Writers of Pro Football Prospectus 2008

Most Recent FO Features

CoxMar15.jpg

» Varsity Numbers: Honing in

Bill Connelly again looks at which college football teams the F/+ ratings are sure about, and which teams remain a mystery (led by Appalachian State).

30 Sep 2009

Week 4 FEI Ratings

by Brian Fremeau

One thing that distinguishes college football from the NFL is its overtime. In the pros, the first team to put points on the board wins in “sudden death." The team fortunate enough to start the extra period in possession of the ball is determined by coin flip and is often the victor.

The college football overtime involves alternating “guaranteed” possessions for each team starting at the opponent’s 25-yard line. Effectively, the outcome of the game is determined by which offense is more successful in short field position efficiency.

Each system has its advocates and detractors. Many football fans lament the fact that the NFL system rewards the team that wins the coin flip with an opportunity to drive a short distance for a game-winning field goal, never giving the opponent a chance to respond. Many others hate the fact that in college, kickoff and kick return strategy are entirely removed from the game in the overtime period(s) -- by staging a short-field shootout, the game itself becomes something other than “real” football.

I’m not sure either approach is ideal, and personally, I have a fondness in my heart for the old college system that let tied games at the end of regulation remain tied. Reflecting on the college overtime system prompted a question for me recently -- what if “real” football was more like the college overtime system?

What if we eliminated kickoffs and kick returns and gave each team the same starting field position on an equal number of possessions in every game? What if the end result of the possession was judged solely on the team’s ability to advance to the end zone or into field goal range regardless of the ability of the team’s own field goal kicking unit? And all other outcomes -- fumbles, interceptions, punts, turnovers on downs -- were treated as failures but did not have an impact on the opponent’s next possession? It wouldn’t be “real” football, no. But it would be a pure, isolated measure of the success rates of each offense versus the opposing defense. And it is precisely what is measured by Offensive Efficiency (OE) and Defensive Efficiency (DE).

Offensive Efficiency is the total drive-ending value an offense earns divided by the total drive-starting value of its field position over the course of the game. Defensive Efficiency is its counterpart, the success rate of a team limiting the opponent’s cumulative offensive drive-ending value. Offensive and defensive efficiency are calibrated as a rating above or below zero -- a good offense has a positive rating and a good defense has a negative one. If all other factors -- special teams play, turnovers creating field position, etc. -- were eliminated, how often would a pure OE versus DE football game produce a different result than the actual games played?


Year FBS vs. FBS
Games
Winning Team
Had Better OE
Pct. of Games
2007 712 622 0.874
2008 717 623 0.869
2009 188 169 0.899
Total 1617 1414 0.874

Obviously, the team that plays more efficient offensively than the opponent wins the game more often than not. However, teams that were out-dueled offensively have won 19 games this season by tipping field position and/or special teams in their favor. Texas and Michigan -- both undefeated -- have each won a game this year in which it “lost” the OE versus DE battle. Both the Longhorns (against Texas Tech) and Wolverines (against Notre Dame) had kick return touchdowns and produced better starting field position to overcome small OE deficits.

Is this exercise relevant or useful? A season ago, only one football team had a stronger OE than its opponent in every single game it played: the national champion Florida Gators. (Undefeated Utah won four times last season while running an OE deficit in the game). The Gators also happened to be the second best team in the nation in creating a Field Position Advantage (FPA), a big reason why they rated as one of the top teams of the FEI era (since 2003).

This year, 17 teams remain undefeated, but only 13 have won the OE versus DE battle in every game. Of those 13, only Florida, Iowa, and Kansas, also currently boast a top-20 FPA rating through the first part of the season. It’s still early, and as Florida experienced in a loss to Mississippi last year, there are no sure things. But a consistent offensive efficiency and field position advantage can be a lethal combination.

Projections vs. Actual Data

I have received some questions about the use of projected data throughout the beginning part of the season, and I wanted to make sure I cleared up any lingering confusion. The Game Efficiency (GE), Offensive Efficiency (OE), Defensive Efficiency (DE) and Field Position Advantage (FPA) data listed in the weekly ratings is not influenced in any way by pre-season projected data. It is raw data collected from games played to date this season, and it is not adjusted for the relative strength of the opponent.

FEI data will continue to be adjusted and partially influenced by Projected FEI data until Week 7. There are two reasons for this. First, the ratings are much more stable (and hopefully, the changes are a bit more intuitive) on a week-to-week basis by gradually reducing the projected data weight. Second, the hybrid ratings for the first six weeks of the year are, in fact, a better predictor of future game outcomes than "actual FEI" at this point in the year would produce. This is not the primary motivating factor, and certainly some results very obviously don't reflect reality. But Projected Win Expectations have been fairly accurate this year, and all of last week's "upsets" -- Penn State, Mississippi, Miami and California -- were forecasted by FEI in part due to the influence of projected data.

Week 4 FEI Top 25

The principles of the Fremeau Efficiency Index (FEI) 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 as the likelihood that an elite team would win every game on the given team's schedule.

Only games between FBS teams are considered in the FEI calculations. Since limited data is available at the beginning of the season, the ratings to date are a function of both actual games played and projected outcomes based on the 2009 Projected FEI Ratings. The weight given to projected outcomes will be reduced each week until mid-October, at which point the projections will be eliminated entirely.

NEW THIS WEEK: FEI ratings for all 120 FBS teams are now listed in the stats page section of FootballOutsiders.com. Click here for current ratings, or here for 2008 ratings. Other past seasons will be added over the next few days. We're also planning on adding stats pages with Bill Connelly's S&P metrics as well.

Rank Team FBS
W-L
FEI Last
Week
GE GE
Rank
SOS SOS
Rank
OE OE
Rank
DE DE
Rank
FPA FPA
Rank
1 Florida 3-0 0.215 1 0.403 3 0.385 79 0.459 18 -0.677 7 0.571 16
2 Virginia Tech 3-1 0.207 12 0.185 24 0.229 23 0.313 29 -0.521 16 0.516 43
3 Iowa 3-0 0.203 15 0.273 11 0.272 38 0.003 57 -0.833 1 0.579 13
4 Texas 4-0 0.201 2 0.388 4 0.319 57 0.315 28 -0.694 6 0.573 15
5 Auburn 4-0 0.199 5 0.261 12 0.319 58 0.787 5 -0.259 33 0.501 59
6 West Virginia 1-1 0.196 6 0.025 53 0.272 39 0.366 22 -0.078 50 0.429 109
7 Alabama 4-0 0.195 10 0.376 5 0.286 46 0.634 9 -0.726 5 0.521 40
8 Boise State 4-0 0.195 14 0.404 2 0.629 115 0.483 17 -0.621 8 0.574 14
9 Miami 2-1 0.185 7 -0.007 63 0.112 2 0.266 32 0.064 68 0.479 79
10 TCU 2-0 0.183 27 0.241 15 0.390 82 0.115 45 -0.794 2 0.487 72
11 Florida State 1-2 0.183 8 0.079 42 0.107 1 0.220 38 0.402 101 0.591 7
12 Oklahoma 1-1 0.182 9 0.310 9 0.131 3 0.348 25 -0.561 13 0.580 11
Rank Team FBS
W-L
FEI Last
Week
GE GE
Rank
SOS SOS
Rank
OE OE
Rank
DE DE
Rank
FPA FPA
Rank
13 Ohio State 3-1 0.173 4 0.197 21 0.300 50 0.100 48 -0.441 19 0.546 26
14 South Florida 2-0 0.158 46 0.247 14 0.254 29 0.421 19 -0.621 9 0.488 71
15 Clemson 2-2 0.156 11 0.087 38 0.203 12 -0.413 99 -0.474 18 0.564 19
16 Oregon 3-1 0.153 29 0.109 31 0.268 36 -0.031 63 -0.386 25 0.507 55
17 USC 3-1 0.147 3 0.207 19 0.210 17 0.073 51 -0.727 4 0.514 48
18 Michigan 4-0 0.140 17 0.253 13 0.264 35 0.536 11 -0.209 39 0.526 36
19 Texas Tech 1-2 0.139 20 0.138 26 0.226 22 0.391 21 -0.254 35 0.516 45
20 Notre Dame 3-1 0.137 18 0.204 20 0.338 67 1.222 2 0.020 62 0.458 100
21 Georgia Tech 2-1 0.136 19 -0.003 60 0.185 7 0.057 53 0.151 81 0.537 30
22 BYU 3-1 0.131 16 0.103 33 0.337 66 0.830 4 -0.019 54 0.433 107
23 Kansas 3-0 0.128 30 0.345 7 0.275 41 0.507 13 -0.510 17 0.590 9
24 Georgia 3-1 0.126 21 0.010 56 0.186 8 0.083 50 -0.148 44 0.429 108
25 Utah 3-1 0.123 25 0.119 28 0.311 56 -0.007 58 -0.390 23 0.520 41

Posted by: Brian Fremeau on 30 Sep 2009

12 comments, Last at 02 Oct 2009, 11:09pm by Brian Fremeau

Comments

1
by Kal :: Wed, 09/30/2009 - 6:41pm

I'm confused how the projected FEI scores predicted that Cal would lose to Oregon, given that Cal was ranked #16 by Projected FEI and Oregon #29. Is this to say that Oregon+7 was picked? What would FEI normally say the result of that game was?

8
by orang3b :: Thu, 10/01/2009 - 9:14am

Week 4 Forecasts
FEI Forecasts represent the expected Margin of Victory at the conclusion of Competitive Possessions in the game, calculated as a function of the current FEI rating of each team, the projected number of Competitive Possessions to be played, and the projected scoring pace of the game.
Oregon was projected to win 22-21 (based on Week 3 FEI).

2
by tide182 :: Wed, 09/30/2009 - 7:11pm

I guess I'm a little lost on Alabama's ranking. They're 4-0, they have the 9th-best OE rank and the 5th-best DE rank along with a 40th FPA rank (better than three teams ahead of them, including Va. Tech who they BEAT on a neutral field, badly statistically) and yet they're behind teams like Auburn, West Virginia, and Virginia Tech? The Auburn ranking, in particular, is baffling, using only the data you list. The teams have the same records, Alabama against a tougher schedule. Auburn's offense is rated slightly higher, but Alabama's defense is significantly higher rated and they also hold a fairly healthy edge in FPA. Further, I doubt anything in the preseason projections still included in your numbers would weigh too heavily in favor of Auburn (or West Virginia, though perhaps Virginia Tech) over Alabama. How, then, does Auburn get ahead of Alabama?

And I hate to be the guy that says "oh but School A romped School B head-to-head" because I get that isn't entirely the point here - and here I've gone and done it twice - but Virginia Tech got beaten by Alabama head-to-head, sports a lower rated Offense, Defense, FPA, and overall GE against a schedule superior - but not vastly so according to your metrics - to Alabama's, and yet is ranked 5 spots higher?

All I'm trying to pry out here, is what is it about Alabama that the computer doesn't like relative to these other schools? Please advise.

6
by Pat F. :: Thu, 10/01/2009 - 1:02am

I'm guessing its the preseason projection. They have Virginia Tech 4, Auburn 11, Alabama 13 (linked in article).

7
by TV_Pete (not verified) :: Thu, 10/01/2009 - 6:56am

I expect Alabama could be at or near the top of the rankings come Mid-October (assuming their current performance continues) when the preseason projections are removed. There is little doubt in my mind that Alabama-Texas-Florida are the Top 3 teams right now.

Of course, some people think that Houston is the best (most proven) team in the country, but my subjective rankings/analysis suggest otherwise.

3
by Anonymous222 (not verified) :: Wed, 09/30/2009 - 8:15pm

The team fortunate enough to start the extra period in possession of the ball is determined by coin flip and is often the victor.

Hmm... I thought the winner of the coin toss wins ~50% of the time.

Unless by "often" you mean 50%, in which case we agree.

5
by Arkaein :: Wed, 09/30/2009 - 9:28pm

I don't have the exact numbers on hand, but the trend in the last few years has heavily favored the winners of the coin flip. I think that team has won something like 60%-70% of NFL overtime games, much better than the historic trend close to 50%.

4
by navin :: Wed, 09/30/2009 - 9:26pm

I would argue that the college system is even more unfair. I thought the winner of the toss (who goes on defense first) wins the game quite often in college. I don't have the exact numbers, but I thought it was > 60%.

9
by HarrisFan1 (not verified) :: Thu, 10/01/2009 - 1:12pm

Is there something about the system that elevates ACC teams? Last year some ACC teams ended up really highly ranked in this and it's looking like that again.

12
by Brian Fremeau :: Fri, 10/02/2009 - 11:09pm

Probably still a lingering effect of preseason projections, which were rather high on several ACC teams including Virginia Tech, Clemson, Georgia Tech, Miami, Florida State and Boston College. That, and there have already been several games between those teams, whereas some other leagues haven't pitted top teams against one another yet.

10
by Corey (not verified) :: Thu, 10/01/2009 - 10:30pm

I am confused about how the GE relates to FEI and SOS. If FEI is a SOS adjusted GE, shouldn't Nebraska (#1 in GE and #30 in SOS) be much, much higher? How are teams like BYU, Utah, Notre dame and Oregon still ahead of them? are the preseason projections still a major factor?

11
by Brian Fremeau :: Fri, 10/02/2009 - 11:07pm

The pre-season projections are still a factor, yes, but note also that SOS represents a team's entire 2009 schedule, not games played to date. Some of Nebraska's toughest games lie ahead.