Writers of Pro Football Prospectus 2008

Most Recent FO Features


» 2017 Defeats

The Cardinals had a winning record with backup quarterbacks last year thanks in large part to their high-profile edge rusher who terrorized opposing offenses. We look at defeat leaders for every position, as well as overall leaders over the past few seasons.

09 Sep 2015

FEI Week 1: Game Splits

by Brian Fremeau

The first weekend of the college football season produced fewer big surprises than usual, and the updated FEI ratings this week don't look much different than they did a week ago as a result. FEI game projections picked 35 out of 39 game winners correctly, including a 3-1 straight-up record in games in which the FEI ratings projected an underdog to win outright. That's a nice start and one that likely will not be seen again soon. College football promises to deliver a more turbulent and unexpected weekend of game results at some point in the coming weeks, and there will be bigger shake-ups in these ratings when that happens.

This week is as good a time as ever to dig into the details of my data analysis approach through an examination of what I call game splits, the single-game components of margin of victory or defeat. This is the unadjusted data that does not take opponent strength into consideration but instead establishes a baseline set of expectations for offensive, defensive, and special teams efficiency based on national averages. Let's use Ohio State's victory over Virginia Tech on Monday night and split up the drive data in that win.

Non-Garbage Time Drives

The first step is to identify which possessions are included in the non-garbage portion of the game. In total, the Buckeyes and Hokies combined to start 28 total offensive possessions, and there was one other possession opportunity lost by Ohio State on a fumbled punt return in the first half. I start by eliminating the final possession of the first half, a one-play drive that killed the final 15 seconds from the game clock and gained 9 yards but was not contested in the same manner as the rest of the first-half possessions. As a rule, one-play and two-play drives at the end of the first half qualify as garbage according to my methodology unless they pose a legitimate scoring threat.

In the second half, I calculate the start of garbage time retroactively based on the scoring margin at the conclusion of each possession and the remaining number of possessions to be played. On Monday night, Ohio State took a 42-17 lead with 10:43 left in the fourth quarter that put the chance for a Virginia Tech comeback victory out of reach. The Hokies had trailed by 18 points before that last score pushed it to 25 points and would possess the ball only two more times before the end of the game.

The basic calculation for the start of garbage time in my methodology is when the scoring margin exceeds eight times the number of remaining team possessions plus one. In other words, had Virginia Tech not given up the final touchdown, they would have had at least two chances to score and threaten to bring the game back to a single-possession lead. Once Ohio State tagged on that final touchdown, however, two more scoring chances would not have brought the Hokies back to within a single possession. That's my definition of the start of garbage time.

Removing these possessions from consideration leaves a total of 24 non-garbage possessions -- 12 offensive possessions for Virginia Tech, 11 for Ohio State, and one zero-play possession with the special teams turnover for the Buckeyes. These 24 possessions resulted in a 42-17 non-garbage final score for the game.

Scoring Margin Value

Ohio State put 42 points on the scoreboard and gave up 17 points but that isn't a very precise record of the performance of their offensive, defensive, and special teams units in the game. Each of those units contributes some value to the scoring margin on every possession. Special teams or defensive efficiency may position the offense for success, or the offense may earn the lion's share of credit by digging its way out from deep in its own territory. The value of field position is the key to game splits, and this chart is the key to measuring the value of field position.

Points Per Possession 2007-2014

A drive that starts on a team's own 20-yard line is less valuable than one that starts on the opponent's 20-yard line, and so on. The chart and its underlying data also indicate that there is a scoring expectation above zero on every drive, regardless of where it begins. On average, possessing the football on offense is worth 2.1 points. That's the first baseline in the game splits data and one that is often forgotten about. When there are an odd number for possessions in a game, the team with an extra possession has an extra 2.1 points of value in their ledger that contributes to the overall scoring margin of the game.

Let's look at the first two drives of Monday night's game as an example. Virginia Tech received the opening kickoff and started the drive on its own 25-yard line. That position on the field has a value of 1.8 points. That is 0.3 points less than the average possession value of 2.1 points. The touchback on the kickoff cost Virginia Tech 0.3 points of special teams value before the first play from scrimmage. Ohio State's kickoff reduced its 2.1-point deficit to only 1.8 points by kicking it into the end zone.

The Hokies' offense went nowhere on the drive, going three-and-out before settling for a punt. Virginia Tech's offense cost itself the 1.8 points of starting field position value by not moving the ball or scoring, and Ohio State's defense earned those 1.8 points on its side of the ledger. Zero points were on the scoreboard at the conclusion of the possession, but the OSU defense and special teams had already generated value by taking the possession value away from Virginia Tech.

Virginia Tech booted a 40-yard punt and Ohio State took over on its own 36-yard line. The resulting field position was worth 2.1 points, equal to the value of the possession itself. Ohio State's offense took advantage, driving 64 yards in eight plays to the end zone. The extra point attempt was successful and the Buckeyes now led 7-0. Most of the seven-point margin belonged to Ohio State's offense (4.8 points), some to the defense (1.8 points), and the rest to the Ohio State special teams (0.4 points).

Offense, defense, and special teams values are calculated similarly over the course of the game. Each drive and possession exchange provides an opportunity for some value to be generated or forfeited, and the game splits reflect the individual unit contributions as a result. Ohio State's fumbled punt in the second quarter counts against its special teams, a loss of 3.0 points of value as a result of both the forfeited possession and the favorable field position for Virginia Tech on the Buckeyes' 38-yard line at the end of the play.

In total, Ohio State's offense earned 25.3 points, the defense earned 3.9 points, and special teams cost the Buckeyes 4.2 points. Those unit totals add up to the 25-point scoring margin at the conclusion of non-garbage time.

Other Splits

The game splits data tables also include the total field position value available to each team over the course of the game. Ohio State's starting field position value was worth 20.0 points against the Hokies, and Virginia Tech's was worth 24.5 points. That is, if both teams had average efficiency performances in the game, Virginia Tech would have defeated Ohio State by a score of approximately 24-20 based on starting field position alone. Last year, teams that held a net field position value advantage won 65.4 percent of the time.

As with field position, the game splits tables also break out turnover value generated and lost over the course of the game. Virginia Tech had the advantage in turnover value as well, even though the total turnovers in non-garbage time were equal. Ohio State turned the ball over twice in non-garbage time, worth a total of 5.7 points to Virginia Tech, while the Hokies turned it over twice, worth 4.9 points to Ohio State. Last year, teams that held a net turnover value advantage won 72.4 percent of the time.

The final columns of data identify the percentage of available yards earned by each team in the game. These are calculated as a function of the total yards earned divided by the total yards available as measured from starting field position to end zone. Ohio State earned 63.9 percent of available yards on non-garbage drives and Virginia Tech earned only 33.6 percent. Last year, teams that had an available yards advantage won 83.1 percent of the time.

I post the game splits for every FBS matchup on my site.

FEI Week 1 Ratings

The Fremeau Efficiency Index (FEI) is a college football rating system based on opponent-adjusted drive efficiency. Approximately 20,000 possessions are contested annually in FBS-vs.-FBS games. First-half clock-kills and end-of-game garbage drives and scores are filtered out. Unadjusted game efficiency (GE) is a measure of net success on non-garbage possessions, and opponent adjustments are calculated with special emphasis placed on quality performances against good teams, win or lose. Other definitions:

  • SOS: Strength of Schedule, measured as the likelihood of an elite team going undefeated against the given team's regular season schedule. Schedule strength data based on FEI ratings and calculated across other dimensions can be explored in this interactive visualization.
  • FBS MW: Mean Wins, the average number of games a team with the given FEI rating would be expected to win against its regular season schedule of FBS opponents.
  • FBS RMW: Remaining Mean Wins, the average number of games a team with the given FEI rating would be expected to win against the remainder of its regular season schedule of FBS opponents.

Preseason projections are a function of program strength and trajectory as well as transition factors including returning starters and recruiting success. The weight given to preseason projections is reduced in the FEI formula over the first half of the season until it is eliminated entirely following the results of Week 7. Preseason projection data receives 85 percent of the weight in this week's ratings. Ratings for all FBS teams can be found here.

Rank Team FBS
1 Alabama 1-0 .279 2 .357 10 .113 5 9.4 8.5
2 Ohio State 1-0 .274 1 .298 14 .461 77 11.3 10.6
3 Oregon 0-0 .246 3 - - .229 37 9.5 9.5
4 Baylor 1-0 .228 4 .400 7 .363 65 9.6 8.6
5 Georgia 1-0 .211 5 .346 13 .192 29 8.7 7.7
6 Michigan State 1-0 .193 6 .161 21 .158 18 9.8 8.9
7 TCU 1-0 .190 8 .061 31 .275 48 8.8 7.9
8 Florida State 1-0 .188 7 .531 2 .420 70 9.3 8.4
9 UCLA 1-0 .186 10 .349 11 .258 42 9.6 8.6
10 LSU 0-0 .180 9 - - .088 2 7.6 7.6
11 Mississippi 0-0 .178 11 - - .125 11 7.9 7.9
12 Notre Dame 1-0 .171 15 .421 5 .300 52 9.3 8.4
13 Arizona State 0-1 .169 12 -.182 59 .209 35 7.9 7.4
14 Georgia Tech 0-0 .166 14 - - .189 28 7.6 7.6
15 USC 1-0 .166 13 .411 6 .121 10 8.4 7.4
Rank Team FBS
16 Boise State 1-0 .163 16 .033 35 .641 110 9.8 8.9
17 Missouri 0-0 .158 17 - - .292 51 8.2 8.2
18 Stanford 0-1 .152 18 -.130 56 .234 38 8.6 7.8
19 Oklahoma 1-0 .148 19 .348 12 .195 31 8.8 7.8
20 Texas A&M 1-0 .147 25 .182 20 .107 3 6.4 6.0
21 Clemson 0-0 .141 20 - - .286 50 7.7 7.7
22 Auburn 1-0 .139 24 .095 25 .108 4 6.3 5.7
23 Virginia Tech 0-1 .132 23 -.298 65 .271 47 7.8 7.6
24 Wisconsin 0-1 .132 21 -.357 69 .313 56 9.7 9.6
25 South Carolina 1-0 .131 22 .057 32 .164 20 6.7 6.0
26 Kansas State 0-0 .120 26 - - .357 64 7.7 7.7
27 Arkansas 1-0 .120 28 .667 1 .059 1 5.9 4.9
28 Mississippi State 1-0 .117 27 .198 18 .125 12 6.0 5.1
29 Utah 1-0 .102 29 .091 28 .158 19 7.2 6.5
30 Nebraska 0-1 .094 31 -.051 46 .480 79 8.6 8.0

Posted by: Brian Fremeau on 09 Sep 2015