FEI Week 16 Ratings

The final game of the college football regular season resulted in a solid 31-7 victory for Navy over Army on Saturday. It was the first victory by the Midshipmen in the annual rivalry since 2015, and the first game in the rivalry that was decided by more than a single score since 2013. Army had a very disappointing season this year, opening with a preseason FEI projection of No. 30 overall (and a 4% likelihood to run the table to an undefeated regular season) and giving Michigan an overtime scare in Week 2, but dropping seven losses in their last eight games against FBS opponents to finish the year. Navy, on the other hand, had a very solid season -- 9-2 against FBS opponents, with both losses coming against teams (Memphis and Notre Dame) that finished among the top 15 in the final Associated Press top-25 rankings.

Navy's loss to Memphis was competitive -- the Midshipmen led the eventual AAC champions at the half -- but their loss to the Fighting Irish on November 16 was decidedly not. Notre Dame scored on each of its first-half possessions and nearly blanked Navy over the same span, opening up a 38-3 lead at the half. The Irish forced a punt and tacked on another touchdown to start the second half, ushering in garbage time with a dominant 45-3 lead amassed over only 15 non-garbage game possessions.

Blowout victories are rarely as efficient as this one was, and even more rarely do efficient blowouts come against otherwise decent competition. If we calculate game efficiency as the non-garbage margin of victory divided by the number of non-garbage game possessions, we can compare Notre Dame's raw efficiency victory over Navy on November 16 with every other single-game performance in FBS games this season. There have been a total of 734 FBS vs. FBS games thus far in 2019, meaning that there have been a total of 1,468 single-game performances. The most dominant in terms of raw game efficiency are listed in the table below. Notre Dame's victory over Navy ranks fifth (out of 1,468) by this measure.

Note the opponent FEI ratings of the teams that suffered these defeats this year. Half of the list is comprised of teams ranked in the 100s in FEI, and all but Navy rank 60th or worse. If we expanded our list to the top 50 most dominant game efficiency victories of the year, none of the others came against an opponent ranked among the top 30 in FEI.

Game efficiency in and of itself is prone to exaggerated results at its extremes. Wisconsin crushed Kent State 41-0 in only 12 non-garbage possessions, the most ruthlessly efficient victory of the season, resulting in a game efficiency rating of 3.42 (41 divided by 12). Baylor crushed Kansas 41-0 in 24 non-garbage possessions this year, resulting in a game efficiency of 1.71 (41 divided by 24). Is it accurate to say that Wisconsin was twice as efficient in its victory over Kent State as Baylor was in its victory over Kansas? I think that's fair, but I'm not so sure it is fair to conclude that Wisconsin's win was twice as impressive. Though Wisconsin's offense was effectively twice as efficient as Baylor's offense, an argument could be made that Baylor's defense was significantly more impressive than Wisconsin's, pitching a shutout in non-garbage time despite its opponent having twice as many offensive opportunities.

How might we go about measuring the most dominant single-game performances of the year, considering the exaggerated extremes of game efficiency data and also taking into account the strength of the opponents faced? There are several ways to approach this question, and there is not necessarily a right answer. I've taken two approaches and have posted the results on my site this week, under a new page called Game Ratings.

Every team's raw game efficiency data (GE) in every game is listed there, including the overall rank of that performance and the percentile rating of the performance in comparison to all other game results this season. The adjusted game efficiency data (AGE) most closely represents the per-game component data that formulates the FEI ratings, extracting only the offensive, defensive, and field goal game splits results for each game and comparing that data to the average splits surrendered by its opponents. (Note that FEI is actually a computation not of the per-game data, but of per-possession data, so extraordinary outlier efficiency games are muted somewhat in the overall team ratings). According to AGE, Notre Dame's victory over Navy ranks as the No.1 opponent-adjusted single-game efficiency performance of the year.

The top of the AGE list is still comprised of blowout victories over relatively poor opponents, again because of the extreme outliers of exceptionally efficient victories. That isn't necessarily wrong, but the notion of "best opponent-adjusted game efficiency" points me to think about strong wins over much stronger opponents. I created the "game ratings" (GR) data to more closely approximate that notion, using the percentile game efficiency performance as the baseline, then adjusting according to the overall FEI rating of the opponent.

GR data still spits out Notre Dame over Navy as the best opponent-adjusted game of the season, with a 2.62 rating (equivalent to 2.62 points per possession in the game better than an average opponent). The second-best opponent-adjusted game of the season (and virtually tied for first with a 2.62 rating) is LSU over Georgia in the SEC Championship Game. LSU crushed the Bulldogs 34-3 over 17 non-garbage possessions. That 1.82 GE ranked 81st overall this season (.945 percentile raw efficiency performance). It was great, but it wasn't an outlier in terms of extraordinary per-possession efficiency like Wisconsin over Kent State. But it came against what is now ranked as the eighth-best team in college football, effectively a 95th percentile performance against a 95th percentile opponent. Notre Dame had a 99th-percentile performance against an 84th-percentile opponent. GR, as currently constructed, judges each of these opponent-adjusted single-game performances to be about equal.

As I mentioned earlier, there is no right answer to the question. I could put even more weight on opponent strength to swing GR more in favor of a result like LSU against Georgia, or more weight on absolute dominant efficiency to swing GR more in favor of a result like Notre Dame against Navy. I'll be retroactively running GR numbers for previous seasons to add more context to this conversation, but would love to hear from others as to how they would best credit a single-game performance. I'll also run a new set of final GR numbers after bowls and playoff games wrap up in January.

2019 FEI Ratings (through Week 16)

The Fremeau Efficiency Index (FEI) is a college football rating system based on opponent-adjusted possession efficiency, representing the per possession scoring advantage a team would be expected to have on a neutral field against an average opponent. Unadjusted possession efficiency (PE) is calculated as a function of offensive, defensive, and special teams game splits. Schedule strength is represented by each team's average per possession opponent adjustment (OA). Opponent-adjusted offense ratings (OFEI), opponent-adjusted defense ratings (DFEI), and opponent-adjusted special teams ratings (SFEI) are calculated in a similar manner as overall FEI ratings. Team records against all FBS opponents (W-L) and against opponents ranked in the FEI top 10 (v10), top 20 (v20), top 30 (v30), top 40 (v40), and top 50 (v50) are also provided.

Ratings and supporting data are calculated from the results of non-garbage possessions in FBS vs. FBS games.