In this week's Varsity Numbers, Bill Connelly revisits some measures and concepts: Adjusted Scores, Covariance, and momentum (or whatever you choose to call it).
24 Oct 2007
by Brian Fremeau
The Fremeau Efficiency Index (FEI) 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, 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.
Only games between FBS (Division 1-A) teams are considered. The Week 8 Ratings represent the results of games played through Sunday, October 21, 2007.
|Rank||Team||W-L||FEI||Last Wk||GE||GE Rank|
|Rank||Team||W-L||FEI||Last Wk||GE||GE Rank|
Click here for rankings of all 119 teams.
Wait a second. Didn't South Florida lose last Thursday? Didn't the national media close the book on their Cinderella season? Didn't we all agree to relegate the Bulls to the bottom of the one-loss heap?
Not FEI, and unlike the polls, the reason behind FEI's rating of USF has nothing to do with the name on the uniform or the history of the program.
LSU and South Florida are currently the only programs in the country with multiple wins over top-10 FEI teams, and each have dropped one game to a top-15 FEI team. But while LSU maintains a comfortable position at No. 3 in the BCS, South Florida tumbled to No. 10 in the BCS this week and now has virtually no shot at the championship game. USF finds itself behind, among others, one-loss Oklahoma and one-loss West Virginia in the BCS, two good teams, but neither with a better resume than the Bulls. The Sooners lost to a now four-loss Colorado team, and the Mountaineers lost to none other than South Florida themselves. Poll anchoring and loss-consequences are one thing, but can WVU over USF on a ballot truly be justified?
The one top-10 team bested both by LSU and South Florida is Auburn, a team that hangs in at No. 5 despite another loss over the weekend. Three-loss teams don't typically garner this much respect, but the Tigers have faced a murderous schedule to date and have fared pretty well. Auburn is one of four teams with multiple victories over FEI top-20 teams, they have faced more top-10 (three) and top-20 (four) teams than any other team in the FEI top-25, and they lost to both No. 1 and No. 2 on the final play of the game.
Here's a look at this week's Underrated and Overrated teams in the polls.
Rutgers (4-2; NR BCS, No. 25 AP, No. 14 FEI)
By knocking off South Florida, Rutgers earned the biggest FEI bump from last week, and put itself in position to make another run at the Big East title that eluded them last season. Unlike last season, though, Rutgers is doing more with less, at least as far as field position is concerned. They have faced the largest drop-off in Field Position Advantage among FBS teams from last season, plummeting from an eight yard per possession advantage in 2006 to a nearly seven yard per possession disadvantage in 2007. Their excellent running game has barely missed a step, however, and spearheaded the victory last Thursday despite facing a six-yard per possession field position deficit for the game to USF. Up next are contests with No. 6 West Virginia and No. 21 Connecticut, two winnable games regardless of FPA. If Rutgers can shift the field in their favor, watch out.
Georgia Tech (4-3; NR BCS, NR AP, No. 20 FEI)
The Yellow Jackets haven't notched a true marquee victory yet this season, so why the FEI love? Georgia Tech has faced as many top-40 opponents as any other team, winning two of five, and as good teams are supposed to do, they have manhandled the weak competition they have faced. The schedule sets up nicely for a strong finish, with an off weekend to prepare for a date next Thursday with Virginia Tech, followed by a relatively light ACC slate until the annual rivalry game with Georgia.
Virginia Tech (5-1; No. 8 BCS, No. 8 AP, No. 33 FEI)
Virginia Tech's best win over No. 31 Clemson also happens to be their only victory over a team with a winning record. Against LSU in Week 2, the Hokies were as uncompetitive as any of the Tiger opponents this season. Several top teams have seemingly earned their lofty poll ranking by default this season, perhaps none more so than Virginia Tech. A Thursday night showdown with undefeated Boston College (another team with a little more to prove) gives the Hokies a second opportunity in the spotlight to assert themselves and actually earn poll recognition.
USC (6-1; No. 12 BCS, No. 9 AP, No. 34 FEI)
In the eyes of poll voters, the Trojans virtually cashed in a get-out-of-jail-free card with their 38-0 thrashing of Notre Dame on Saturday. Why hasn't FEI yet forgiven the "historic" loss to Stanford on October 6? Well, not only has USC not yet defeated a top-40 opponent, they haven't yet played one. In fact, four Southern Cal victories have come against teams ranked among the bottom 40 in FEI, and facing their toughest competition to date, USC escaped with victories over No. 51 Washington and No. 61 Arizona. With five FEI top-40 match-ups coming up, including No. 4 Oregon this week, plenty of opportunities remain for USC to improve their resume. Until they do, the jury is out.
On Saturday afternoon, North Dakota State defeated Minnesota 27-21 in the Metrodome, the seventh loss in an entirely forgettable season for the Golden Gophers. Yet Minnesota's FEI ranking climbed two spots this week and their record remains listed at 1-6. What gives?
As mentioned in the lead-in to each week's ratings, FEI considers only games between FBS (formerly known as Division-1A) opponents. All other FBS games, both wins and losses to FCS (formerly Division-1AA) teams, are ignored. None of the 60 victories by FBS over FCS teams, nor the eight losses, impact the weekly FEI ratings in any way. In any other year, ignoring those 68 games would go largely unnoticed. But one result seems to make this year different. See if you can pick it out:
I haven't received one inquiry this season about how FEI is going to handle Louisiana Lafayette's loss. I likewise haven't received one inquiry this season about how to deal with Ohio State's Week 1 victory over Youngstown State. Why does that epic Michigan game deserve special treatment?
FEI is built on the principle that wins over poor teams are relatively unimportant while losses, particularly to poor teams, are given significant weight. By this standard, a fair evaluation of Michigan ought to include the Appalachian State data point in some fashion. But ignoring that outcome has less to do with a fair evaluation of Michigan than it does with a fair evaluation of Appalachian State. With only one game against FBS competition, can we properly determine how good or bad Appalachian State truly is? Fellow FCS team North Dakota State has played (and won) twice as many games against FBS competition -- can we confidently assign a strength value to them, either?
The challenge of any college football ranking system to properly evaluate 119 teams is difficult enough, particularly with most teams within that large set playing only 11 or 12 games total, seven or more against a small subset of conference teams. And yet, the "connectivity" of each FBS team to every other FBS team dwarfs that of FCS teams. At regular season's end, the average FBS team will have played 11 FBS opponents, and have at least one common opponent with 53 other teams. But Appalachian State will share only one common opponent (Michigan) with each of only 11 other FBS teams. By the second generation of connectivity, an average FBS team will share more than 25 common opponent-opponents with every other FBS team. Appalachian State, meanwhile, will share an average of only two common opponent-opponents with less than half of the FBS teams by season's end.
By this measure of connectivity, the individual game data point for each FBS vs. FCS game is about ten times less valuable than the individual game data point in FBS vs. FBS games. Appalachian State's victory over Michigan may still be the headline of the year, but is it statistically significant in the context of Michigan's other 11 games this season?
While some computer ranking systems have employed different solutions to this problem, I refuse to assign an arbitrary value to Appalachian State, particularly one that would carry anywhere near the same weight as the other opponents on Michigan's schedule. Though it is certainly possible that the Wolverines could "unjustly" disrupt the national championship picture by ignoring this one result, I believe that a fair evaluation of Michigan can be made from their other 11 games, just as a fair evaluation of their rival Ohio State can be made by excluding its victory over Youngstown State.
27 comments, Last at 26 Oct 2007, 7:10pm by zlionsfan