FEI 2021 Projections: Alabama's No. 1, Again

NCAA - The day will come again, eventually, when I will not rank Alabama at the top of my preseason projections. Today is not that day. The Crimson Tide program operates under head coach Nick Saban's mantra to "trust the process" and not to focus on outcomes, so let us do the outcome-focusing on their behalf. Alabama has won six national championships in the last 12 seasons, including three College Football Playoff championships since the four-team postseason format started in 2014. They have ranked no worse than fourth nationally in my season-ending FEI ratings every year since 2009 and have claimed the top spot eight times. This year marks the sixth straight season I have Alabama at the top of my FEI preseason projections. The last time Alabama was not No. 1 in my preseason ratings (2015), the Crimson Tide went on to win the title anyway.

The defending national champions are replacing superstars across their roster, notably at every record-breaking offensive skill position, but the blue-chip talent stepping into key starting roles at Alabama will almost certainly be stellar themselves. Nick Saban has signed a top-five recruiting class per 247Sports.com every year since 2008—often the No. 1 class, as Alabama inked this past cycle. No program has proven it can reinvent itself as well as Saban's has either. His 2011 team owns almost all of the best defensive season metrics I have measured, by a healthy margin. His 2020 team owns almost all of the best offensive season metrics I have measured, also by a healthy margin. Does defense win championships or does offense win championships? The answer really does not matter; Saban wins championships.

Click on the chart for enlarged version

Alabama leads the way once again this year, but several other contenders also perennially plant themselves atop my preseason FEI projections. Ohio State, Clemson, Oklahoma, and Georgia are college football's gold-card members vying for Alabama's platinum status. Though each has some roster question marks and scheduling landmines to navigate, it is hard to imagine any of them flailing this fall. The College Football Playoff field will not be set until December, but even before a single game kicks off it feels like three (if not four) of the seeds are already spoken for. The 2021 preseason top-five teams have claimed 21 of 28 playoff bids to date, and the concentration of power at the top of the sport remains as strong as ever.

Where it gets most interesting this season is just beyond that elite tier. Preseason forecasts are surely going to miss the mark on a number of teams, and the schools ranked from No. 6 to No. 66 (at least) are all on notice. I admit my projection model is particularly vulnerable to misfires coming out of the unusual circumstances of the 2020 season and the uncertain reliability of its data. Non-conference games disappeared last year, and wildly imbalanced schedules cast a long shadow of skepticism on team and unit data, both raw and opponent-adjusted. Alabama was certainly very good and Massachusetts was certainly very bad, but almost everything in between was and is justifiably up for some debate.

I wrestled with the "2020 junk data" question throughout the offseason. My base formula for preseason projections takes five years of opponent-adjusted efficiency data into account, weighted for the most recent results. I modeled and back-tested a number of adjustments to this approach, ultimately settling on a 2021 projection formula that gives equal weight to previous-previous season data as it does to previous season data. FEI ratings from 2016 to 2020 are all inputs this season, with 2019 and 2020 data each receiving the same extra weight. The success of the new projection model in forecasting early-season game results remains the same as the old model, and as I have done in the past, I will progressively filter out preseason numbers as new game results data comes in.

I am comfortable with the approach even as I acknowledge and accept its blind spots. I recalculated preseason projected FBS win totals for every team since 2012 based on the revised projection model and compared them with actual wins for each season. (I included all regular-season and postseason games played in this exercise.) The median delta between projected wins and actual wins for all teams was 1.5 wins per team, and the average delta was 1.9 wins. Among Power 5 teams, the average delta was 1.7 wins. Among non-Power 5 teams, the average delta was 2.0 wins. Among the 10 teams that the FEI projection model whiffed on most since 2012 (Table 1), seven were non-Power 5 teams.

Note that most of the teams in the table had a massive swing in success from the previous season. The Auburn Tigers, for example, won only two FBS games in 2012 and then turned around and won 11 the following year. The Southern Mississippi Golden Eagles won 11 FBS games in 2011 but laid a goose egg one year later. The FEI projection model will never predict these kinds of extreme whiplash results, but it will forecast most teams within range of their most probable season outcomes. The teams on the margins that play well above or below their projection ratings will reveal themselves every Saturday.

We entered last season without a data compass to guide us. This year, we have a compass in hand, but I suspect it may need a bit of recalibration. As always, I am excited to collect and share college football data and analysis here throughout the season.

2021 FEI Preseason Projections

FEI ratings (FEI) represent the per-possession scoring advantage each 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. Strength of schedule ratings represent the expected number of losses an elite team two standard deviations better than average would have against the each 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). Preseason projections are based on efficiency ratings over the previous five seasons, weighted for the most recent results.