O.R./Analytics in Action Blog
Since President Obama famously fills out his NCAA men’s basketball tournament bracket each year, we thought we’d ask Ed Kaplan, president of INFORMS, for his picks. Kaplan, a professor at Yale University whose research generally focuses on informing public policy on such critical issues as healthcare and antiterrorism, teaches a course on sports analytics and co-authored (with Stan Garstka) an article, “March Madness and the Office Pool,” for the highly respected academic journal Management Science. Kaplan’s pick to win it all this year is Kansas, “even though that means picking UConn to lose,” says the 30-year resident of Connecticut.
Following the logic espoused in his article, Kaplan used a dynamic program to fill out the brackets backward. Instead of starting by picking first-round winners and advancing them in a similar fashion, the self-described “Member-in-Chief” of INFORMS first asked who he expects to win the entire tournament (Kansas), and then enters Kansas to win in all earlier rounds. The next question: Who loses to Kansas in the final? The model suggests that the best pick is Michigan State. Continuing backward to the national semifinals, the model picks Oklahoma and North Carolina to round out the Final Four. If you’re one of the more than 40 million people filling out a bracket this week, you can use the same model at a website created by Tom Adams.
Kaplan is one of several prominent members of INFORMS who have put some serious science into their “bracketology.” For example, Sheldon Jacobson of the University of Illinois at Urbana-Champagne and the treasurer of INFORMS, has organized a website along with his students that focuses on bracket odds. Meanwhile, a team of professors at Georgia Tech led by Joel Sokol developed the LRMC (Bayesian) ranking system, which the New York Times described as “influential in the seeding process” in its coverage of March Madness in the March 14 edition. Laura Albert McLay, associate professor at the University of Wisconsin and vice president of marketing, communications and outreach for INFORMS, explains bracketology and Markov chains in a YouTube video.
Let the games (and the madness) begin.