The calendar has turned to November, and by the end of the upcoming weekend nearly all FBS teams and conferences will have successfully played at least one game. We're still grappling with disruptions of one kind or another, but the 2020 season continues to press forward. Pac-12 and MAC conference teams will jump into the ring this week, and though they'll have some catching up to do in some perception circles, they'll each have played as many games as Wisconsin has; the Badgers have canceled their second straight game while dealing an coronavirus outbreak in their program.
The Badgers dominated Illinois two weeks ago, and based on that very limited sample they have a profile of a team that should contend for, if not run away with, the Big Ten West division. But running away with anything may be an impossible task in a truncated season. Wisconsin can't afford to have any other games canceled, lest it not meet the six-game minimum requirement for Big Ten conference championship eligibility. Far less consequentially for their postseason goals, but just as importantly to my numbers, a six-game slate won't be easily compared with teams that play eight games, ten games, or more.
I haven't specifically published my methodology for factoring in and progressively phasing out preseason data over the course of the season, but I have referenced that it varies by team based on the number of games played. BYU has seven FBS games under its belt through Week 9, and its preseason projection data -- which very modestly pegged the Cougars as a middle-of-the-pack team before they kicked Navy up and down the field in September -- is still a 22% factor in their FEI ratings. Is that too much? Should preseason data at this point in the year mean nothing for BYU or others that have played at least seven games? Their showdown with Boise State this Friday will be among the most important data points we'll collect for the Cougars this year, but I still plan on keeping preseason ratings in the mix to some degree.
Preseason projections and the phase-out process are a big reason why LSU hasn't fallen further as well. To be sure, the Tigers haven't performed anything like their championship season in 2019, and it's very possible with a tough schedule remaining that they'll keep skidding every week for the remainder of the year. But there's a balance that must be struck as well. It isn't that preseason projections are infallible, it's that we simply don't have enough new information early in the year, so we rely on priors that have proven overall to be effective. For situations like LSU that are way off the mark, there are plenty of top teams that, thus far, appear to have performed according to expectations.
There may need to be model testing after the season is over to see if pulling the preseason projection data out faster would result in better ratings, and by that I mean ratings that produce better projected results. I ran a set of test ratings in which I pulled the projection data out more rapidly -- ratings for teams that have already played at least five opponents would be exclusively based on in-season data, and those with fewer than five opponents would carry less preseason weight than they do in the current model. BYU jumps up to No. 3 in this alternate version, and LSU drops 30 spots to No. 46. On the surface, that feels more palatable, but is it more predictive?
I've said it before, and I'll repeat it again here: everything about the FEI ratings in 2020 feels like an experiment. Our control subjects -- Alabama, Ohio State, and Clemson -- appear to be stable and reliable, but it's hard to get a good read on anyone else. We need more games played and fewer games canceled, of course, and I'm not planning on throwing out all the data at year's end just because of the peculiarities of the season. But it is odd, and we'll still be seeing some teams for the first time this month.
2020 FEI Ratings (through Week 9)
FEI ratings (FEI) represent the per-possession scoring advantage a team would be expected to have on a neutral field against an average opponent. Offense ratings (OFEI) and defense ratings (DFEI) represent the per-possession scoring advantages for each team unit against an average opponent unit. FEI, OFEI, and DFEI ratings are based on a combination of opponent-adjusted results to date and preseason projections.
Net points per drive (NPD) is the difference between points scored per offensive drive and points allowed per opponent offensive drive. Net available yards percentage (NAY) is the difference between offensive available yards percentage and opponent offensive available yards percentage. Net yards per play (NPP) is the difference between drive yards per offensive play and drive yards allowed per opponent offensive play. Three different schedule strength ratings for games played to date are provided, based on current FEI ratings, representing the expected number of losses an elite team two standard deviations better than average would have against the given team's schedule (ELS), the expected number of losses a good team one standard deviation above average would have against the schedule (GLS), and the expected number of losses an average team would have against the schedule (ALS).
Ratings and supporting data are calculated from the results of non-garbage possessions in FBS vs. FBS games.