Writers of Pro Football Prospectus 2008

14 Sep 2016

FEI Week 2: Early Dominance

by Brian Fremeau

The second weekend of college football wasn't nearly as splashy as the first. None of the teams ranked in the polls faced off against one another, and the game outcomes were overall fairly predictable as a result. It isn't uncommon for power programs to feast on a light schedule at the beginning of the year, and that was certainly the case for most of the top teams in the FEI ratings this week as well.

There were 46 games played last weekend between FBS opponents, and 21 of them were decided by at least four scores (45.7 percent). A total of 30 of the 46 games (65.2 percent) were decided by at least 17 points. Those are both significantly higher rates of domination than average when we consider the entire season. Since 2007, only 26.0 percent of FBS games have been decided by 25 or more points, and 44.9 percent of games have been decided by 17 or more points. But in comparison to early weeks of the year, it's precisely what we would expect since that's when power teams load up on non-power opponents. 51.7 percent of FBS games in Weeks 1 and 2 from 2007 to 2015 were decided by at least 17 points -- and 51.7 percent of FBS games so far this year have as well.

It's still very early, and projection data still accounts for 71 percent of the weight in this week's FEI ratings. But several teams have already produced a few dominant wins worth highlighting. Alabama sits atop the FEI ratings once again, but they're exceptionally far removed from the rest of the pack. The ratings differential between Alabama and No. 2 Stanford is currently larger than the ratings differential between Stanford and No. 15 Notre Dame. That's due not only to the Crimson Tide's raw efficiency data to date -- a combined overall victory margin of 90-16, 69-6 in 48 non-garbage possessions -- but also to the quality of opponents they've faced thus far. Alabama has defeated two teams ranked among the FEI top 30 and crushed both of them. No other program has even faced two top-30 opponents yet.

The most dominant team in terms of raw efficiency so far this year is the Michigan Wolverines. They're 2-0 on the year with a combined overall victory margin of 114-17, and a margin of 86-14 in only 36 non-garbage possessions. Their competition hasn't been anywhere near the level played by Alabama thus far -- Central Florida ranks 81st and Hawaii 128th this week -- which is why they haven't vaulted forward in the FEI ratings yet. Joining the Wolverines with a pair of efficient wins against FBS opponents is the team ranked right behind them, the Washington Huskies. Like Michigan, Washington has looked great against lousy opponents to date, with a combined non-garbage margin of 83-3 in 41 non-garbage possessions versus Rutgers and Idaho.

Team data isn't very well connected yet, which is why I ease out the impact of preseason projection data over the first six weeks of the year. But it's also important to note that blazing hot starts don't always lead to strong finishes.

Including Michigan and Washington, there have been 10 teams since 2007 that emerged from Week 2 with a pair of wins over FBS opponents and a game efficiency rating of at least .500 through two games. Only two of the previous eight teams to start the season that way ended up finishing the year ranked in the FEI top 10. The two most recent on the list (2014 Oklahoma and 2015 USC) combined to lose 11 games.

Teams With Two FBS Wins in First Two Weeks and .500+ Game Efficiency Since 2007
Year Team Wk 2 GE Final Rank FBS Record
2008 Arizona Wildcats .517 23 8-5
2009 Nebraska Cornhuskers .571 11 10-4
2010 Oregon Ducks .552 5 11-1
2011 Florida Gators .638 43 6-6
2011 Wisconsin Badgers .730 5 10-3
2013 Central Florida Knights .639 11 12-1
2014 Oklahoma Sooners .557 27 8-5
2015 USC Trojans .537 17 8-6
2016 Michigan Wolverines .571 ??? ???
2016 Washington Huskies .557 ??? ???

It may be premature to get too excited about the strong starts for Michigan and Washington, but their respective fanbases have every right to feel great about the recent trendline. Since the start of the 2007 season, the Wolverines have recorded a total of 22 FBS victories by at least four scores, and eight of those dominant victories have come since Jim Harbaugh took the reins of the team last fall. At Washington, the Huskies have had only 15 total wins over FBS opponents by 25 or more points since the start of the 2007 season, and five of those 15 dominant wins have come since Halloween last year.

Those programs have a long way to go to match the ten years of dominance from Alabama, of course. Since 2007, a total of 48 of the Crimson Tide's 99 victories against FBS opponents have been dominant wins by at least 25 points. Alabama is also the only program in the nation that hasn't suffered a single loss by 17 or more points in the same span.

FEI Ratings Through Week 2

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 represent approximately 71 percent of this week's ratings. Ratings for all teams are linked here.

Rk Team Rec FEI GE Rk SOS Rk
1 Alabama 2-0 .350 .375 12 .077 7
2 Stanford 1-0 .257 .161 26 .170 22
3 Clemson 2-0 .244 .060 47 .278 47
4 LSU 0-1 .231 -.023 66 .077 8
5 Oregon 1-0 .221 .257 14 .232 35
6 Oklahoma 1-1 .215 .105 35 .274 45
7 Florida State 1-0 .213 .121 31 .206 26
8 Mississippi 0-1 .210 -.121 85 .037 4
9 Houston 1-0 .206 .119 32 .363 71
10 Arkansas 2-0 .194 .024 56 .081 9
11 Ohio State 2-0 .182 .489 6 .213 28
12 Georgia 1-0 .182 .112 33 .318 60
13 Louisville 2-0 .173 .439 9 .149 19
14 Tennessee 2-0 .172 .154 27 .136 14
15 Notre Dame 1-1 .165 .196 24 .292 51
Rk Team Rec FEI GE Rk SOS Rk
16 Michigan 2-0 .157 .571 1 .342 67
17 Washington 2-0 .149 .557 3 .206 27
18 Michigan State 0-0 .141 - - .304 54
19 BYU 1-1 .137 .006 61 .320 62
20 Texas A&M 1-0 .134 .063 46 .037 5
21 Nebraska 2-0 .129 .351 13 .274 44
22 UCLA 1-1 .124 .080 41 .241 36
23 USC 1-1 .124 -.026 67 .030 3
24 Baylor 1-0 .122 .257 15 .342 68
25 Utah 1-0 .107 .011 59 .331 64
26 Boise State 2-0 .107 .233 18 .737 122
27 Wisconsin 2-0 .102 .208 22 .174 23
28 Iowa 2-0 .099 .457 7 .537 93
29 Western Kentucky 1-1 .097 .007 60 .148 18
30 Georgia Tech 1-0 .096 .043 52 .264 40

Posted by: Brian Fremeau on 14 Sep 2016

14 comments, Last at 15 Sep 2016, 8:55pm by ramirez

Comments

1
by techvet :: Wed, 09/14/2016 - 5:16pm

Wisconsin and LSU have practically identical SOS and yet Wisconsin is 23 slots lower despite beating LSU which is the only game reflected for LSU, right? Interesting.

2
by Brian Fremeau :: Wed, 09/14/2016 - 9:17pm

As indicated, the preseason projections still account for approximately 71% of this week's ratings. So LSU's preseason No.2 rating and Wisconsin's preseason No.40 rating still carry more weight than the results to date.

Also, the SOS ratings represent season-long strength of schedule, which like the FEI ratings themselves, will evolve over the course of the year.

3
by ramirez :: Wed, 09/14/2016 - 10:01pm

The problem with the strength of schedule rankings is that they work on the assumption that SEC games are harder than those of any other conference. That's why of the top 11 teams in SOS, 9 of them are SEC teams, including the entire SEC West. But the proposition that SEC games are more difficult just doesn't stand up to scrutiny. It's one of the great myths in sports.

What a lot of people don't know is that SEC teams face fewer power 5 opponents than the other major conferences. The Big Ten and Pac-12 both play 9 conference games, the SEC only 8. And Pac-12 and Big Ten teams play just as many power 5 teams out of conference as the SEC teams do. As an example, let's compare Mississippi State's schedule to that of California.

Miss State plays 8 SEC games, and one game OOC against BYU, a power 5 level team. Their other 3 games are against FCS Samford, Umass, and South Alabama (who already beat Miss St.). California plays 9 conference games, a game against Texas, and 2 against weaker opponents, San Diego State and Hawaii. So Cal plays 10 power 5 teams, and two mountain west teams. Miss St plays 9 power 5 teams, an FCS team, a Sun Belt team, and Umass, one of the worst programs in FBS. It's clear that Cal is playing more major conference opponents, and also the the weakest teams on Cal's schedule are better than the worst teams playing Miss St. But the FEI says that Mississippi St has the 2nd-toughest schedule in the country, Cal just the 24th-toughest. The only way you can get those results is by beginning the process by assuming SEC games are tougher than Pac-12 games.

Miss St is being rewarded for playing SEC doormats like south carolina, kentucky, and arkansas, who are all being overrated by the formula.

4
by Dennis :: Wed, 09/14/2016 - 11:39pm

It would be really great if we could take some teams from other conferences and put them in the SEC and see how they do. Oh wait, that actually happened when Missouri and Texas A&M joined in 2012.

A&M was decidedly average before joining, 26-25 in the previous four seasons, including 15-18 in the Big 12. And they've been very average in the SEC, 6-2 in conference their first year and 4-4 each of the next three.

Missouri was pretty good before joining, 36-17 overall the previous four seasons, 20-13 in the Big 12. They have two SEC East championships at 7-1 in conference in between 2-6 and 1-7 seasons.

Granted this is the latest show at small sample size theater, but if the SEC hype is accurate, it should have taken Mizzou and A&M years to become competitive in the conference.

5
by zenbitz :: Thu, 09/15/2016 - 12:38am

There simply aren't enough interconference Power-5 games to determine which conference is better. The SEC schedule is tougher because the pre-season rankings of SEC teams is higher on average than PAC-12.

6
by ramirez :: Thu, 09/15/2016 - 9:50am

Right, but that's the point. Why are the preseason rankings higher for SEC teams? What have schools like Tennessee and Mississippi State done to justify those rankings? I'm tired of hearing about how amazing the entire conference is. ESPN wants you to believe it, because they want to get two SEC teams into the playoff. So far, apart from Alabama, the other schools haven't cooperated, but I have no doubt they will try again this season. When you stack the deck by ranking all the SEC teams highly, it artificially enhances the rankings of the entire conference. Look at ESPN's preseason FPI rankings, which were a joke. That's why polls and formulas shouldn't have any rankings before the middle of the season, when we actually have a feel for the relative strengths of the teams.

9
by Brian Fremeau :: Thu, 09/15/2016 - 11:15am

There are two teams that have played two top 50 opponents so far this year, according to FEI: Alabama and Tennessee. They each won both games. None of those four victories was a conference win. If FEI is overcrediting those wins, it must have some kind of bias toward the Pac-12, Conference USA, Sun Belt, and ACC since those wins are among the most valuable data points of the season so far.

I understand that the SEC gets more credit in the preseason projections than any other conference due to its recent history, and I understand how that might not sit well, but it will work itself out as the season goes forward.

EDIT: actually one other team has played two top 50 opponents (and lost both games): Arkansas State

EDIT2: there is only one other team with two top 70 wins so far, and they're also in the SEC: Arkansas (wins over the Big 12 and Conference USA)

7
by Aaron Brooks Go... :: Thu, 09/15/2016 - 10:16am

Which is fine, but that doesn't refute garbage in-garbage out.

8
by hoegher :: Thu, 09/15/2016 - 10:47am

I am not privy to the inner workings of FEI, but I can almost guarantee that the "SEC" label carries zero weight in the ratings beyond a label.

These ratings are likely totally agnostic as to conference affiliation and the only inputs are the yards/drives teams put up on the field. The reason SEC teams rate out highly is not because of some code "IF CONF=SEC THEN FEI=FEI+10" but because those teams have performed well in the past and current games.

Also, FEI is pretty clear about not including FCS games, so not sure why that complaint is listed.

10
by Muldrake :: Thu, 09/15/2016 - 2:31pm

The problem of the SEC playing so many FCS teams is that those games are excluded and therefore reduce sample size and connectivity. It's possible that the SEC is being overrated as a whole when they're effectively playing 25% fewer of the games that create the connectivity with the sport as a whole.

11
by Brian Fremeau :: Thu, 09/15/2016 - 2:55pm

I would prefer far fewer FCS games overall for this connectivity reason, but I don't think it impacts the SEC any more than the many other conferences that also play many FCS opponents. I actually think FEI has tended to favor the Pac-12 more than other systems due in part to the greater connectivity in the conference and fewer FCS games played overall. I wonder if we'll see the Big Ten get that boost as the season progresses.

12
by oaktoon :: Thu, 09/15/2016 - 6:46pm

right now all I see is Texas defeating Notre Dame; Houston beating Oklahoma; and Wisconsin besting LSU. And all three teams that won are rated below the teams they beat-- with two of them not even close... So by what empirical evidence do pre-season ratings deserve to be part of this metric, let alone 71%?? Let them play a few more weeks and then we'll see who deserves to be where. Right now this is no better than a chimp throwing darts...

13
by Brian Fremeau :: Thu, 09/15/2016 - 7:24pm

Preseason ratings have proven to be effective at predicting results. No ratings could reasonably be expected to predict all results, of course. FEI game projections through the first two weeks have picked 82.8 percent of game winners and are 54.2 percent successful against the spread (the betting line has predicted 79.3 percent of game winners, for reference): http://www.thepredictiontracker.com/ncaaresults.php

I suppose it would feel better to have dropped LSU like a rock and put Wisconsin into the top 10 after the first week, etc. I doubt the game projections would have been as accurate doing so, however.

14
by ramirez :: Thu, 09/15/2016 - 8:55pm

Anyone know why the spam filter is blocking my comment?