Ball. Leading the nation in rushing yards and the whole of humanity in rushing touchdowns, it would only make sense that the Badger would finish on top. Oregon's James and Alabama's Richardson round out the top 3. The first major surprise is Georgia Tech's Orwin Smith at number 4. That someone named Orwin could be one of the most productive backs in the nation in the 21st century is shocking enough, but this is compounded by the fact that he did it in only 61 carries. He did it with 95 and 77 yard touchdowns, a touchdown on a fourth down conversion, and even 18 yards on a 3rd and 17. This brings us to the overall theme of these results, which is that point impact is built on big plays and high yards per attempt, not by grinding it out.

Oregon, Navy, Georgia Tech, and Utah St. managed to land two players in the top 25 while 83% of teams are absent from the list. A number of quarterbacks would have made this list just from their rushing yards: C. Harnish (51.5), C. Klein (50.5), D. Robinson (42.1), B. Miller (34.0), C. Shaw, CJ Brown, J. Franklin, G MarQueis, and Logan Thomas. In 2010, the best 5 seasons by point impact on running players were all quarterbacks. Third among running backs in 2010 was Montee Ball, a surprising result since he was effectively the #3 back on his own team.

Point Impact Estimate
	Name	Team	PI	PIA	PIB	YPC	C
1	Ball, Montee	Wisconsin	88.7	89.6	87.9	6.4	290
2	James, LaMichael	Oregon	73.9	77.1	70.7	7.4	246
3	Richardson, Trent	Alabama	54.3	53.9	54.7	6	283
4	Smith, Orwin	Georgia Tech	50.2	52.3	48.1	10.3	61
5	Ganaway, Terrance	Baylor	48	42.3	53.8	5.9	240
6	Pierce, Bernard	Temple	43.3	42.9	43.7	5.6	272
7	James, Waymon	TCU	41.9	44.1	39.8	7.7	121
8	Randle, Joseph	Oklahoma St.	41.3	41.4	41.2	6	208
9	Josey, Henry	Missouri	39.1	42.5	35.8	8.1	145
10	Maples, Raymond	Army	35.6	38.9	32.4	7.4	146
11	Clark, Asher	Air Force	35.4	35.3	35.5	7.3	163
12	Gray, Jonas	Notre Dame	36.1	39.6	32.6	6.9	114
13	Stacy, Zac	Vanderbilt	30.8	27.7	33.9	6.2	201
14	Mark, Lampford	Nevada	30.3	27	33.7	5.6	159
15	Turbin, Robert	Utah St.	30.2	34.7	25.8	6.2	249
16	Thomas, Adonis	Toledo	33	29	37	6.5	171
17	Smith, Michael	Utah St.	28	26	29.9	7	114
18	Sims, Charles	Houston	26.8	26.2	27.3	7.5	110
19	Stephens, Eric	Texas Tech	25.4	25.1	25.6	5.2	108
20	Howell, John	Navy	25.3	26.5	24.1	9.7	35
21	Blue, Alfred	LSU	25.3	24.8	25.8	6.9	78
22	Douglas, Ja'terian	Tulsa	24.8	20.6	28.9	8.2	112
23	Barner, Kenjon	Oregon	24.5	28.3	20.7	6.3	152
24	Peeples, Embry	Georgia Tech	23.6	22.9	24.3	11.1	47
25	Greene, Gee Gee	Navy	23.4	24.1	22.6	8	64