Writers of Pro Football Prospectus 2008

Most Recent FO Features


» Futures: Josh Rosen

UCLA's quarterback clearly has the talent to succeed as an NFL starter. The question is whether or not he can avoid enough mistakes to become a superstar.

15 Oct 2008

FEI Week 7 Ratings

by Brian Fremeau

The Fremeau Efficiency Index principles and methodology can be found here. Like DVOA, FEI rewards playing well against good teams, win or lose, and punishes losing to poor teams more harshly than it rewards defeating poor teams. Unlike DVOA, it is drive-based, not play-by-play based, and it is specifically engineered to measure the college game.

FEI is the opponent-adjusted value of Game Efficiency (GE), a measurement of the success rate of a team scoring and preventing opponent scoring throughout the non-garbage-time possessions of a game. Like DVOA, it represents a team's efficiency value over average. Strength of Schedule (SOS) is calculated from a privileged perspective (explained here) and represents the likelihood that an Elite team (top-5) would post an undefeated record against the given team's opponents.

Now that enough game data is available, projected outcomes are no longer included in the FEI ratings. The following ratings are calculated from all FBS games played through Saturday, October 11. Only games between FBS teams are considered.

Rank Team Record FEI Last Week vs. Top 10 vs. Top 40 GE GE Rank SOS SOS Rank
1 Penn State 6-0 0.280 2 0-0 0-0 0.445 2 0.784 98
2 Missouri 4-1 0.257 1 0-1 0-1 0.295 11 0.533 41
3 Virginia Tech 4-1 0.257 5 2-0 2-1 0.065 41 0.343 8
4 Georgia Tech 3-1 0.256 7 0-1 1-1 0.219 17 0.490 33
5 Texas 6-0 0.252 8 1-0 1-0 0.469 1 0.638 69
6 Alabama 6-0 0.243 6 0-0 2-0 0.322 8 0.558 49
7 Oklahoma State 5-0 0.233 18 1-0 1-0 0.282 12 0.543 43
8 North Carolina 4-1 0.231 11 0-1 2-1 0.183 19 0.466 29
9 USC 4-1 0.227 4 0-0 1-0 0.310 9 0.771 95
10 Oklahoma 4-1 0.217 3 0-1 0-1 0.308 10 0.541 42
11 Georgia 4-1 0.215 12 0-1 1-1 0.124 25 0.424 19
12 Florida 5-1 0.214 9 0-0 0-1 0.351 6 0.637 68
Rank Team Record FEI Last Week vs. Top 10 vs. Top 40
13 Vanderbilt 5-1 0.204 10 0-0 2-0 0.122 26 0.481 31
14 Wake Forest 4-1 0.200 14 0-0 2-1 0.079 32 0.516 38
15 South Carolina 4-2 0.194 21 0-0 2-2 0.075 34 0.302 7
16 Mississippi 2-3 0.193 15 0-0 1-3 0.024 50 0.256 4
17 Ball State 6-0 0.182 13 0-0 1-0 0.331 7 0.778 97
18 Texas Tech 4-0 0.179 16 0-0 0-0 0.365 5 0.887 114
19 Michigan State 6-1 0.178 26 0-0 3-1 0.175 20 0.563 52
20 Pittsburgh 4-1 0.159 23 0-0 2-0 0.072 37 0.634 66
21 Boise State 4-0 0.146 25 0-0 0-0 0.248 14 0.899 115
22 South Florida 4-1 0.133 31 0-0 1-1 0.156 22 0.685 81
23 Navy 3-2 0.129 34 0-0 2-1 -0.028 64 0.456 27
24 Iowa 3-3 0.121 29 0-0 0-3 0.133 24 0.502 34
25 Utah 6-0 0.114 22 0-0 1-0 0.239 15 0.809 105

The Week 7 FEI Ratings for all 120 FBS teams can be found here. Expanded FEI Ratings data can be found here.

The Georgia Tech Yellow Jackets haven't been ranked in the Associated Press top 25 since early September last season. Running a triple-option offense at a BCS conference school in the 21st century under new coach Paul Johnson, Georgia Tech has defeated only three FBS teams. Last weekend, they ran up a whopping 10-7 victory over Gardner-Webb. What in the world are they doing at No. 4 in this week's FEI ratings?

First, the Gardner-Webb game, like all FCS games, isn't counted. But even if I did include it, I would have to also include the Yellow Jackets' dominant victory over Jacksonville State back in August, so it would probably be a wash.

Second, the projected results have now been entirely eliminated, and team ratings from now on will exclusively reflect results on the field in 2008. Teams like Georgia Tech actually received a boost simply by eliminating the remaining dead weight of their more modest projected rating. Meanwhile, two projected heavyweights, USC and Florida, dropped somewhat in the rankings even after posting a solid victory last week. Why? Their respective opponents, Arizona State and LSU, have done little to nothing yet this season to meet their projected expectations, and most of their other opponents didn't help improve their profiles either.

Third, line up Georgia Tech's resume with, say, Georgia's and try to defend the Bulldogs' more lofty public perception. Georgia Tech handed Boston College its only loss of the year in Boston, and Georgia won at South Carolina. Georgia Tech steamrolled Mississippi State (1-4) and Duke (2-2), and Georgia let Arizona State (1-4) and Tennessee (2-4) hang in against them. Georgia Tech lost a squeaker to Virginia Tech (4-1) and Georgia had its doors blown off against Alabama (6-0).

This isn't to suggest that Georgia Tech will be running all the way to a BCS championship. More than likely, they are going to play well down the stretch, never get blown out, stick around the FEI top 20, and be one of those mild-mannered but dangerous three-or-so-loss teams in late November -- just in time for the Georgia game.

Going Undefeated

Texas meets Missouri Saturday night in Act II of Burnt October, a four-part consecutive week mini-series drama co-starring Oklahoma, Missouri, Oklahoma State, and Texas Tech, each of whom have dominated opponents outside their circle. The Longhorns are actually one of ten remaining undefeated teams through seven weeks this season, three more than any other year since 2003. Five of the ten are non-BCS teams (Ball State, Boise State, Utah, BYU, and Tulsa). From 2003 to 2007, only six total non-BCS teams remained undefeated through seven weeks. None of this year's five appear to be world-beaters, but its too early to tell if they'll be remembered like 2007 Hawaii (worst undefeated regular-season team in memory) or 2004 Utah (should have had a chance against the best).

BYU and Utah play each other at the end of the year, and Texas, Texas Tech, and Oklahoma State still must play a round robin. Therefore, only seven of the ten current undefeated teams can mathematically run the table. How likely is it to happen for any of them? The table below shows each team's likelihood of winning all of its remaining scheduled games (possible conference title games and bowl games are not figured into the equation), as a function of the current FEI ratings and the Projected Win Expectation (PWE) model.

Likelihood of Remaining Undefeated
FEI Rank Team W-L Liklihood Pct.
21 Boise State 4-0 73.0%
1 Penn State 6-0 50.3%
6 Alabama 6-0 47.1%
17 Ball State 6-0 46.7%
25 Utah 6-0 33.2%
37 Tulsa 5-0 24.0%
5 Texas 6-0 15.2%
7 Oklahoma State 5-0 14.0%
35 BYU 5-0 8.1%
18 Texas Tech 4-0 2.6%

Through seven weeks last year, Hawaii had only a 10.6 percent chance of winning out. Combined, FEI expected one or two of the six remaining undefeated teams in 2007 to succeed in doing so, and only the Warriors were able to pull it off. This year, FEI expects three of the remaining ten to complete a perfect regular season.


This Sunday, the first BCS Ratings of 2008 will be released. You can count on the following to be mentioned ad nauseum by the talking heads next week: The number one team in the first BCS ratings has appeared in the BCS championship game in each of the last five seasons. Its a coveted spot, for sure. What you might not hear from the talking heads is that four of those front-runners lost the BCS championship game, and the eventual national champion has been ranked outside the BCS first-release top 5 twice in five years.

BCS Champions Through Week Seven, 2003-2007
Year Team W-L FEI Rank BCS Rank
2007 LSU 6-1 1 4
2006 Florida 6-1 2 6
2005 Texas 6-0 3 2
2004 USC 6-0 1 1
2003 LSU 4-1 5 12

I'm pretty sure you won't hear from the talking heads that the BCS champion and the FEI No. 1 team at season's end has appeared in the Week Seven FEI Top 5 every year since 2003. Nine of the ten BCS championship game participants since 2003 have appeared in the Week Seven FEI Top 5 (Ohio State last season was No. 12 in Week 7, the lone exception).

The 2007 LSU Tigers held the FEI No. 1 ranking from Week Seven through the BCS championship game last year. Will Penn State match the same feat? Will this weekend's Missouri vs. Texas clash be only a precursor of a Big 12 championship matchup for a berth in the BCS title game? If the SEC and Big 12 are won by two-loss teams, could one of the under-the-radar ACC teams really sneak in? Or will both BCS participants emerge from outside this week's FEI top 5 for the first time ever? Five years of data is a trend, not a guarantee, Nittany Lions fans. But for the rest of the nation, fans everywhere can start wringing their hands now over another good chance at a BCS championship berth for the Big Ten.

Posted by: Brian Fremeau on 15 Oct 2008

17 comments, Last at 16 Oct 2008, 6:03pm by Kirk


by emcee fleshy .S... :: Wed, 10/15/2008 - 3:28pm

Thanks for the explanation on Tech. Still don't really believe you. But hope you're right.

ATL's FOMBC Wasn't My Fault!

by White Rose Duelist :: Wed, 10/15/2008 - 3:29pm

Boise State's and Utah's combined probability of going undefeated is 106.2%, despite having to play each other.

by emcee fleshy .S... :: Wed, 10/15/2008 - 3:32pm

Dude, you have to multiply the numbers, not add them.

(If you're joking, I apologize for this note.)

ATL's FOMBC Wasn't My Fault!

by Anonymous (not verified) :: Wed, 10/15/2008 - 10:44pm

I think WRD meant the probability of one or the other going undefeated, which would be the sum if they did in fact play each other.

by emcee fleshy .S... :: Thu, 10/16/2008 - 4:31pm

That still wouldn't be the sum. The odds of one or both of them going undefeated would be the odds of them each -not- going undefeated, multiplied by one another, then subtracted from one.


by Anonymous (not verified) :: Wed, 10/15/2008 - 3:37pm

Boise State does not play Utah, BYU does.

by Travis :: Wed, 10/15/2008 - 3:37pm

BYU plays Utah, not Boise State.

by abernethyj :: Wed, 10/15/2008 - 3:50pm

Could you add a column in next week's table with a "Remaining SOS" for the updated data each week? Enjoyed reading about SOS in general last week.

by Brian Fremeau :: Wed, 10/15/2008 - 5:03pm

The complete Strength of Schedule data is available in the "Expanded FEI Ratings data" link provided under the FEI table above.

by Anonymous (not verified) :: Wed, 10/15/2008 - 4:01pm

BYU has been very dominant this year een though their ranking is low. How bad is their SOS that it could place them out of the top 30?

btw, the talking heads also won't mention South Florida's Top 10 ranking last year before getting raped by the baby Leaf-led Oregon Ducks.

by Brian Fremeau :: Wed, 10/15/2008 - 5:09pm

BYU's SOS is No. 119. They have mostly dominated, yes, defeating No. 113 Washington (0-5), No. 105 UCLA (2-4), No. 115 Wyoming (1-5), No. 117 Utah State (1-5), and No. 103 New Mexico (3-4) to date.

by Will Allen (not verified) :: Wed, 10/15/2008 - 6:14pm

Given the PAC 10's poor showing against the Mountain West this year, it would appear to me that the SOS for many PAC 10 teams is too high, and the SOS for many Mountain West teams is too low. I acknowledge that I am likely overlooking something.

Also, Brian, have you measured yet the predictive quality of the FEI rankings vs. the predictive quality of the well known ranking systems yet? Is the sample size big enough yet for such a comparison to be strongly suggestive of anything?

by Brian Fremeau :: Wed, 10/15/2008 - 11:43pm


I'm not sure the FEI SOS numbers don't already reflect your instincts. The only two Pac-10 teams with strong SOS at this point are Oregon State (they have played Penn State, USC and Utah already) and Arizona State (Georgia, USC and California). The Mountain West SOS numbers are low, and that has a lot to do with the weak Pac-10. That poor showing by the Pac-10 in those games shouldn't boost the MWC SOS, it should lower it.

I haven't done any systematic comparisons of FEI with other well-known ranking systems. FEI does appear along with a number of other systems in a comparison and consensus rating produced by Kenneth Massey. A comprehensive comparison would probably have to be an off-season project.

by Anonymous (not verified) :: Thu, 10/16/2008 - 3:10am

"The FEI No. 1 team at season's end has appeared in the Week Seven FEI Top 5 every year since 2003."

I'm not sure that this is a bragging point. Not to say your system is broken but if we set Ho = system broken, then the mere fact that it stays broken from week 7 to the end of the season certainly gives us no reason to reject the null.

"Nine of the ten BCS championship game participants since 2003 have appeared in the Week Seven FEI Top 5"

Also using one flawed system (BCS) to validate your own system seems flawed as well.

I would tend to think that the only decent indicator of your systems accuracy could only come by way of projecting the results of actual matchups as using another flawed ranking system doesn't really prove much and remaining SoS may be a bigger influence than a team's own ranking when it comes to predicting final standing.

Sorry for the negativity - definitely a neat approach. I was just underwhelmed by the comments you made in the last two paragraphs to talk up your past results.

by Anonymous (not verified) :: Thu, 10/16/2008 - 5:35am

I couldn't help but notice that in the ratings the MAC is at parity with the PAC-10. The top 10 MAC teams are rated 17, 36, 41, 50, 52, 56, 57, 65, 77 and 84; the PAC-10 comes in 9, 26, 47, 49, 58, 63, 66, 105, 113, 118.

There is only one head to head matchup between the conferences, 47 Arizona beating 57 Temple 41-16. But there have been ten common opponents, against which the MAC (average FEI rating 67) has been outscored 36.8-21.4 and is 1-9; the PAC-10 (average FEI rating 56) outscored the same teams 31.6-26.4 and is 5-5. The PAC-10 teams were 20.6 points - three touchdowns - better. (They also were at home much more often in these matchups).

If you just focus on the six common opponents where the MAC team has a higher rating, Boise St. beat 52 Bowling Green by 13 and won at 66 Oregon by 5; Georgia beat 56 Cent. Mich. by 39 and (at) 58 ASU by 17; Purdue beat 56 Cent. Mich. by 7 and lost to 66 Oregon by 6; Tennessee beat 41 N Illinois by 4 and lost at 105 UCLA by 3; Fresno St. beat (at) 86 Toledo by 1 and beat 105 UCLA (at) by 5; Idaho lost to 36 West. Mich by 23 and lost at 47 Arizona by 70. The PAC-10 teams (average 75 rating) are 14.5 points better than the MAC teams (average 55 ranking). The MAC teams were only home once and the PAC-10 teams five times....

I'm skeptical that even in a bad year for the PAC-10 and a very good year for the MAC, that the two conferences would be even close to roughly comparable in terms of quality.

by tide182 :: Thu, 10/16/2008 - 9:50am

wow - that is one crazy set of rankings at the top. i sure will be interested to see how this plays out because any set of rankings that includes missouri (what have they done, to use the GT argument?), va tech (really?), and ga tech ahead of texas, alabama, usc, georgia and florida is certainly eye-opening. my subjective eyes - having watched all of the aforementioned teams at some point this season - can't imagine any of the 2-4 ranked teams beating any of those teams i just mentioned below them at a neutral site. also hard to believe a champion is coming from that former group, so does that mean penn state - texas in the BCS title game?

ive always been taught to trust the numbers more than my own eyes, and certainly here's an example of a showdown between the two...

also, as a last point, isnt in true that polls using strength of schedule can be very misleading until you get to the last week or two when people have had the chance to play out their full conference schedule thereby mitigating some of the effects of fluke performances vs. exceptionally strong or weak competition?

by Kirk (not verified) :: Thu, 10/16/2008 - 6:03pm

Can you explain why PSU is higher than Texas, a team with a higher GE and tougher SOS than them? Thanks.