The NCAA Men’s Division I Basketball Tournament attracts millions of viewers each year, many of whom like to “play along” at home or at the office by filling out tournament brackets. The odds of filling out a perfect bracket are astronomical – there are more possible brackets than stars in the galaxy – so how can a March Madness fan be confident in filling out, if not a perfect bracket, a pretty strong one? Sheldon H. Jacobson, a professor of computer science and of mathematics who has performed many data analytics studies on tournament bracketology, talked about how seeds advance through the tournament with News Bureau physical sciences editor Liz Ahlberg.
How can data analytics help with something as seemingly unpredictable as the March Madness tournament?
How the seeds advance can be modeled using advanced analytical techniques. We created two models that capture the seeds that consistently either win or lose in certain rounds. The models allow checks and balances to assess whether a given bracket is reasonable, or outrageously unusual. It does not mean that any reasonable bracket will be a winning bracket, but it does give you some sense of how much risk you are taking with a particular bracket.
That being said, 2014 was a highly unusual year. It was the fifth-rarest Final Four seed combination over the past 30 years. In fact, four of the last five tournaments have yielded relatively unusual Final Four seed combinations. We may be due for more predictable years, like a Final Four made up of 1-1-2-3 seeds.
Isn’t it pretty safe to pick the No. 1 seeds to advance?
While everyone loves to focus on the No. 1 seeds, our research suggests that the odds are high that two or three of the top seeds will not make the Final Four. All our models indicate that there is a 70 percent chance that only one or two of the No. 1 seeds will advance this far.
Despite such odds, No. 1 seeds do win the National Championship more than any other seed. No. 1 seeds are most vulnerable in the Elite Eight and National Semifinals. If they survive such games, they are often quite successful in winning the National Championship.
You created the BracketOdds website to help people determine the statistical likeliness of their brackets, based on your research. How has your most recent work improved the site?
Working with Douglas M. King, a lecturer of industrial and enterprise systems engineering, and Arash Khatibi, a graduate student in that department, we modeled combinations of seeds that advance through the tournament. We introduced an earlier model for how seeds advance in the Elite Eight, the National Semifinals, and the National Finals, and implemented it on the website back in 2012. The new models provide odds for these rounds as well as the Sweet 16 and the Round of 32, providing a more complete analysis of any given bracket.
The hardest rounds to predict are those early ones. What does your research say about the first rounds and the Sweet 16?
The Round of 64 is where the real upsets occur. On average, three teams seeded No. 11 or No. 12 will advanced from the Round of 64 each tournament. Among the teams seeded No. 13 or No. 14, picking one such team (or two, if you feel daring) is always a good idea. No. 15 teams are hard to predict. They have been successful the past few years, so we may be due for a dry period. No. 16 teams have never won; best to leave your bracket that way.
For the Sweet 16, make sure you include one team seeded No. 11 or worse; this has occurred in more than 86 percent of the past 30 tournaments. In fact, in more than half the tournaments, two such teams make this round. Among these seeds, No. 11 and No. 12 have been most successful. In fact, the total number of teams seeded No. 11 and No. 12 that reached the Sweet 16 has been almost as much (37) as the total number of teams seeded No. 8, No. 9, and No. 10 (38) over the past 30 tournaments.
So how can college basketball fans feel confident they have a pretty good bracket?
Fill out your bracket backwards, starting with the national champion, then the national runner-up, then the Final Four, and so forth. Mix a reasonable number of upsets into the Rounds of 64 and 32 and the Sweet Sixteen. Our website provides some assistance with the odds in the Elite Eight and the Final Four. If that all sounds like too much work, our BracketOdds website will fill in a bracket for you, based on our analytics model.