The NCAA men’s basketball tournament – popularly known as “March Madness” – begins next week. Millions of college basketball fans around the world will hope that their bracket will predict all the games correctly. In an interview with News Bureau physical sciences editor Liz Ahlberg, computer science professor Sheldon H. Jacobson discussed the statistics of bracketology and how the math can be used to predict each round of the tournament. Working with U. of I. students, he also oversees the BracketOdds website to provide fans with some help in filling out their brackets.
How can computer science and statistics make any sense out of the unpredictability of March Madness?
Bracketology is an exercise in data, analytics and probability. Since 1985, a No. 1 seed has never lost to a No. 16 seed in the round of 64. However, out of 124 games since 1985, a No. 2 seed has lost to a No. 15 seed seven times in the round of 64. The probability of a No. 15 seed pulling such an upset is very small, yet it does happen. That is how the power of probability expresses itself in the tournament. Extreme upsets are rare events but occur with regularity, given a sufficient number of opportunities.
Are you saying that predicting upsets is easy?
Yes and no. Yes, it is easy to predict that in the round of 64, between one and three teams seeded 13, 14 or 15 are highly likely to pull an upset. What we cannot say with certainty is which of those teams will pull the upset.
How much stock should people put in seeds when filling out their brackets? Upsets do happen, as you’ve said, but isn’t it still the safest choice to always pick the higher seed to win?
Picking the higher seed may work, but it depends on the round. For example, over the past 31 tournaments, an average of nearly three out of the eight teams seeded 11 or 12 advance in the round of 64. Moreover, in 27 of the past 31 tournaments, there has been at least one team seeded 11 or worse that has reached the Sweet Sixteen. On the other hand, during this period, more than 70 percent of Final Four teams have been seeded 1, 2 or 3. What this means is that seeds matter, particularly as we advance further along in the tournament.
How does your Bracketodds website incorporate statistics to help people fill out their brackets?
We like to think of our website as STEM (science, technology, engineering and math) in action. We launched the site in 2012, and in 2015 we launched a bracket generator so that people unfamiliar with the tournament can have a bracket generated for them that contains the right mix of upsets and favorites advancing.
This year, we introduced our picks for upsets amongst teams seeded 13, 14 or 15 in the round of 64. This tool is based on historical upsets using an advanced analytics technique. Our predictions for upsets will be available Monday morning, after the teams are announced on Selection Sunday.
Can you give me a glimpse of how you fill out your own bracket?
I do not fill out a bracket. The best suggestion for people who want to fill out a bracket is to fill in which seeds you want to advance and, once the teams are announced, fill in the teams based on the seeds you chose. History suggests that you will be no worse – or better – than trying to do it based on knowledge of the teams.