NCAA tournaments are getting measurably wilder

Last night, the #1 seed Kansas Jayhawks won the NCAA men’s basketball tournament. Maybe an unsurprising outcome… but they defeated #8 seed North Carolina in the championship game. Only four #8 seeds have made it that far in the nearly four decades of the current format.

This year’s tournament also saw #15 seed Saint Peter’s advance to their regional final, the first #15 seed ever to do so.

And there were plenty of other unlikely upsets as well. Considering the tournament as a whole, that is, all 63 games (ignoring the abomination of the four play-in games) and their seeding match-ups, how “wild” was this year’s outcome compared to past tournaments?

Or, put another way, what was the prior probability of picking a “perfect bracket,” guessing every game outcome correctly, and how does that probability compare with past years?

This question has been discussed here before; I won’t rehash the details of this past article that describes the details of methodology for modeling game probabilities. This post is really just bringing that analysis up to date to reflect recent tournaments (all source historical tournament data is available on GitHub). Here are the results:

Probability of a perfect bracket, 1985-2022.

The constant black line at the bottom reflects the 1 in , or 1 in 9.2 quintillion probability of guessing any year’s outcome correctly, if you simply flip a fair coin to determine the winner of each game. The constant blue and red lines at the top are for comparison with the actual outcomes, indicating the probability of a “chalk” bracket outcome, always picking the higher seeded team to win each match-up. (The blue and red colors reflect two different models of game probabilities as a function of difference in “strength” of each team; as discussed in the previous article, blue indicates a strength following a normal distribution density, and red indicates a linear strength function.)

By either the normal or linear models, three of the last four tournaments (2018, 2021, and 2022– 2019 was pretty by the book, and 2020 was, well, 2020) were among the top five “wildest,” most unlikely outcomes in the last four decades.