FEI Week 2 Ratings and Fumble Recovery Value

by Brian Fremeau
In one of the most thrilling games played thus far in 2018, Clemson fended off an upset bid on Saturday night against Texas A&M to win by a final score of 28-26. With only 14 seconds remaining in the second half, the Aggies scored on a 14-yard touchdown pass from Kellen Mond to Kendrick Rogers to pull within striking distance. But Mond's pass on the ensuing two-point conversion attempt was intercepted. Clemson escaped with a victory, and according to regular season projected win likelihoods the rest of the way, the Tigers held onto the nation's clearest path back to the College Football Playoff.
Texas A&M nearly scored on its penultimate possession as well. A five-play, 85-yard drive inside four minutes to play in the game ended in somewhat controversial fashion when Mond connected with Quartney Davis for a potential touchdown , thwarted by a fumble forced by K'von Wallace as Davis lunged for the goal line. The ruling on the field, unable to be reversed on replay, was a fumble through the end zone and a touchback for Clemson.
How costly was the fumble at the pylon for Texas A&M? On the one hand, the Aggies were able to force a three-and-out, get the ball back, and drive for the potential game-tying touchdown on their next possession. There's no telling whether Texas A&M would have had more success on a two-point attempt one series earlier, and there's no telling how the final few minutes of the game would have played out differently had the Aggies tied it up and forced Clemson to play more aggressively on offense late in the game. But ignoring all of that, we can measure the impact of fumbles based on the position on the field they occur, and it turns out A&M's goal-line fumble was as costly as they come.
From 2007 to 2017, there were a total of 9,008 non-garbage offensive drives that ended in a fumble and were followed by a non-garbage possession by the opponent. Defenses returned 702 of those fumble recoveries (7.8 percent) for a touchdown on the play. Defenses fumbled back to the offense on the play on 23 occasions in that span. On three occasions, the offensive fumble resulted in a recovery by the defense, retreat into their own end zone, and a safety. All other situations resulted in a defensive fumble recovery followed by an offensive possession for the recovering team.
When an offense fumbles the ball away to the defense, there are three component values in that event to consider:
- the potential scoring value based on starting field position of the offensive drive lost as a result of a stop,
- the additional scoring value lost by the offense if it moved into scoring range on the drive,
- and the resulting field position value granted to the opposing team that recovers the fumble.
Consider a fumble at the offense's own 1-yard line on a drive that starts from that position. The starting field position forfeited by the stop is worth only 0.96 points, and the offense failing to move into scoring range on the drive doesn't result in any other offensive value forfeited. But on average, the resulting field position for the opponent recovering the fumble in this situation is worth 6.89 points. In fact, 88 percent of these situations in the last 11 years resulted in the defense scoring a touchdown on the fumble recovery. All told, then, a fumble at a team's own 1-yard line on a drive that starts on the team's own 1-yard line produces an average fumble recovery value of 7.85 points (0.96 + 6.89 = 7.85).
Starting field position value is important when considering the total value lost on the possession, but we can also extract out the value of the fumble event alone based on the value of the offensive field position where the fumble occurs and the average resulting value generated by the opposing defense that recovers it. On all fumbles that occur at the offense's own 1-yard line, the average total value generated by the defense recovering the fumble is 6.78 points. That's a costly area of the field to fumble, but it isn't as costly as coughing it up as a fumble at the goal line as the offense is heading in for a score.

The spike in the chart at the goal line is a result of the forfeited value of a touchdown (6.96 points on average) plus the resulting field position for the team that recovers the fumble. Of the 226 fumbles at the goal line recorded from 2007 to 2017, 209 (92.5 percent) resulted in a touchback. Fourteen of the goal-line fumbles were returned by the defense to an average starting field position at the 30-yard line. Three were returned 100 yards the other way for a touchdown.
In total, the average total value generated on the turnover event at the goal line is 8.72 points -- that is, a goal-line fumble like Texas A&M had against Clemson with just over two minutes left in the game is equivalent to an 8.72-point swing. Had they merely fumbled the ball away at the 1-yard line instead, it would have been half as costly -- only 4.36 points in value are exchanged on average on fumbles at that point on the field.
FEI Week 2 Ratings
The Fremeau Efficiency Index (FEI) is a college football rating system based on opponent-adjusted possession efficiency. Preseason projections (86 percent weight in this week's ratings) are based on five-year results, recruiting success, and returning offensive and defensive production. Strength of Schedule ratings (SOS) represent the average number of losses an elite team (two standard deviations better than average) would have against the team's regular season schedule. Strength of Schedule ratings against opponents played to date (PSOS) are also calculated from the perspective of an elite team.
Click here for ratings for all 130 FBS teams.
Rk | Team | Rec | FEI | SOS | Rk | PSOS | Rk |
1 | Alabama | 2-0 | .274 | 1.10 | 46 | .14 | 46 |
2 | Clemson | 1-0 | .259 | .94 | 59 | .16 | 41 |
3 | Auburn | 1-0 | .257 | 2.16 | 2 | .68 | 5 |
4 | Georgia | 1-0 | .255 | .94 | 58 | .11 | 50 |
5 | Ohio State | 2-0 | .254 | 1.04 | 55 | .02 | 96 |
6 | Notre Dame | 2-0 | .223 | 1.17 | 38 | .16 | 42 |
7 | Oklahoma | 2-0 | .216 | .78 | 67 | .04 | 81 |
8 | Washington | 0-1 | .211 | 1.27 | 30 | .47 | 11 |
9 | Stanford | 2-0 | .185 | 1.56 | 12 | .17 | 40 |
10 | Michigan State | 1-1 | .182 | 1.10 | 45 | .17 | 37 |
11 | Wisconsin | 2-0 | .179 | .95 | 57 | .03 | 89 |
12 | Mississippi State | 1-0 | .162 | 1.62 | 10 | .10 | 54 |
13 | Michigan | 1-1 | .157 | 1.84 | 5 | .47 | 12 |
14 | Penn State | 2-0 | .149 | 1.23 | 33 | .06 | 65 |
15 | USC | 1-1 | .149 | 1.23 | 35 | .36 | 19 |
Rk | Team | Rec | FEI | SOS | Rk | PSOS | Rk |
16 | Florida State | 0-1 | .139 | 1.55 | 14 | .10 | 55 |
17 | Miami | 0-1 | .134 | .91 | 61 | .17 | 38 |
18 | LSU | 1-0 | .130 | 2.07 | 3 | .17 | 35 |
19 | Virginia Tech | 1-0 | .122 | 1.02 | 56 | .24 | 26 |
20 | Boise State | 2-0 | .115 | .32 | 108 | .05 | 75 |
21 | Iowa | 2-0 | .112 | .68 | 80 | .05 | 72 |
22 | Duke | 2-0 | .109 | 1.30 | 28 | .11 | 52 |
23 | Central Florida | 1-0 | .109 | .25 | 119 | .02 | 100 |
24 | TCU | 1-0 | .109 | 1.31 | 27 | .01 | 107 |
25 | Oklahoma State | 1-0 | .103 | 1.13 | 43 | .01 | 121 |
26 | Texas | 1-1 | .100 | 1.07 | 52 | .03 | 85 |
27 | Texas A&M | 0-1 | .096 | 2.20 | 1 | .37 | 16 |
28 | Boston College | 1-0 | .090 | 1.27 | 31 | .02 | 95 |
29 | Arizona State | 2-0 | .090 | 1.49 | 18 | .21 | 27 |
30 | Oregon | 1-0 | .088 | .93 | 60 | .01 | 110 |
Comments
3 comments, Last at 18 Sep 2018, 10:19pm
#1 by nottom // Sep 12, 2018 - 2:37pm
The goal line fumble analysis seems flawed to me. You can't assume the offense would have scored a TD, if they hold on to the ball they will often just be down at the 1. I might expect a small spike due to a higher percentage of touchbacks giving better field position than a recovery at the 1 or 2, but a 4 point spike just doesn't make any any sense.
#2 by Brian Fremeau // Sep 12, 2018 - 5:26pm
I considered this, and I admit that there is a blind spot in the data that I can't fully overcome. There have been situations in which a ball carrier simply fumbles as he is headed in for a score, and a touchdown was literally forfeited on the play. And there are situations like the one in the Texas A&M vs Clemson game in which the ruling on the field could very well have resulted in a 1st-and-goal situation for the Aggies. The spike in the data/chart implies that the value of the fumble at the goal line is like taking a touchdown off the board, but it may be better as you suggest to consider all such fumbles at the goal line as equivalent to a fumble at the 1-yard line instead.
#3 by Brian Fremeau // Sep 18, 2018 - 10:19pm
There have been two fumbles recorded at the goal line through Week 3 in FBS non-garbage drives. The first was the Texas A&M play discussed here against Clemson in Week 2. The second is this Vanderbilt fumble sequence in Week 3: https://www.sbnation.com/college-football/2018/9/15/17860882/vanderbilt-notre-dame-2018-results-touchback
In this case, the strip/fumble occurs at the 1-yard line, but in the chaos, the ball is "advanced" into the end zone, nearly recovered for a touchdown by Vanderbilt, but ultimately recovered for a touchback by Notre Dame. Unlike the A&M game in which the best case for the Aggies would have been first-and-goal at the 1-yard line, the Commodores' best case would be a fumble recovery for a touchdown.
I'll try to pay close attention to these instances over the course of the season.