This time last week, the College Football Playoff picture looked pretty tidy. Conveniently for the selection committee, there were four undefeated major-conference teams — Alabama, Clemson, Michigan and Washington — a perfect fit for the all-important top four slots of the CFP rankings. And although chaos is always around the corner in college football, the favorites seemed relatively safe going into the weekend, with each of the top four either playing at home or installed as heavy favorites. By the Elo-like college football rating system I developed with FiveThirtyEight editor-in-chief Nate Silver a few seasons ago,1 all four were more than 75 percent likely to win, with Alabama and Clemson’s chances sitting above 90 percent.
And then they played the games.
First, Associated Press No. 3 Clemson fell to Pittsburgh by a point Saturday evening, losing on a last-minute field goal. Then No. 4 Washington saw its undefeated season end against Southern California in a 26-13 home defeat later that night. And the weekend’s final big domino fell soon after, when Iowa upended No. 2 Michigan, also by a point, also on a game-winning field goal in the closing seconds. (For its part, Alabama cruised over Mississippi State, 51-3.)
Will your favorite team make the College Football Playoff? See all of our predictions for the 2016 season »
The carnage was notable for both its scale — it was the first time Nos. 2, 3 and 4 in the AP poll all lost on the same day since 1985 — and how unexpected it was. To the latter point, it wasn’t as though Saturday’s slain giants were playing other dominant teams; Pitt and Iowa were mediocre teams according to Elo, and although USC graded fairly well, Elo still had them as 11-point underdogs at Washington. These upsets were difficult to see coming. According to Elo, there was just a 0.5 percent chance of three of the top four teams losing.
Compare that to Oct. 19, 1985, the last time this particular combination of highly ranked teams lost: No. 2 Michigan fell on that date, but they were playing No. 1 Iowa — an upset wasn’t a total shock. And although No. 4 Arkansas slipped against good-but-not-great Texas, No. 3 Oklahoma fell to a Miami Hurricanes squad with a higher Elo; their loss was only an upset because they were at home.
So not all weekends with highly ranked teams losing are created equal. To get a better sense of where this past weekend ranks historically, I went back to 1950 (the first season the AP released a preseason poll, meaning there were rankings for each week of the season) and tracked Elo upsets for AP top-five and top-10 teams. For each game, I calculated an “upset score,” which was equal to the underdog’s pregame win probability (in terms of percentage points), according to Elo. The biggest upset weeks were the weeks that featured the largest cumulative upset score for highly ranked teams.
And according to this method, this past weekend ranks third among the most upset-laden weeks for AP top-10 teams, and fourth for top-five teams, since 1950:
|YEAR||WK||TEAM||RK||WIN PROB.||TEAM||RK||TOP 5 UPSET TOTAL|
|Okla. St.||2||84||Iowa St.||—|
|W. Virginia||5||63||South Florida||18|
|2002||11||Notre Dame||4||81||Boston College||—||201|
|1983||4||Notre Dame||4||86||Michigan St.||—||196|
As shocking as this weekend was, there were weeks in the past that saw even more chaos atop the AP rankings. On Nov. 3, 1990, four of the AP’s top five lost on the same day, and only one of those games was close.
As for how this affects the playoff picture, I’ll have more on that in my weekly column on Wednesday. You can see how our current CFP forecast model looks, though the release of the latest committee rankings will likely shake things up even more Tuesday night. But for now, know that there have rarely been college football weekends as crazy as the one we just witnessed. Right when things were looking like they might work out neatly for the committee, they’re going to have a lot of interesting quandaries to address over the next handful of weeks.
Check out our college football predictions.