Writers of Pro Football Prospectus 2008

printable version

Fremeau Efficiency Ratings

College football power ratings and analysis



29 Oct 2008

FEI Week 9 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 to date.

The following ratings are calculated based on data from all FBS games played through Sunday, October 26. 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 Texas 8-0 0.302 1 1-0 3-0 0.421 1 0.362 31
2 Penn State 8-0 0.284 3 0-0 2-0 0.388 4 0.586 71
3 North Carolina 5-2 0.277 10 0-1 4-2 0.160 18 0.304 16
4 USC 6-1 0.267 6 0-0 3-0 0.346 6 0.563 64
5 Florida 6-1 0.266 4 0-0 1-1 0.397 2 0.566 65
6 Alabama 8-0 0.245 2 1-0 2-0 0.291 8 0.474 51
7 Georgia 6-1 0.217 5 0-1 2-1 0.144 23 0.379 36
8 Virginia Tech 4-3 0.200 9 2-0 2-3 0.022 49 0.198 4
9 Georgia Tech 4-2 0.200 7 0-1 2-2 0.134 24 0.451 46
10 Oklahoma 6-1 0.200 8 0-1 1-1 0.291 9 0.370 32
11 Oklahoma State 6-1 0.197 11 0-1 1-1 0.246 12 0.309 18
12 Texas Tech 6-0 0.194 17 0-0 0-0 0.396 3 0.869 116
Rank Team Record FEI Last Week vs. Top 10 vs. Top 40
GE GE Rank SOS SOS Rank
13 Florida State 4-1 0.192 26 1-0 2-1 0.090 28 0.494 54
14 Missouri 5-2 0.184 15 0-1 1-2 0.220 14 0.272 11
15 Ohio State 6-2 0.166 18 0-2 2-2 0.088 29 0.210 5
16 Connecticut 5-2 0.184 15 0-1 1-1 0.093 27 0.396 38
17 Mississippi 3-4 0.157 14 1-1 1-4 0.022 50 0.214 7
18 Vanderbilt 5-3 0.141 13 0-1 2-2 0.071 35 0.377 35
19 Boston College 4-2 0.138 12 1-2 1-2 0.064 36 0.272 10
20 Ball State 7-0 0.135 21 0-0 0-0 0.327 7 0.861 113
21 Miami 4-3 0.134 41 0-2 2-3 0.025 48 0.174 3
22 Pittsburgh 5-2 0.129 16 0-0 2-0 0.073 34 0.614 77
23 Iowa 4-3 0.128 22 0-0 0-3 0.161 17 0.584 70
24 Wake Forest 4-3 0.120 28 0-0 2-1 0.005 57 0.421 41
25 South Carolina 4-3 0.119 19 0-1 1-2 0.054 38 0.397 39

Adjusted Offensive Efficiency (AOE) and Adjusted Defensive Efficiency (ADE) are the opponent-adjusted values of Offensive Efficiency (OE) and Defensive Efficiency (DE), explained here. Like FEI, the multiple-order adjustments are weighted according to both the strength of the opponent and the relative significance of the result; efficiency against a team's best competition faced is given more relevance weight. AOE and ADE represent a team's value over/under average. Positive AOE and negative ADE are the most valuable.

Rank Team Record AOE AOE Rank ADE ADE Rank OE OE Rank DE DE Rank
1 Texas 8-0 0.649 1 -0.386 7 1.083 1 -0.199 36
2 Penn State 8-0 0.530 3 -0.353 11 0.543 9 -0.526 4
3 North Carolina 5-2 0.177 34 -0.505 1 -0.043 59 -0.355 17
4 USC 6-1 0.431 6 -0.376 9 0.446 13 -0.580 2
5 Florida 6-1 0.365 10 -0.395 4 0.538 10 -0.590 1
6 Alabama 8-0 0.356 11 -0.401 3 0.199 33 -0.575 3
7 Georgia 6-1 0.423 7 -0.175 39 0.214 31 -0.150 40
8 Virginia Tech 4-3 0.041 57 -0.280 18 -0.214 82 -0.107 46
9 Georgia Tech 4-2 0.278 18 -0.181 37 -0.011 54 -0.398 12
10 Oklahoma 6-1 0.539 2 -0.234 26 0.876 4 -0.125 43
11 Oklahoma State 6-1 0.295 16 -0.248 24 0.608 8 0.031 64
12 Texas Tech 6-0 0.300 15 -0.281 17 1.018 3 -0.057 52
Rank Team Record AOE AOE Rank ADE ADE Rank
OE OE Rank DE DE Rank
13 Florida State 4-1 0.078 50 1-0.265 19 -0.065 63 -0.254 26
14 Missouri 5-2 0.463 4 -0.050 58 0.730 6 0.098 75
15 Ohio State 6-2 0.037 58 -0.393 6 -0.172 77 -0.409 9
16 Connecticut 5-2 0.226 26 -0.394 5 0.060 45 -0.371 15
17 Mississippi 3-4 0.275 21 -0.206 33 0.011 51 -0.043 55
18 Vanderbilt 5-3 -0.024 67 -0.369 10 -0.207 79 -0.313 21
19 Boston College 4-2 0.140 41 -0.350 12 0.052 47 -0.448 8
20 Ball State 7-0 0.277 20 -0.217 31 0.773 5 -0.249 27
21 Miami 4-3 0.090 48 0.097 75 -0.194 78 -0.035 57
22 Pittsburgh 5-2 0.278 19 -0.105 48 0.352 18 0.050 67
23 Iowa 4-3 0.018 64 -0.337 13 -0.015 57 -0.481 7
24 Wake Forest 4-3 -0.297 103 -0.478 2 -0.387 107 -0.312 22
25 South Carolina 4-3 -0.126 79 -0.380 8 -0.263 89 -0.359 16

The Week 9 FEI Ratings for all 120 FBS teams can be found here. Expanded FEI Ratings data can be found here.

Offense versus Defense

Every top 10 offense according to AOE was in action last weekend, and nine of those teams notched victories. Nine of the top 10 defenses were on the field, and six were victorious. Only one game over the weekend featured a top 10 AOE team versus a top 10 ADE team (Penn State versus Ohio State), the tenth such clash this season. The top 10 offenses have won five, and the top 10 defenses have won five. Georgia (No. 7 AOE) faces Florida (No. 1 ADE) on Saturday. What should we expect?

The following chart illustrates the relationship between the Offensive Efficiency measured in games and the relative strength of the offensive and defensive units in each match-up. The relative strength is calculated as the difference in standard deviation between one team's AOE and the opponent's ADE. Every game played in 2007 is plotted in blue. The marquee matchups in 2008 described in the previous paragraph are plotted in red.

A linear regression could be used to project offensive efficiency for each team in each matchup, and if no defensive or special teams turnovers or scores occur in a given game, AOE and ADE can be useful (if not precise) in making projections. The number of possessions in a given game is obviously critical, though variable, as well -- Penn State possessed the ball eight times total against Ohio State; Oklahoma possessed the ball seventeen times offensively against Kansas State.

That said, if Georgia begins an average number of possessions (12) from average starting field position (own 30) against Florida, I expect the Bulldogs to score 24 points. They managed 24 points in 13 possessions against Vanderbilt (No. 10 ADE), 14 points in 9 possessions against South Carolina (No. 8 ADE), and 16 points in 10 offensive non-garbage possessions against Alabama (No. 3 ADE).

Athletic Coast Conference Bias

Seriously, what in the world is going on with FEI's ACC fetish? Believe me, I'm asking myself the same question. North Carolina leaped up to No. 3, six ACC teams are in the FEI top 24, multiple teams are moving up after losses and none of it seems to make any sense at all.

Neither does the conference itself. Virginia lost its first three games of 2008 by 45 (to USC), by 35 (to Connecticut), and by 28 (to Duke). The Cavaliers now have the best record in ACC conference play. Maryland lost to one of the worst teams in the Sun Belt Conference (Middle Tennessee) in one week and defeated one of the best teams in the Pac-10 Conference (California) the next. Yet at 5-2, the Terrapins are ranked behind 1-4 Clemson in FEI. Are FEI's ACC brains fried, or can we try and figure this thing out?

I decided to remove all ACC games (both conference and non-conference) from the system to see if it could bring things into focus. If we vote the entire conference off the island, the resulting top 10 looks awfully familiar.


FEI Top 10 sans ACC
Rank Team W-L FEI BCS Rank AP Rank
1 Texas 8-0 0.318 1 1
2 Penn State 8-0 0.298 3 3
3 Florida 5-1 0.264 8 5
4 Alabama 7-0 0.260 2 2
5 USC 5-1 0.256 5 7
6 Georgia 6-1 0.247 6 8
7 Oklahoma 6-1 0.221 4 4
8 Oklahoma State 6-1 0.214 9 9
9 Texas Tech 6-0 0.208 7 6
10 Missouri 5-2 0.197 14 14

Throwing in a bunch of ACC conference games with every team hovering near .500 shouldn't disrupt things, right? So it has to be those pesky non-conference games. Has the ACC really impressed? Yes, apparently. Without anyone really noticing, the ACC has posted an 11-7 record to date against BCS conference (plus Notre Dame) opposition, including an undefeated record against the Big 12 (the lightweights, but still). The Big 12 is 7-7 against similar foes. The SEC is 5-6, including a loss by Vanderbilt at home to Duke last Saturday. Not for nothing, but Vanderbilt currently has the fourth-best conference record in the vaunted Southeastern Conference. Duke has the 11th-best conference record in the ACC.

Its hard to imagine that three-loss Virginia Tech, two-loss Georgia Tech, and two-loss North Carolina are actually "better" than Oklahoma, Oklahoma State or Texas Tech. But since the latter three still haven't played one another, FEI is going to have to wait and see. Discard outcomes and dismiss teams as you wish.

Posted by: Brian Fremeau on 29 Oct 2008

8 comments, Last at 30 Oct 2008, 9:58am by Tar Heel Mania

Comments

1
by mm (not verified) :: Wed, 10/29/2008 - 12:12pm

Well, we'll get to measure the ACC against the SEC some more in the rivalry games that always come up at the end of the year. Florida vs. Florida State, Georgia vs. Georgia Tech, and South Carolina vs. Clemson could shift the ACC up or down, both in public perception and in the FEI.

Throw in Vanderbilt vs. Wake Forest (not sure if that's a rivalry, but noticed it was scheduled), and you still have 4 SEC vs. ACC games that feature 7 teams in the FEI top 25!

2
by sethburn :: Wed, 10/29/2008 - 12:34pm

I wonder how one would solve for an a situation where a conference clearly outperformed during non-conference play. It is hard for a computer to know that teams aren't as good as their stats would indicate.

3
by chappy (not verified) :: Wed, 10/29/2008 - 2:40pm

Somethings wrong with the stats for the sans ACC breakout. Alabama's record doesn't seem right. If it is, how do they get credited with a loss?

4
by Travis :: Wed, 10/29/2008 - 4:10pm

It looks like USC's record is listed twice, once as "USC" and once as "Alabama".

5
by Brian Fremeau :: Wed, 10/29/2008 - 5:14pm

Yep, Alabama's line was listed incorrectly. It has been corrected.

6
by Pat (filler) (not verified) :: Wed, 10/29/2008 - 7:19pm

The ACC bias is all in the opponent adjustments, which probably implies there's something not quite right there. How are you doing those? If it's iterative (correct raw game efficiency based on opponent's game efficiency), you might try doing a second iteration (correct raw game efficiency based on opponent's FEI to get a second-order correction) and seeing what that does.

All of the ACC teams have massive upward adjustments - about ~0.1, at least, upwards for all of them. That seems a little weird.

7
by Brian Fremeau :: Wed, 10/29/2008 - 10:11pm

Pat --

I am already doing multiple-order adjustments, but I do agree that there is something about the adjustments themselves that needs to be investigated. An entire conference having an upward adjustment from GE to FEI in and of itself isn't alarming, especially if that conference has performed better than others in non-conference games. But the ACC hasn't been giant killers, they have just beaten more mid-level non-conference teams than anyone else.

There's something to be said for the schizophrenic quality of the ACC teams (Maryland and Virginia especially) that might need to be looked at, too. An abnormally high variance in play over the course of the season should perhaps be its own factor in the opponent adjustments. A team plays way over their heads one week and way below expectations the next. FEI might treat those results as equally "valid" whereas a fan might give the second result more weight than the first.

8
by Tar Heel Mania (not verified) :: Thu, 10/30/2008 - 9:58am

Statistically it makes sense, since both results are two extremes that help to establish a mean. It's no less frustrating, though.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd>
  • Lines and paragraphs break automatically.

More information about formatting options

CAPTCHA
To skip this, please log in.