Fremeau Efficiency Index Bowl Forecast

Guest column by Brian C. Fremeau

The college football bowl season may be the least predictable series of contests in the world of sports. Teams haven't seen the field in anywhere from 17 to 50 days. From across the country or just down the road, some teams travel with hoards of jubilant fans in tow, while others shoulder the heavy baggage of disappointing losses and season-hinging what-ifs. Coach-less, coach-in-transition, or new-coach-at-the-helm squads find themselves grasping for motivation and inspiration in the midst of a week that is equal parts vacation, parade float, press conference and business trip. And teams accustomed to playing eight or more conference and rivalry games a year face off against unfamiliar competition for the first time in months, years, or ever. (13 bowls, including three of the five BCS bowls, will match up opponents that have never before met.)

Against this backdrop, we revisit the Fremeau Efficiency Index (FEI), first introduced to Football Outsiders back in November. Taking a systematic approach to evaluating Division-1A football through the prism of drive-by-drive possession efficiency, FEI attempts to measure how well a team played its games over the course of the season and ranks all 119 teams accordingly.

Fremeau Efficiency Index -- Final Pre-Bowl Ratings

  • FEI = Opponent-adjusted average Game Efficiency
  • GE = Raw average Game Efficiency
  • OE/DE = Raw Offensive/Defensive Efficiency

1 Ohio State 12-0 0.500   0.338 2 0.436 5 0.133 3
2 Michigan 11-1 0.456   0.218 12 0.373 17 0.165 8
3 LSU 10-2 0.429   0.306 3 0.427 7 0.148 5
4 Florida 11-1 0.410   0.164 18 0.319 37 0.166 9
5 Louisville 11-1 0.408   0.278 6 0.433 6 0.168 10
6 Wisconsin 10-1 0.390   0.258 8 0.332 29 0.120 2
7 Tennessee 9-3 0.375   0.155 20 0.386 13 0.254 48
8 Arkansas 9-3 0.363   0.118 28 0.309 42 0.221 32
9 West Virginia 9-2 0.353   0.259 9 0.498 2 0.256 49
10 USC 10-2 0.324   0.209 13 0.382 15 0.183 17
11 Notre Dame 10-2 0.324   0.228 11 0.417 8 0.249 44
12 Auburn 10-2 0.307   0.140 23 0.355 22 0.208 25
13 Georgia 7-4 0.289   0.068 39 0.248 68 0.22 31
14 South Carolina 6-5 0.28   0.100 34 0.376 16 0.246 43
15 Rutgers 9-2 0.276   0.149 21 0.304 44 0.178 14
16 Boise State 11-0 0.266   0.287 5 0.491 4 0.180 16
17 Clemson 8-4 0.266   0.263 7 0.383 14 0.174 12
18 Alabama 6-6 0.250   0.034 48 0.239 76 0.218 29
19 Georgia Tech 8-4 0.249   0.085 38 0.272 55 0.156 6
20 Virginia Tech 9-2 0.243   0.172 16 0.255 64 0.118 1
21 Wake Forest 10-2 0.240   0.094 36 0.265 60 0.177 13
22 Boston College 8-3 0.230   0.132 24 0.274 52 0.191 21
23 Penn State 7-4 0.222   0.049 44 0.223 87 0.161 7
24 Cincinnati 6-5 0.203   -0.007 63 0.208 90 0.210 26
25 California 8-3 0.200   0.148 22 0.324 33 0.227 36

For a complete FEI ranking of all Division-1A teams, click here. For a more detailed explanation of Game Efficiency data and FEI methodology, click here.

The extent to which FEI provides a better measure of team strengths than other college football ranking models cannot be adequately measured without more data and an extensive (read: off-season) analysis. That it provides a different perspective on relative team strengths from other models is clear, thanks in part to Kenneth Massey's College Football Ranking Comparison. Of the 100+ rating systems included in Massey's comparison, FEI rates as one of the least "correlative to consensus." In the BCS era, correlation to consensus may be a rating system's most treasured attribute, but should that be so? The three consensus-overrated teams identified in the 11/9 FEI article (Texas, Oregon, and Tulsa) went a combined 1-7 over the remainder of the season. Maybe correlation to consensus itself is overrated. In that spirit, let's take a closer look at ten of the most intriguing upcoming bowl match-ups from the perspective of Game Efficiency and FEI.

Note: A breakdown of team field position efficiency contributes to some of the bowl game commentary below. As a reference, the following graph illustrates the national offensive/defensive efficiency rates for drives starting from each yard line on the field, and breaks down field position into three broad categories:

  • Long (70-99 Yards from End Zone; 60.8% of Possessions; 45.4% of Scoring)
  • Mid (45-69 Yards from End Zone; 28.8% of Possessions; 33.6% of Scoring)
  • Short (1-44 Yards from End Zone; 10.4% of Possessions; 21.0% of Scoring)

Average starting field position in college football is between the 29- and 30-yard lines. There were 14,314 "competitive drives" in Division-1A during 2006.

FEI Forecasts represent the expected Margin of Victory at the conclusion of competitive possessions in the game listed, calculated as a function of the current FEI rating of each team and the expected number of competitive possessions to be played. Each team's expected offensive efficiency measured against its opponent's defensive efficiency determines the forecasted scores. FEI Forecasts for all 32 bowl games can be found here.

Emerald Bowl, December 27, Florida State vs. UCLA

One play can sometimes make a season, and Eric McNeal's deflection-spin-dive-interception to seal the victory for UCLA over USC qualifies for Play of the Year consideration. But is one play -- or one spectacular game, for that matter -- truly more meaningful than UCLA's other less-than-efficient defensive efforts over the course of the season (e.g. versus Washington State, California, and Oregon)? The Bruins' #23 raw defensive efficiency for the season rates one spot above Florida St. Offensively, FSU has a slight edge. Turnovers may tip the scales and field position for either team, but expect to see the Bruins team that stumbled five times this season more often than the one that shocked the Trojans.

FEI Forecast: Florida St. 24, UCLA 14

Points 24 14
Off Possessions 12 12
Long Poss/Pts 7 / 10 7 / 7
Mid Poss/Pts 4 / 7 4 / 7
Short Poss/Pts 1 / 7 1 / 0

Holiday Bowl, December 29, California vs. Texas A&M

Generally, 8-3 records are nothing to sneeze at in BCS conferences. Then again, Cal suffered a beating at Tennessee and lost to 5-6 Arizona. A&M was only 6 total points away from undefeated, but also 13 points away from 3-8. Aaachoo Aaachoo. This game will be another squeaker for the Aggies, but their inability to generate Short Fields (they generated only eight Short Field opportunities all season; the Bears only allowed seven) will find them on the wrong end of the final score.

FEI Forecast: California 24, Texas A&M 20

  Cal A&M
Points 24 20
Off Possessions 11 11
Long Poss/Pts 7 / 10 7 / 10
Mid Poss/Pts 3 / 7 4 / 10
Short Poss/Pts 1 / 7 0 / 0

Alamo Bowl, December 30, Texas vs. Iowa

Since their last victory on November 4th, the Longhorns lost two games, the Big 12 South, a BCS berth, and a whole herd of pride. Iowa hasn't won since October 28th, and last won a Big Ten game on October 7th (2-6 in conference, 0-5 in conference down the stretch). Isn't this what bowl season is all about? This game could get completely out of hand in Texas' favor, but FEI likes Iowa's ability to limit the Texas Long and Mid field efficiency.

FEI Forecast: Texas 24, Iowa 21

  Texas Iowa
Points 24 21
Off Possessions 11 11
Long Poss/Pts 7 / 10 7 / 7
Mid Poss/Pts 3 / 7 3 / 7
Short Poss/Pts 1 / 7 1 / 7

Chick-Fil-A Bowl, December 30, Virginia Tech vs. Georgia

Georgia was written off for dead by many halfway through the season; FEI, however, had a 5-4 Bulldogs team rated #21 back on November 9, just before back-to-back upsets of Auburn and Georgia Tech turned the season around. The Hokies boast a better record, a stronger defense (#1 in raw efficiency) and a stronger offense than Georgia, but how would things shake out if the teams traded schedules? Georgia made enough plays against a Yellow Jackets team on a weekend that showcased three big SEC over ACC wins. Look for the Bulldogs to do it again.

FEI Forecast: Georgia 17, Virginia Tech 14

  UGA VaTech
Points 17 14
Off Possessions 11 11
Long Poss/Pts 7 / 7 6 / 0
Mid Poss/Pts 3 / 7 3 / 0
Short Poss/Pts 1 / 3 2 / 14

Capital One Bowl, January 1, Arkansas vs. Wisconsin

One of the most promising bowl matchups offers the undercard to the BCS Championship Game, a Big Ten versus SEC clash featuring a mostly overlooked Wisconsin team hoping to make a statement against Heisman runner-up Darren McFadden and the Razorbacks. Arkansas' only notable out-of-conference game was a beatdown at the hands of USC in September, while Wisconsin's notable OOC matchup ... is this bowl game. The Badgers may be untested, but they mostly dominated their schedule with a solid running game and an especially stingy defense. Expect Arkansas to get McFadden the ball in every way imaginable, but don't expect it to be enough.

FEI Forecast: Wisconsin 21, Arkansas 17

  Wisc Ark
Points 21 17
Off Possessions 11 11
Long Poss/Pts 6 / 7 6 / 7
Mid Poss/Pts 4 / 7 4 / 7
Short Poss/Pts 1 / 7 1 / 3

Rose Bowl, January 1, USC vs. Michigan

While there may be disappointment in both locker rooms with the vanishing of BCS Championship hopes in the final weekend of the year, there should be no shortage of motivation, especially on the Wolverines sideline, since a Michigan win and an Ohio State loss might garner UM some AP vote recognition when all is said and done. The Trojans' wildly inconsistent offensive play cost them in two losses to significantly inferior opposition; Michigan's only true fault is that they happen to play a yearly series against the only team in the country better than them this year. Michigan's ball-control offense will keep USC out of Short Field opportunities (UM leads the nation with only 3.7% of opponent possessions starting in Short Fields) and the swarming defense will produce a chunk of Mid Field possessions -- enough to seal the win.

FEI Forecast: Michigan 27, USC 17

Points 27 17
Off Possessions 11 11
Long Poss/Pts 6 / 10 7 / 7
Mid Poss/Pts 4 / 10 3 / 7
Short Poss/Pts 1 / 7 1 / 3

Fiesta Bowl, January 1, Boise State vs. Oklahoma

The last time a team crashed the BCS party, undefeated Utah cruised through a Fiesta Bowl pounding of overmatched Pittsburgh two years ago. Undefeated WAC-champ Boise State, behind an excellent offense and underrated defense, will test its mettle against a much stronger opponent in Oklahoma -- or are they? According to FEI, Oklahoma is one of the most overrated teams in the country, hanging their hat on solid but not outstanding performances against almost entirely ordinary competition. They don't have a victory that compares with either of Boise State's best victories (Oregon State, in dominating fashion, and Hawaii), and both Sooner losses came when their backfield featured a healthy Adrian Peterson. Historically, sure, the Big 12 champ is a stud, but not this year. The Broncos will notch one more victory -- and hope for the Gators to take down the Buckeyes.

FEI Forecast: Boise St. 28, Oklahoma 20

  Boise Okla
Points 28 20
Off Possessions 11 11
Long Poss/Pts 6 / 14 7 / 10
Mid Poss/Pts 3 / 7 3 / 7
Short Poss/Pts 2 / 7 1 / 3

Orange Bowl, January 2, Louisville vs. Wake Forest

Louisville needed an overtime loss by Rutgers on the final day of the season to get to Miami, but their BCS ticket was effectively punched at the conclusion of their November 2nd pinball game versus West Virginia, after a 44-point scoreboard barrage. Wake punched their ticket in an almost unwatchable 9-6 victory over Georgia Tech in the ACC Championship. Which style of play will triumph? The Demon Deacons' apparent conservative field position philosophy (Coach of the Year Jim Grobe went for it on fourth down on only six occasions this season, least in the nation) plays right into Louisville's strength (the Cardinals are third in the nation in Long Field efficiency at 41.4%). Expect Louisville to jump out early and stay a couple of scores ahead the rest of the way.

FEI Forecast: Louisville 27, Wake Forest 14

  L'Ville Wake
Points 27 14
Off Possessions 11 11
Long Poss/Pts 6 / 14 6 / 7
Mid Poss/Pts 4 / 10 3 / 0
Short Poss/Pts 1 / 3 2 / 7

Sugar Bowl, January 3, Notre Dame vs. LSU

Twice each this season, Notre Dame and LSU dropped a high profile big game against a top opponent. LSU's losses to Auburn and Florida are generally attributed to their own mistakes and underperformances; ND's losses to Michigan and USC are attributed to their being overmatched by superior opponent team speed. Notre Dame's ability to defend LSU's balanced attack might decide the game, but the Irish offense may be an even bigger factor. ND last threatened a team-speed-advantaged opponent in their down-to-the-wire loss to USC last season, a game in which the Irish controlled the clock for 44+ minutes and executed three Long Field touchdown drives. The difference between last year's ND offense and this year's? A 7% drop in Long Field efficiency that cost the Irish more than 35 points on the season and placed added pressure on the much-maligned defense. Notre Dame could surprise the Tigers with a return to 2005 offensive form, but LSU (who, like Ohio State, fields a top-seven offense and defense) should be able to control the game.

FEI Forecast: LSU 24, Notre Dame 17

Points 24 17
Off Possessions 10 10
Long Poss/Pts 6 / 14 6 / 7
Mid Poss/Pts 3 / 7 3 / 3
Short Poss/Pts 1 / 3 1 / 7

BCS Championship Game, January 8, Ohio State vs. Florida

If the debate over whether or not Florida should be playing Ohio State for the title doesn't get in the way, the BCS Championship game promises to be a very good game, and will hopefully deliver on at least some of the magic of last year's USC vs. Texas epic. In addition to controversy, what do the Gators bring to the table? Defensively, they present the Buckeyes with much the same challenge as did Michigan, and even though the games will have been played 50 days apart, OSU is not likely to light up the scoreboard for 40+ against a top-ten defense twice in a row. The edge in the game lies with Ohio State's advantage in converting Mid Field drives into points (48.3% efficiency on the season), possessions that will likely be created by the Buckeyes' own top-ten defense forcing three-and-outs for Florida -- 17 of OSU's 42 Mid Field drives this year followed opponent three-and-outs; 23 of 42 followed opponent drives of under 20 yards; only two of 42 were initiated via opponent turnovers (the national average was 13%). If Florida can sustain drives -- even those that don't result in scores -- in order to force Ohio State into more Long Field possessions, they can pull off the upset. A defensive or special teams score might do the trick as well. In the end, though, look for Ohio St to cap off a perfect season and a (relatively) undisputed championship with a victory.

FEI Forecast: Ohio St. 24, Florida 17

  Ohio St. Florida
Points 24 17
Off Possessions 10 10
Long Fields / Pts 6 / 7 7 / 7
Mid Fields / Pts 3 / 10 2 / 3
Short Fields / Pts 1 / 7 1 / 7


22 comments, Last at 01 Jan 2007, 11:33pm

#1 by beedubyuh (not verified) // Dec 23, 2006 - 2:26pm

Great work! This is neat stuff. I look forward to seeing more next season. Thanks!

What's that? Oh, alright. First! (sigh)

Points: 0

#2 by RecoveringPackerFan (not verified) // Dec 23, 2006 - 4:01pm

Interesting stuff. I look forward to seeing the further development of this system.

Points: 0

#3 by Jesse (not verified) // Dec 23, 2006 - 6:47pm

ND is going to get destroyed by LSU, Oklahoma is going to win by about 14, and Michigan-USC is going to be closer...interesting read, though I'm not sure how much faith I have in the formulas.

Boise's win over Oregon State isn't as impressive as it would be if it happened now. OS wasn't the team they are now, just as USC wouldn't spank Nebraska the way they did way back in September. A 7 point win over the #31 team in the country isn't that impressive either.

OU's only real loss came against Texas on a really bad day. If Boise State even stays within 10 I'll be impressed

Points: 0

#4 by Atul (not verified) // Dec 23, 2006 - 6:59pm

what about the gator bowl? surely that matchup is more interesting than a horrible Seminoles team playing against the lucky bruins

Points: 0

#5 by El Angelo (not verified) // Dec 23, 2006 - 7:03pm

There's a lot of 24's and 17's in those predictions, FWIW.

Points: 0

#6 by Tally (not verified) // Dec 23, 2006 - 8:08pm

Re 5:

Just two...

Points: 0

#7 by toomuchtimeonmyhands (not verified) // Dec 23, 2006 - 9:54pm

5 & 6,

Here's a tally of each final score (not by match-up, but by team):

14 3
17 5
20 2
21 2
24 5
27 2
28 1

There are more 17s and 24s than any other score, but, in answer to comment 5's implicit question, this information isn't worth much of anything if we don't know how often teams end up with these scores in general.

Points: 0

#8 by Erasmus (not verified) // Dec 24, 2006 - 12:52am

Alabama at 18? Is that good?

Points: 0

#9 by jimmo (not verified) // Dec 24, 2006 - 12:56am

here's hoping Brian! I'm riding the index in my pool this year (starting with today's games), and am one of 25 (of 127) at 3-0.

I picked 'em straight off the final rankings on the site, and I've been rewarded with a win already; years past, I would have surely gone with New Mexico over San Jose because of the home field.

Points: 0

#10 by Chris (not verified) // Dec 24, 2006 - 1:26am

Too bad bowl games are determined more by intangibles than straight up stats, or else you may have something.

Points: 0

#11 by Kal (not verified) // Dec 24, 2006 - 1:41am

#10: So far he's 5-6 with the scores being very reasonable in comparison to the final outcome. What the FEI won't do is predict reasonably things like missed extra points and whatnot, so the scores are going to be less exact - but the margins should be pretty good.

The one that he got wrong he got REALLY wrong - Troy just owned Rice, but he had Rice and Troy in a very close matchup. Makes me wonder if there are tweaks that can be done to get that prediction closer while getting the rest pretty good. For instance, I still don't know how special teams are weighted (or if they are implicit in the ratings given the success rate as the determining factor) nor do I know what kind of weights are being performed to make teams more or less quality.

Points: 0

#12 by Kal (not verified) // Dec 24, 2006 - 2:30am

Sorry, that's meant to read '5 OF 6', not '5 AND 6'. He's picked 5 of the 6 games right, damnit.

Points: 0

#13 by Disco Stu (not verified) // Dec 24, 2006 - 2:46am

Vegas has UCLA by 4, Va Tech by 3, Ark by 1.5, USC-Mich as a pickem, and Okla by 7.5... it sounds like there's money to be made...

Points: 0

#14 by james (not verified) // Dec 24, 2006 - 5:53am

rice played the backup qb for most of the game and he threw at a bunch of picks or else the prediction would prob be right

Points: 0

#15 by DrewTS (not verified) // Dec 24, 2006 - 10:44pm

I love this kind of stuff, so I hate to criticize. But the final scores all look very generic, and we all know that every game can't be 24-17 or 28-20. There are no shootouts or defensive struggles predicted, but they almost certainly will occur.

So I guess my $0.02 is that showing the final score prediction makes it look computerish, which might make it lose some credibility in some eyes. I like the suggestion in post #11 -- go with the margins, rather than the exact scores.

Points: 0

#16 by Dan (not verified) // Dec 26, 2006 - 3:23pm

UC at 24...finally my boys are getting a little recognition for their schedule. The Bearcats are definitely better than Oklahoma (#34). Okay that was a joke but I'm serious about the schedule.

Points: 0

#17 by Basilicus (not verified) // Dec 26, 2006 - 4:51pm

Very nice work. Like most innovative systems, it sheds some light as to what to look for. I don't think it's the whole tomato, so to speak, but it is a very clear system and much more than I've seen anyone else do in terms of college statistical analysis.

Points: 0

#18 by james (not verified) // Dec 27, 2006 - 9:53am

alright man,
fsu to score 24...if that's right than sell your FEI for a million dollars now...avid gambler who doesn't think that can happen is dropping his .02 here.

Points: 0

#19 by zlionsfan (not verified) // Dec 27, 2006 - 11:38am

I agree with DrewTS: I noticed the lower scores across the board in these predictions. Ten teams scored 40 or more points in last season's bowl games (in 156, I mean 28 bowls), eight in 2004, and nine in 2003, so it would be reasonable to expect a few scores in that range this year.

If the system is predicting what amounts to the mean result of each drive, then it doesn't seem to make as much sense to predict a final score as it would simply to predict margin of victory.

Then again, I guess that's how most systems predict final scores, as some sort of aggregation of individual results.

Intangibles? I thought bowl games were determined by touchdowns, field goals, extra points, and safeties ...

Points: 0

#20 by zlionsfan (not verified) // Dec 28, 2006 - 11:11am

By the way, I thought last night's Emerald Bowl telecast was excruciatingly bad. I'm almost wondering if ESPN is now openly campaigning for a 16-team playoff. I don't think I've ever heard Dan Fouts doing play-by-play instead of color, and I hope I never hear it again. It was like the Three Stooges all over again, except instead of arguing, they were just joking about random things.

Fortunately, there was a lot of South Park to watch on Comedy Central.

Points: 0

#21 by Disco Stu (not verified) // Dec 29, 2006 - 11:38am

Looks like Rutgers-K State was the biggest FEI mismatch of the bowl season, and that one proved pretty accurate

Points: 0

#22 by Rocco (not verified) // Jan 01, 2007 - 11:33pm

Well, the prediction about BSU-OU is looking smart right about now.

Points: 0

Save 10%
& Support Brian
Support Football Outsiders' independent media and Brian Fremeau. Use promo code FREMEAU to save 10% on any FO+ membership and give half the cost of your membership to tip Brian.