DATA MODELING – Method to the March Madness

Die-hard college hoop fans always mark their calendar with a fateful day in March known as Selection Sunday. That’s the day the NCAA Basketball Tournament Selection Committee unveils the 65 teams fortunate enough to be invited to the big dance?the NCAA playoffs.

Every year it’s full of controversy. There’s always a team or two that thinks it got cheated, and the committee’s deliberations are shrouded in secrecy. While the factors the committee is supposed to consider are public knowledge, no one knows exactly how it arrives at its final decisions. Jay Coleman, an operations management and quantitative methods professor at the University of North Florida in Jacksonville, is trying to bring some method to the March Madness.

Coleman and Allen Lynch, an economics professor at Mercer University in Macon, Ga., have derived a statistical modeling procedure called the Dance Card. Using statistical software called SAS, Coleman crunches all the factors the selection committee considers when it decides who to take as “at large” teams (teams that didn’t get an automatic bid by winning their conference). All this data is plugged into the dance-card equation, which produces a “power index” figure. Coleman then ranks the teams by power index and determines which should make the cut.

Coleman used this model to predict the past two tournaments. In 2000, it correctly predicted all but three spots. In 2001, it missed only one. “The committee chose Missouri, and we picked Richmond,” he says, “but after the selections were made, most commentators said Richmond was the one team that got the shaft.” He also analyzed all the selections from 1994 to 1999 and found his model would have correctly predicted 94 percent of the available at-large bids during that time.

Coleman believes this type of statistical modeling also applies to a business context. “Anytime you have a binary variable?either you’re in or you’re out, or in a business context, you decide to fund a project or not?this could be applied,” he says. “This system can take major criteria in a decision and at least estimate what kind of weight each of these criteria received in the decision, and that’s valuable information.” Coleman’s next project might be to tackle college football?deriving a better model to determine who makes the Bowl Championship Series.