One of the most common sticking points between humans and computers—the likes of SP+ and FEI—is overall record projections. Regardless of how they evaluate teams, metrics tend to be very conservative in this regard; it's fairly rare for a team to be projected as a 10-game winner, and incredibly rare for playoff contention be a virtual lock. Only one team in FEI's preseason projections had greater than 90% odds to reach double-digit wins in the regular season. Either Ohio State or Alabama, two teams that almost any human observers would consider obvious 10-win teams, finished worse than that in nearly a third of projections.
There are a number of reasons for this—the inherent uncertainty present in the preseason is perhaps the biggest—but part of the reason is a simple and unavoidable fact of probability: doing something a number of times, even if it's relatively easy, is hard. Consider a season of games which all have a 90% win probability, equivalent to a spread of over two touchdowns. Setting aside the outlier opponents such Penn State and Michigan, this is about what Ohio State's schedule looks like this season, for example. Winning any one of these games is a cakewalk ... but the odds of going undefeated, even against such a cushy schedule, are under 30%! Replace two of those games with two tougher ones with a 70% win probability and the chances at an unbeaten season fall to 17%.
These conclusions clash with our human expectations, but we see the results all the time. Half of the preseason top 10—Notre Dame, Texas A&M, Utah, Oklahoma, and Baylor—have already lost multiple games, and all have lost games in which they have seemed like an obvious favorite. The Fighting Irish's preseason bowl odds of only 99.8% may have looked low back then, but now their postseason odds are on the brink of disaster if they can't take care of Navy and Boston College. Oklahoma's 43.4% odds of winning 10 games looked dubious a couple of months ago, but that mark became unreachable as early as Week 6. Outlier results, as unlikely as they may seem in any one game, are a constant factor in college football.
There are no games between top-10 teams this week. College GameDay is visiting FCS Jackson State, and Big Noon Kickoff is attending a game with a spread of over two touchdowns. In theory, not much should shift in the playoff landscape this week—but this week, as every week, will see some 90% favorites lose and some 10% underdogs win. In theory, the likes of Ohio State, Georgia, and Tennessee should dispatch weaker opponents without much fuss, but the odds suggest that surprises will crop up, dark horses will pull off big wins, and even one of college football's most dominant teams may find itself on the brink of a stunning defeat.