No. 16 Seeds Are Due*

It’s the sort of statistic that seems ripped from pages of the Washington Generals media guide. Since the men’s NCAA Tournament went to a 64-team format in 1985, No. 16 seeds are winless: an imperfect 0-124 record.

No. 16 seeds can be pretty bad basketball teams, of course. Often, they’re teams from small conferences that won automatic bids by winning their conference tournament in a series of upsets after having barely cleared .500 during the regular season. (Small-conference teams that win both the regular season and their conference tournaments will usually wind up with No. 13, 14 or 15 seeds instead.) Furthermore, No. 16s have the misfortune of being matched up against No. 1 seeds, which are theoretically the four best teams in the country.

But being bad is one thing; going 0 for 124 is another. My hunch is that No. 16 seeds have been unlucky not to have pulled off at least one upset.

Consider that in the 1998 women’s NCAA Tournament, No. 16 seed Harvard (those plucky upstarts) beat No. 1 seed Stanford. And in the men’s tournament, several No. 16 seeds have come close to winning. Two of them, Princeton and East Tennessee State, lost by a single point in 1989 to Georgetown and Oklahoma, respectively. The next year, No. 16 seed Murray State took Michigan State to overtime before losing by four.

Meanwhile, plenty of No. 15 seeds have won. Well, not plenty, but seven of them have upset No. 2 seeds. And No. 14 seeds have beaten No. 3 seeds 20 times. These results suggest that 16-versus-1 upsets ought to be possible, especially because there isn’t always a whole lot of daylight separating teams from one seed to the next. As its choices this year made clear, the selection committee is not infallible. Sometimes a team gets seeded as a 16 when it should probably have been a 14 or 15. Sometimes a No. 1 seed should have been a No. 2 seed. If a No. 15 seed can beat a No. 2 seed, then surely a 15-seed that’s mis-seeded as 16 can beat a 2-seed that’s mis-seeded as a No. 1.

But we can be more precise about this. In building our NCAA Tournament forecasts this year, we developed an Elo ratings system for college basketball. Although our forecasts for this year blend Elo with several other computer ratings, we can run Elo-based projections for past tournament games going back as far as we like.

Here, then, is how Elo would have forecast every past 1-versus-16 matchup on the day it was played. Once we figure out the No. 1 seed’s odds of winning each game, we can simply multiply the probabilities to figure out their cumulative odds of winning all 124. It’s a long table, so scroll down to the bottom for the punch line.

This data suggests that No. 16 seeds have in fact been pretty unlucky. On average, Elo would have given the No. 1 seed a 97.6 percent chance of winning each individual game; the range runs from 99.9 percent (Duke against Florida A&M in 1999) to 87.1 percent (Memphis against Oral Roberts in 2006). But given 124 chances to pull a rabbit out of their hats, No. 16 seeds “should” have come away with about three victories, according to Elo. Furthermore, the probability of them having gone winless is only about 5 percent. I wouldn’t call the No. 16s phenomenally unlucky — we’re talking about odds of about 20-to-1 against, not 20,000-to-1 against — but this confirms my intuition that they haven’t caught very many breaks.

The good news for No. 16 seeds is that their situation has been improving very slightly. Since the tournament introduced its play-in game in 2001, the average No. 16 to play a No. 1 had an Elo rating of 1504; before that, their average rating was 1470. The play-in games are helpful to the cause of the No. 16 seeds in two ways. First, the truly execrable No. 16s, like Florida A&M in 1999 (which came into the tournament with a 12-18 record in the Mid-Eastern Athletic Conference), will be routed into the play-in game and will usually lose it instead of wasting one of the 16-seeds’ four opportunities. Second, the play-in winners will have a game of NCAA Tournament experience under their belts. That helps both in real life and for a team’s Elo rating, since Elo weights recent games (and especially recent tournament games) more heavily.

So cheer up, Holy Cross, Hampton, Florida Gulf Coast University and Austin Peay State. Yes, you’re probably going to lose by 30 points. But sooner or later, one of you is going to make history.

Check out FiveThirtyEight’s 2016 March Madness Predictions.

Our sports podcast Hot Takedown previews March Madness.