Writers of Pro Football Prospectus 2008

Most Recent FO Features


» Snap-Weighted Age: 2017 NFL Rosters

The Rams and Jaguars finally turned their young talent into playoff berths in 2017. Cleveland looks to be taking over the mantle for "team of the future." Also: Buffalo's wild turnover and the youngest secondary in eight years.

22 Oct 2008

FEI Week 8 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.

Projected outcomes are not included in the following FEI ratings. The ratings are calculated based on data from all FBS games played through Saturday, October 18. 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 7-0 0.303 5 1-0 2-0 0.466 1 0.464 37
2 Alabama 7-0 0.278 6 1-0 3-0 0.286 10 0.386 29
3 Penn State 7-0 0.266 1 0-0 1-0 0.421 2 0.758 94
4 Florida 5-1 0.259 12 0-0 1-1 0.351 5 0.555 55
5 Georgia 5-1 0.252 11 0-1 2-1 0.127 26 0.277 6
6 USC 5-1 0.245 9 0-0 1-0 0.380 3 0.674 80
7 Georgia Tech 4-1 0.240 4 0-1 1-1 0.191 17 0.464 38
8 Oklahoma 5-1 0.230 10 0-1 1-1 0.298 9 0.353 19
9 Virginia Tech 4-2 0.222 3 2-0 2-2 0.046 48 0.289 8
10 North Carolina 4-2 0.202 8 0-1 1-1 0.141 21 0.497 43
11 Oklahoma State 6-0 0.200 7 0-0 1-0 0.314 8 0.675 81
12 Boston College 4-1 0.187 30 1-1 1-1 0.140 22 0.419 33
Rank Team Record FEI Last Week vs. Top 10 vs. Top 40
13 Vanderbilt 5-2 0.187 13 0-1 2-1 0.324 12 0.443 22
14 Mississippi 2-4 0.183 16 1-1 1-4 0.017 54 0.168 2
15 Missouri 4-2 0.182 2 0-1 1-2 0.147 20 0.266 4
16 Pittsburgh 5-1 0.180 20 0-0 2-0 0.134 25 0.629 73
17 Texas Tech 5-0 0.166 18 0-0 0-0 0.327 7 0.880 114
18 Ohio State 6-1 0.165 28 0-1 2-1 0.121 27 0.456 36
19 South Carolina 4-3 0.158 15 0-1 2-3 0.052 45 0.264 3
20 South Florida 5-1 0.131 22 0-0 0-1 0.197 16 0.663 78
21 Ball State 6-0 0.122 17 0-0 0-0 0.332 6 0.901 118
22 Iowa 4-3 0.119 24 0-0 0-3 0.158 18 0.561 57
23 Boise State 5-0 0.117 21 0-0 0-0 0.259 12 0.890 116
24 Tulsa 6-0 0.114 37 0-0 0-0 0.374 4 0.913 120
25 LSU 4-1 0.109 55 0-1 1-1 0.087 30 0.428 34

The Week 8 FEI Ratings for all 120 FBS teams can be found here. Expanded FEI Ratings data can be found here.

Offensive and Defensive Efficiency

As my FO colleague Bill Connelly has addressed this season in his weekly Varsity Numbers columns, traditional stats in college football can be particularly misleading if not examined in the proper context. The NCAA ranks each team in "Total Offense" and "Total Defense" simply by yards earned and allowed per game. If every team began every possession from their own 30-yard line, played the same number of possessions per game as everyone else, never played a garbage-time possession, and faced the same opponent as every other team on every down, then yards per game would be a fine way to rank offenses and defenses.

FEI was built as a drive-based system for two main reasons. The first was practical: College football play-by-play data was too massive to collect, too unreliable, or in some cases several years back, too unavailable. The second reason was philosophical: Drives are essential and fundamental in football, a game played in alternating possessions that gives teams equal opportunity to score and be scored upon. Game Efficiency, therefore, is the starting point for FEI, a measure of the collective success of a team maximizing its own possessions and minimizing those of its opponent. But what about breaking that success down by unit efficiency standards? We'll start with offense and defense. We'll take a closer look at special teams and field position efficiency next week.

To measure a team's Offensive Efficiency in each game, I first discarded clock-kill and garbage-time possessions. Second, I measured the offensive scoring expectation of a team from each drive's starting field position according to national baselines. Third, I modified the actual result of non-touchdown drives to account for national punting/kicking averages (the opponent's 12-yard line is worth 2.528 points offensively regardless of the success of the field goal attempt; the opponent's 35-yard line is worth 0.810 points offensively regardless of a field goal or punt), and credited touchdown-scoring drives with 6.958 points (the average touchdown value neutralized for average rates of point-after and two-point conversion attempts). Offensive Efficiency (OE) represents the sum of the value of offensive points earned divided by the sum of its offensive score expectations over the drives of the game, minus one (to calibrate OE as a zero-sum measurement). Defensive Efficiency (DE) is equal to the opponent's OE.

For example, on October 11, Texas defeated Oklahoma in the Red River Rivalry 45-35 in 24 total possessions. The final two possessions of the game, following Oklahoma's turnover on downs with 2:31 left in the game, were discarded as garbage-time possessions.

10/11/2008 Texas vs. Oklahoma
Stat Texas Oklahoma
Final Score 45 35
Non-Offensive Points 7 0
Offensive Possessions 10 11
Offensive Expectation 18.97 20.73
Offensive Points Earned 35.38 34.79
Offensive Efficiency 0.865 0.678
Defensive Efficiency 0.678 0.865

Both offenses played very efficiently, scoring well over national average expectations. In a vacuum, Texas' offense outplayed Oklahoma's. To date over the course of the season, Texas' average OE is second only to Tulsa's and Oklahoma ranks third. But how do we adjust these results for the level of competition faced? Is the Longhorns' defense good, bad or average?

Offenses versus Texas to date
Opponent OE vs. UT Season OE
Florida Atlantic -0.493 -0.144
UTEP -0.380 0.047
Rice -0.293 0.587
Arkansas -0.708 -0.209
Colorado -0.473 -0.258
Oklahoma 0.678 0.955
Missouri -0.051 0.744

Texas' stifling defense didn't just show up last weekend against Missouri. Every Texas opponent to date, including Oklahoma and their five-touchdown day in the Cotton Bowl, has been held far below its average Offensive Efficiency against the Longhorns. Clearly, OE and DE need to be adjusted to account for the quality of the opposition faced.

Adjusted Offensive Efficiency (AOE) and Adjusted Defensive Efficiency (ADE) are the opponent-adjusted values of OE and DE. 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.

Top Ten AOE Top Ten ADE
AOE Rank Team AOE OE OE Rank ADE Rank Team ADE DE DE Rank
1 Oklahoma 0.628 0.955 3 1 Alabama -0.486 -0.534 5
2 Texas 0.597 1.132 2 2 Florida -0.470 -0.558 2
3 Penn State 0.557 0.644 8 3 Wake Forest -0.460 -0.277 24
4 Georgia 0.528 0.152 37 4 North Carolina -0.442 -0.324 19
5 Navy 0.506 0.106 43 5 Texas -0.433 -0.259 27
6 Missouri 0.467 0.744 6 6 South Carolina -0.430 -0.349 17
7 USC 0.454 0.554 11 7 Oklahoma State -0.398 -0.082 54
8 LSU 0.402 0.122 40 8 Pittsburgh -0.391 -0.082 53
9 Nebraska 0.396 0.381 17 9 Boston College -0.371 -0.534 4
10 Georgia Tech 0.385 0.035 49 10 Ohio State -0.364 -0.439 13

Offensive and Defensive Efficiency data for all teams can be found in the FEI Expanded Ratings here.

Texas is the only team currently rated in the top ten in both AOE and ADE. Alabama is on the doorstep (No. 11 AOE, No. 1 ADE). Penn State, Florida, and USC currently rank in the top 20 in AOE and ADE. Texas' upcoming undefeated opponent Oklahoma State rates stronger in ADE than AOE, a reversal of its traditional-stat rankings (NCAA Ranks: No. 7 Total Offense, No. 47 Total Defense). Like Texas, the Cowboys held down Chase Daniel and Missouri and kept their other opponent offenses largely in check. Will Oklahoma State present the most formidable challenge yet to Colt McCoy's Burnt October Heisman campaign? Can Ohio State shut down Penn State's offensive juggernaut in Columbus, or are the Nittany Lions capable of a USC-like offensive outburst? Is the showdown between Georgia's offense and LSU's defense actually a mismatch? I'll take a closer look at past season AOE and ADE trends in the coming weeks.

Posted by: Brian Fremeau on 22 Oct 2008

10 comments, Last at 23 Oct 2008, 2:08pm by T-Diddy


by sam :: Wed, 10/22/2008 - 2:59pm

why the big jump for Florida, who didn't play?

sam! or the original sam from the old FO

by Tom Gower :: Wed, 10/22/2008 - 4:53pm

Yeah, the Florida thing is really weird-their FEI jumped from .214 to .259 with no change in GE. The factor looks like SOS-it jumped from .637, 68th, to .555, 55th. I'm guessing the primary culprit is LSU, .016, 55th last week and .109, 25th this week. But, I don't care, the fact that a team can jump from what would be 10th to 4th in one week based on what their opponents do in week 8 doesn't sit well with me at all.

by Jean Sansterre (not verified) :: Wed, 10/22/2008 - 5:40pm

Why is that? I mean, yeah, it seems weird for a team to 'get better' when they haven't played. But the alternative is sillier. Remember that year when Buffalo beat New England 31-0? At the time, it was pretty clear that New England sucked (and its players hated their coach - heh) and that Buffalo was pretty good. However, let's say that Buffalo had 16 byes the rest of the year. An objective assessment of how good Buffalo is should keep going up, because New England proceeded to beat the tar out of the rest of the league. Beating the Patriots 31-0 speaks much better of a team than beating the Raiders 31-0. And since the evaluations of the teams you've played is constantly evolving (since they keep playing new games, and giving more information about their quality) your rating will change purely on the basis of the changes in how good the teams you've played have been.

For example, Oklahoma beat TCU. They've had a pretty good year so far. But last week, TCU throttled BYU. Oklahoma's 35-10 win over TCU speaks better and better of it the more it becomes apparent that TCU is a good team.

It's an understandable intuitive fallacy. It's counter-intuitive to think that the meaning of something has changed even when the event itself hasn't. If you've eaten five eggs (including yolks) for breakfast every day for the past decade, your doctor may have told you that you were an unhealthy eater. However, based on recent research, your doctor may now tell you that you are a healthy eater, and always had been. This makes no intuitive sense, since your eating habits have not changed. This only happens if the original conclusions were flawed, and better information has become available. It's counter-intuitive, but your new conclusion is always more accurate.

Florida shelacking LSU meant that Florida was pretty good when it seemed like LSU was only average. Now that it's become apparent that LSU is actually pretty good itself, it has become correspondingly apparent that Florida is damned good to have blown out a pretty good team by 30 points. It doesn't seem intuitive to think that Florida's better because LSU is. But the alternative is to make conclusions about how good teams are with very little data, and ignore new data that arrives. Making conclusions while deliberately ignoring new data tends to compromise the conclusions themselves.

by Brian Fremeau :: Wed, 10/22/2008 - 5:14pm

Sam and NewsToTom,

Every week, the ratings add new data and recalibrate the opponent-adjustments of Game Efficiency data to date accordingly. Miami moved up from 53 to 41 with a win, Tennessee moved up from 60 to 48 with a win, and LSU moved up from 55 to 25 with a win, all solid Florida victories. And FEI still likes Mississippi. On the flip side, Notre Dame didn't play last week and dropped from 39 to 46. All six of the Irish opponents to date lost last Saturday.

If the recalibration didn't happen, a team like BYU would retain credit for beating a "good" UCLA team; or Alabama would retain credit for beating a "great" Clemson team. I think re-evaluating those outcomes is valuable.

by Tom Gower :: Wed, 10/22/2008 - 6:59pm

Oh, I fully understand and accept that. Indeed, if not for incorporating preseason FEI (your version of DAVE), I'd expect the fluctuations early in the year would be huge (think Blogpoll resume rankers). But, the additional information created by going from Week 7 to 8, when a team doesn't play is, what, 15-18% more SOS information? It's just hugely counterintuitive to me that that additional information produces that much change. I'm also remembering similar discussions last year even later in the year. Maybe I'll start running FEI with a lesser SOS adjustment like I did last year (I think).

by sam :: Thu, 10/23/2008 - 7:49am

OK. That's kind of all I wanted to know. I wasn't questioning it.

It seemed a big jump (into the top 5) and I didn't really think a 24-17 win over South Carolina was going to make LSU jump enough to make Florida look that much better. But your explanation is fine with me.

sam! or the original sam from the old FO

by abernethyj :: Wed, 10/22/2008 - 3:05pm

Thanks for throwing SOS up there this week.

by pawnking (not verified) :: Wed, 10/22/2008 - 3:21pm

Very cool, Brian. I'm always impressed with your willingness to both apply consistent methodology and to show your work on the same.

by Becephalus :: Thu, 10/23/2008 - 12:20am

Let me double down on the idea that the Florida and Notre Dame movements couldn't be more reasonable. There really is not a whole lot of information comparing teams to begin with, I think that is what you are missing. A week in which most of a teams past opponents have a good showing is a very illustrative week, particularly if many earlier games from those opponents were meaningless ones against creampuffs.

The Wire should win the Nobel prize for literature.

by T-Diddy (not verified) :: Thu, 10/23/2008 - 2:08pm

So Brian, is it possible to turn these offensive and defensive efficiency ratings into projected point totals for a given team in a given game since it is based off of expected points per drive?