How Likely Was It That A 1 Seed Would Lose?

On March 16, 2018 the University of Maryland Baltimore County (UMBC) beat the University of Virginia 74-54. This was the first time in the history of the NCAA tournament that a number one seed was upset by the number sixteen seed. Prior to the UMBC win, the one seed was 135-0 when playing the sixteen seed in the NCAA Tournament.

While this upset was historic, how likely was it to happen? To understand that, we can look at spreads of games similar to the typical one-sixteen matchup, then see how often the underdog team wins the game. The idea here being if you want to know how likely it is that a 20 point favorite loses, we just need to look at a lot of games where a team was favored by 20 points and see how often the 20 point favorite loses. There is more detail on this idea in this post.

First, we need to understand the range of spreads for a typical one seed vs. sixteen seed game. A small sample will suffice here given the relatively small total population of 136 games. We will use the median* range of spreads in each year to prevent outliers from influencing the analysis.

*In the event of a half point, I rounded down to avoid quarter point median spreads.

For this analysis, we will use the 4-year median range so as to avoid outliers like the 2015 Kentucky -35 spread and the 2018 Kansas -14 spread. In the population of 517 games from the data set that fell within the range. The favorite won 506 (97.9%) games and the underdog won 11 (2.1%) games. Given the number of games played prior to the Virginia upset, statistically we’d have expected the 16 seed to have won 2.9 games in the 136 total games played.

So, using historical spread data, a reasonable argument could be made that it was about time a 16 seed won in the NCAA tournament.