Writers of Pro Football Prospectus 2008

19 Oct 2016

FEI Week 7: Game Ratings

by Brian Fremeau

We have finally reached the point in the season when all projection data has been eliminated from the FEI ratings formula and every team stands only on the merits of what it has accomplished this year. Some of those programs are in nearly the same spot that our preseason FEI ratings had forecasted. A total of 44 teams are ranked within 10 spots of their preseason position, including current No. 1 Alabama, No. 2 Clemson, No. 5 Ohio State, No. 6 Washington, No. 7 Louisville, and No. 10 Mississippi. A few preseason FEI favorites have fallen off significantly, however, including current No. 36 Oklahoma, No. 50 Stanford, No. 67 Georgia, and No. 69 Oregon. Current FEI top-10 teams Washington State and Western Michigan were projected into the top 50, but have significantly exceeded expectations thus far.

There will of course be some shifting from this point until the end of the year, but the current distribution of team efficiency ratings has a bit more in common with what the final ratings will be than what the preseason ratings projected. Last season, the correlation between preseason FEI and Week 7 FEI was .757, and the correlation between Week 7 FEI and the final FEI ratings was .916. This year, the preseason-to-Week 7 FEI correlation stands at .723.

This is also the point in the season when I debut the offense, defense, and special teams efficiency ratings to round out each team's FEI profile. To date, I have only highlighted raw efficiency data and statistics that have shaped the game results of the first half of the season, but we now have opponent-adjusted data points that amplify how those results produce the overall FEI ratings.

The guts of each team's FEI rating are its GFEI data, the opponent-adjusted single-game efficiency ratings. Every game result is adjusted for the both the strength of the opposition faced and for home field advantage to produce GFEI. Michigan's victory over Hawaii back on the opening weekend of the year ranks as the most efficiently dominant performance to date in terms of raw, unadjusted results -- a 63-3 final score that reached garbage time after only 14 possessions with the Wolverines leading 42-0. But that result was at home for Michigan against the current 113th-ranked Warriors. It's still a very strong performance, but the home field and opponent adjustments rate it in the 94th percentile in GFEI -- better than 94 percent of all single-game performances this year, but not the absolute best.

The absolute best single-game performance of the season thus far came in Week 3, when Louisville throttled Florida State by a final score of 63-20. Garbage time kicked in according to FEI after 23 possessions, with the Cardinals ahead by a score of 56-10. The raw game efficiency for Louisville in that game ranks in the 96th percentile, but adjusting for a performance that came against a good (and nearly very good) team in No. 18 Florida State vaults the GFEI rating to the 99th percentile. Alabama's victory over Tennessee on the road last weekend (35-10 in 24 non-garbage possessions) ranks No. 2 in GFEI, and Washington's dominant win over Stanford in Week 5 (30-0 in 13 non-garbage possessions) ranks third overall.

The question I receive most commonly on Twitter is some variation on "why is Team X rated so high and why is Team Y rated so low?" GFEI data is presented in part to help decipher the ratings and answer those questions. The three teams with the best tally of GFEI results ranked in the 90th percentile -- Alabama (six games in the 90th percentile), Michigan (five), and Clemson (four) -- all rank among the top three of the FEI ratings. The top five teams in FEI each have at least four single-game GFEI performances ranked in the 80th percentile. Eight of the top ten teams in FEI have at least four single-game GFEI performances ranked in the 70th percentile.

The difference between teams that get more love in the polls than in the FEI ratings can also explained by GFEI results. Oklahoma ranks 16th in the Associated Press top 25 poll, but the Sooners have had only two games ranked in the 70th percentile or better according to GFEI -- wins over TCU and Kansas State. Utah (No. 19 in the AP poll, No. 38 in FEI) and Arkansas (No. 17 in the AP poll, No. 39 in FEI) are two other examples of teams with a similar FEI profile that are situated around 20 positions higher in the polls. The Sooners, Utes, and Razorbacks each have the same number of 70th percentile (two) and sub-70th percentile performances (four) as, for example, Iowa and Wake Forest. Those teams are receiving very little attention in the polls but are ranked in the same neighborhood as Oklahoma, Utah, and Arkansas in the FEI ratings.

GFEI ratings for all teams can be found here.

Elite teams rack up a combination of elite raw efficiency results and elite opponent-adjusted efficiency results, and the 90th percentile is generally how the best teams in the country are identified according to FEI. Alabama, Clemson, and Michigan have had the best start, but will they finish strong as well? Are there any teams that haven't yet racked up 90th percentile performances but will do so in the second half of the season? We'll find out soon enough.

FEI Ratings Through Week 7

The Fremeau Efficiency Index (FEI) is a college football rating system based on opponent-adjusted drive efficiency. Approximately 20,000 possessions are contested annually in FBS vs. FBS games. First-half clock-kills and end-of-game garbage drives and scores are filtered out. Unadjusted game efficiency (GE) is a measure of net success on non-garbage possessions, and opponent adjustments are calculated with special emphasis placed on quality performances against good teams, win or lose. Overall SOS ratings represent the likelihood than an elite team (two standard deviations better than average) would go undefeated against the given team's entire schedule.

Offensive FEI (OFEI) is value generated per offensive non-garbage possession adjusted for the strength of opponent defenses faced. Defensive FEI (DFEI) is value generated per opponent offensive non-garbage possession adjusted for the strength of opponent offenses faced. Special Teams Efficiency (STE) is the average value generated per non-garbage possession by a team's non-offensive and non-defensive units.

Strength of schedule ratings for games played to date (SOP) and for scheduled games remaining to be played (SOR), along with the projected number of FBS wins remaining against scheduled opponents (MWR) are also provided.

Rk Team Rec FEI GE Rk SOS Rk OFEI Rk DFEI Rk STE Rk SOP Rk SOR Rk MWR
1 Alabama 7-0 .335 .323 4 .090 22 .63 20 1.71 1 .05 37 .308 30 .291 23 3.5
2 Clemson 6-0 .331 .138 20 .166 41 .62 21 1.21 3 .01 58 .292 27 .569 60 4.7
3 Michigan 6-0 .288 .369 2 .162 38 1.29 3 1.16 4 .23 1 .636 77 .255 18 5.2
4 Texas A&M 5-0 .285 .111 26 .057 10 .60 23 .35 36 .05 34 .425 37 .134 5 4.9
5 Ohio State 6-0 .282 .354 3 .148 35 1.12 8 1.02 7 .07 24 .544 59 .272 20 5.1
6 Washington 5-0 .253 .449 1 .335 68 1.23 6 1.06 6 .20 4 .893 122 .375 34 5.1
7 Louisville 5-1 .236 .323 5 .100 24 1.35 2 .60 18 .00 64 .182 9 .549 56 5.3
8 Washington State 4-1 .219 .214 10 .229 51 .90 12 .84 11 .01 57 .547 60 .418 38 4.9
9 Western Michigan 6-0 .217 .287 6 .613 119 1.43 1 .43 25 -.01 71 .690 90 .888 110 4.8
10 Mississippi 2-3 .207 .099 30 .040 4 1.15 7 .28 39 -.04 88 .229 18 .175 11 4.3
11 Auburn 4-2 .196 .205 13 .019 1 .83 15 .85 10 .10 15 .168 7 .116 4 3.2
12 West Virginia 4-0 .189 .130 24 .505 101 .41 32 .71 15 -.06 98 .788 108 .641 72 6.0
13 Nebraska 6-0 .183 .184 16 .109 28 .42 31 .37 32 .10 14 .646 79 .169 10 3.9
14 Miami 3-2 .177 .137 21 .297 65 .07 58 .73 13 .03 45 .613 73 .484 46 4.5
15 Boise State 6-0 .175 .244 8 .493 99 .99 10 .68 16 -.04 90 .682 85 .723 89 5.3
Rk Team Rec FEI GE Rk SOS Rk OFEI Rk DFEI Rk STE Rk SOP Rk SOR Rk MWR
16 Wisconsin 4-2 .173 .102 28 .056 9 .29 41 1.26 2 .03 42 .132 4 .424 39 4.4
17 LSU 3-2 .166 .137 22 .034 3 .49 28 1.09 5 -.17 124 .454 46 .075 2 2.2
18 Florida State 4-2 .160 .003 64 .041 5 .79 17 -.03 65 -.04 92 .163 6 .254 17 3.2
19 Houston 5-1 .159 .220 9 .417 83 .75 18 .39 31 .04 41 .767 103 .544 55 3.7
20 Northwestern 3-2 .155 .066 37 .102 25 .50 27 .45 22 .05 36 .450 45 .227 15 4.1
21 Tennessee 5-2 .153 .025 53 .089 21 .45 30 .32 37 .09 20 .095 2 .941 122 3.7
22 North Carolina State 3-2 .151 .106 27 .057 11 .55 25 .36 35 -.01 74 .208 14 .274 21 3.7
23 Colorado 4-2 .146 .130 23 .140 33 .49 29 .85 9 -.21 127 .255 22 .550 57 3.4
24 Baylor 5-0 .133 .280 7 .415 81 .81 16 .19 49 -.06 99 .925 126 .448 43 3.9
25 Iowa 5-1 .116 .175 17 .200 47 .39 34 .23 44 .09 16 .677 83 .295 24 2.4
26 UCLA 3-4 .114 .039 46 .077 16 -.20 81 .72 14 -.05 94 .137 5 .564 58 3.1
27 Florida 5-1 .111 .211 12 .278 62 .04 61 .95 8 .00 68 .679 84 .409 36 2.7
28 BYU 4-3 .111 .035 50 .323 67 .22 48 .18 50 .07 27 .491 51 .657 76 3.1
29 South Florida 5-1 .107 .201 14 .557 109 .87 14 .10 58 .09 19 .791 110 .704 85 3.4
30 Minnesota 3-2 .104 .056 39 .256 55 .02 62 .46 21 .01 59 .689 89 .371 33 3.7

Posted by: Brian Fremeau on 19 Oct 2016

2 comments, Last at 20 Oct 2016, 9:49am by Brian Fremeau

Comments

1
by Chappy :: Wed, 10/19/2016 - 4:11pm

"The difference between teams that get more love in the polls than in the FEI ratings can also explained by GFEI results."

Seems to me this sentence is off. The rest of the paragraph is about how teams that are different in the polls have nearly the same FEI result and GFEI percentiles, but are disparate in the polls.

2
by Brian Fremeau :: Thu, 10/20/2016 - 9:49am

You're right, that sentence is awkward. It was meant to call out the "why isn't Oklahoma/Utah/Arkansas ranked higher by FEI?" kind of questions.