Writers of Pro Football Prospectus 2008

Most Recent FO Features


» Weight and Injuries

NFL football is a violent game, and traumatic injuries are unfortunate but unavoidable. But are bigger players more likely to be hurt than their smaller peers?

12 Oct 2016

FEI Week 6: Game Splits

by Brian Fremeau

Due in part to treacherous weather that bombarded the southeastern region of the country, it was a strange weekend in college football. Two games were rescheduled to Sunday, and another -- (and a potentially very important one that that) -- has been indefinitely postponed. Point spread favorites won only 61 percent of the FBS games that were actually played. That was the worst record by point spread favorites in the regular season since Week 6 in 2007, a week that featured Stanford's epic win over USC as a 39-point dog.

We didn't have any upsets of that magnitude this weekend, but there were a half-dozen wins by teams that were double-digit underdogs. The most significant result of that group was Navy's 46-40 victory over Houston, a win that knocked the Cougars out of the running for a potential College Football Playoff bid, and severely jeopardized their potential to even have an opportunity to play in a major bowl. Navy is now in the driver's seat for the American Athletic West division crown, despite having a negative game efficiency and negative net points per drive margin on the season.

A few results weren't upsets, but were still stunning in the other direction. Washington laid waste to Oregon in Autzen Stadium by a final score of 70-21. The Huskies are the nation's top team in game efficiency through the first six weeks and rank fifth overall in FEI -- their crushing defeats of recent Pac-12 North division stalwarts Stanford and Oregon in consecutive weeks were so thorough, both teams are now well outside the FEI top-30. As a result, Washington is the only FEI top-15 team that hasn't played an FEI top-30 opponent yet this year.

One spot ahead of Washington this week are the Michigan Wolverines, who had a spectacularly dominant beatdown of their own. Michigan defeated Rutgers by a final score of 78-0, allowed only 5 yards passing and two first downs, forced 16 punts, and outgained the Scarlet Knights by a total yardage margin of 600 to 39. The top three teams in net points per drive through Week 6 -- Washington, Ohio State, and Michigan -- all have one common opponent to date: Rutgers.

Michigan's defensive domination was thorough, but the most valuable defensive effort of the weekend came out of one of the games impacted by weather -- at least, weather seemed to be a factor that impacted one of the two teams involved. Virginia Tech defeated North Carolina on the road by a final score of 34-3, vaulting to the top of the ACC Coastal division standings and looking every bit like a conference contender improving week by week. The total yardage margin in the victory wasn't eye-popping (264-131), but the way in which the Hokies generated value from its defense was noteworthy.

I break down the non-garbage scoring margins of each game to attribute value gained and lost to the offense, defense, and special teams units that produced that result. Every possession is worth an average of 2.1 points per drive to the team with the ball. If the offense produces a score, it has earned some value above that average; if the offense fails, it has earned some value below that average. The values above and below average are calculated based on field position.

Let's look at North Carolina's first few possessions of the game. The Tar Heels first drive was a turnover on downs after starting with favorable field position. Their second drive went three plays, lost 5 yards, and resulted in an interception that put the Hokies in the red zone on the ensuing possession. North Carolina's third drive went 14 yards in two plays and ended in a fumble that put Virginia Tech in plus territory yet again. The Hokies forced another three-and-out on the next Carolina drive. The Hokies were held to a field goal on each of their short field drives and led 6-0 through the first 12 minutes and eight possessions of the game.

There was a six-point scoring advantage in the game to that point, but my analysis credits Virginia Tech's defense with an even larger margin -- plus-10.9 points of scoring value generated on those first four North Carolina possessions. The Virginia Tech offense should have scored more based on its favorable starting field position after turnovers, and the offense actually cost the Hokies minus-6.0 points in scoring value. The Hokies special teams unit added plus-1.1 points of value on the field goal kicks. The six-point scoring margin was the sum of those values -- +10.9 (defense), -6.0 (offense), +1.1 (special teams) = 6.0 margin.

The Virginia Tech defense didn't relent over the course of the rest of the game either. The Hokies' average starting field position over their next 10 non-garbage drives was their own 44-yard line due to dominant defensive play. Over the non-garbage possessions for the entire game (13 for Virginia Tech), their average starting field position was nearly at midfield. (They tacked on a late touchdown on yet another short field in garbage time). In a game in which the non-garbage final scoring margin was 24 points (27-3 in 27 possessions), the Hokies defense earned 35.6 points in scoring value. That's the highest defensive scoring value generated in any game this year.

Game splits for every team in every FBS game played to date can be found here. Beginning next week, I'll have fully eliminated preseason projection data from the formula, and I'll begin posting offensive, defensive, and special teams efficiency ratings along with the overall FEI ratings. I'll also begin posting opponent-adjusted GFEI game ratings next week.

FEI Ratings Through Week 6

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.

FEI ratings through the first six weeks of the season are based in part on preseason projection data. Preseason ratings (PRE) represent approximately 14 percent of this week's ratings. Strength of schedule ratings for games played to date (SOP) are also provided. Ratings for all teams are linked here.

Rk Team Rec FEI GE Rk SOS Rk PRE Rk SOP Rk
1 Clemson 5-0 .332 .147 24 .206 43 .235 3 .370 26
2 Alabama 6-0 .327 .328 5 .073 11 .290 1 .380 27
3 Texas A&M 5-0 .273 .112 29 .047 5 .111 24 .388 28
4 Michigan 6-0 .270 .370 3 .248 53 .123 22 .691 72
5 Washington 5-0 .248 .450 1 .377 77 .144 15 .814 102
6 Mississippi 2-2 .239 .136 27 .045 4 .182 8 .265 15
7 Ohio State 5-0 .237 .413 2 .173 34 .154 12 .804 96
8 Louisville 4-1 .232 .346 4 .110 23 .137 17 .177 5
9 West Virginia 3-0 .190 .071 35 .424 83 .056 41 .738 84
10 Auburn 4-2 .188 .206 13 .017 1 .103 28 .169 4
11 Tennessee 5-1 .187 .071 36 .095 16 .161 10 .250 14
12 Florida State 3-2 .186 -.022 75 .051 7 .160 11 .163 2
13 LSU 2-2 .185 .076 34 .036 3 .236 2 .472 36
14 UCLA 3-3 .179 .057 40 .105 20 .133 20 .235 12
15 Miami 3-1 .177 .195 18 .343 69 .038 48 .644 65
Rk Team Rec FEI GE Rk SOS Rk PRE Rk SOP Rk
16 Nebraska 5-0 .173 .220 12 .209 45 .134 19 .811 100
17 Wisconsin 4-1 .172 .143 26 .099 17 .058 40 .248 13
18 Western Michigan 5-0 .170 .250 9 .712 121 .028 59 .815 104
19 Boise State 5-0 .163 .286 6 .555 109 .109 25 .807 97
20 Washington State 3-1 .162 .255 7 .230 50 .059 38 .600 57
21 Houston 4-1 .160 .254 8 .423 82 .104 26 .761 88
22 BYU 3-3 .144 .028 55 .263 56 .104 27 .388 29
23 Virginia Tech 3-1 .136 .199 15 .363 74 .084 33 .630 63
24 Baylor 4-0 .130 .201 14 .376 76 .181 9 .936 126
25 Utah 4-1 .112 .108 30 .231 51 .091 30 .706 75
26 Florida 4-1 .098 .196 17 .271 57 .079 36 .615 59
27 Colorado 3-2 .098 .122 28 .145 27 -.052 87 .294 20
28 Oklahoma 3-2 .097 .017 58 .197 39 .200 4 .442 32
29 Kansas State 2-2 .096 .107 31 .280 59 .058 39 .530 44
30 Memphis 3-1 .094 .228 11 .312 64 .035 52 .468 35

Posted by: Brian Fremeau on 12 Oct 2016

6 comments, Last at 17 Oct 2016, 9:46am by Brian Fremeau


by ramirez :: Wed, 10/12/2016 - 9:31pm

When 6 of the top 13 teams all play in the SEC, it really hurts the credibility of the ranking. What have teams like Mississippi and LSU done to deserve to rank that highly? Alabama should be in the top 2, and A&M is deservedly ranked in the AP poll at number 6. Beyond that, what is any SEC team doing near the top 10?

I'm tired of seeing teams with more worthy resumes, like Miami, Nebraska, and Wisconsin, ranked below teams like Auburn and Mississippi. Wisconsin lost a close game on the road at Michigan, and dropped out of the AP top 10. Louisville lost a close game at Clemson, and dropped 4 spots. Tennessee lost a close game at Texas A&M, and didn't drop in the poll. Guess which one of those teams plays in the SEC?

I understand that this FEI ranking is the product of a formula, and not votes. But it's these kind of results that reinforce the myth that the SEC is best by definition. If you think voters in the polls and on the CFP committee aren't influenced by rankings like the FEI and FPI, you're kidding yourself. Mississippi was blown out in the 2nd half by Florida State and Alabama, and their most impressive win was over a Georgia team that's not even in the AP top 25. Putting them at 6th in the country, or even 12th where the AP has them, is a joke.

by Brian Fremeau :: Thu, 10/13/2016 - 8:33am

I won't try to defend the movement of teams in the polls. The difference between Ole Miss and Wisconsin has a little bit to do with preseason projections (14 percent of this week's ratings) and a bit to do with Ole Miss' close loss to Alabama (which currently rates more impressively than Wisconsin's close loss to Michigan). I'll also note that if FEI didn't have LSU ranked so highly, Wisconsin would be further down the ratings themselves.

Miami and Nebraska have played efficiently, but their schedule strength to date (65th and 100th, respectively) holds them back in the FEI ratings. You didn't mention Western Michigan as being underrated, but they have a similar profile with Nebraska right now.

by LionInAZ :: Sat, 10/15/2016 - 8:56pm

The problem with this and any college ranking system is that poll rankings are still the fundamental basis. No real progress will be made until the subjective polls are eliminated altogether.

by Brian Fremeau :: Mon, 10/17/2016 - 9:46am

Polls are not a factor in this rating system.

by Aaron Brooks Go... :: Thu, 10/13/2016 - 9:07am

That VT-UNC game was a bit of a red-herring. Neither team was able to pass at all. This likely adversely affected UNC more than VT -- Evans can at least run.

VT's defense is good (they throttled ECU and it was the TOs that cost them against Tennessee), but it's not *that* good.

by Brian Fremeau :: Thu, 10/13/2016 - 5:57pm

The defensive performances in each of the hurricane-threatened games skew more impressively than they likely would have in more optimal conditions for sure. I don't have a reliable way to correct for weather, however. I think the impact of those games on efficiency ratings won't be too dramatic in the end, but something I should probably pay close attention to.

This piece did raise a few ideas for me on how much I should or should not weight the additional value a defense can provide. The Michigan example is an apt one -- the Wolverines defense did a near-perfect job against Rutgers of shutting them down cold. But they didn't generate much more turnover value on top of that which would have been even better through the lens of game splits. Turnover luck aside, is doing more than just shutting down an opponent an appropriate standard of defensive excellence?