How to Predict a Winner

Professors use SAS® analytics to predict which teams will compete in national championship – with 94% accuracy. by Anne Milley

Every March, basketball fans across America prepare for an annual phenomenon called March Madness, where 64 NCAA college teams compete in the men’s and women’s tournaments to determine the national champions. Along with the tournaments come the nationwide rituals of office pools, brackets, last-second buzzer-beaters and debate over which team might emerge as this year’s Cinderella.

And every year brings controversy over which teams make the cut for the tournament, a.k.a., the Big Dance. The at-large selections by the NCAA Basketball Tournament Selection Committee inevitably leave some teams and fans elated and others feeling snubbed.

But two college professors seem to have discovered a method to March Madness. Professors Jay Coleman, an operations management professor at the University of North Florida in Jacksonville, and Allen Lynch, an economics professor at Mercer University in Macon, Georgia, have predicted the NCAA Division I tournament teams with stunning accuracy. Over the past 10 years, their analytic-powered NCAA "Dance Card" has boasted an impressive 94 percent accuracy rate. This year, Coleman and Lynch correctly picked 31 out of 34 (91 percent) at-large tournament selection slots, missing only Utah State, Notre Dame and LSU. In 2003, Coleman and Lynch correctly picked 32 out of 34. Their best year was 2002, when they picked 33 out of 34 teams. Coleman and Lynch also made selections for the women’s NCAA Division I tournament for the first time this year, correctly picking 31 of 34 teams for a 91 percent accuracy rate.

So, how do they consistently select the right teams year after year? The professors have tapped into predictive modeling technology from SAS.

Coleman and Lynch have applied statistics to the factors used by the selection committee to predict the selection of at-large teams – teams that did not get an automatic bid to the tournament by winning their conference tournament.

The Dance Card application then produces a "power index" used in determining at-large tournament slots. Since its inception, the application, powered by SAS, has achieved an accuracy rate of between 91 and 97 percent.

According to the professors, the accuracy of their Dance Card – and the factors and weights it includes – suggests that the tournament selection committee is fairly predictable in its decisions, despite annual barbs from fans, teams and the media.

The Dance Card is produced in two steps. It starts by simply assuming that all teams with a Ratings Percentage Index (RPI) ranking of 25 or better will make the tournament, and all teams with an RPI ranking of 80 or worse will not make the tournament. All remaining teams (those with RPI rankings of 26 through 79) are then ranked according to the Dance Card scoring code, and the remaining available at-large slots are assigned based on those rankings.

Using this procedure, the Dance Card has never missed on more than three spots over the last decade. From 1994 to 2003, the Dance Card correctly predicted 321 of the 342 available at-large tournament slots – or 93.9 percent.

The Dance Card uses a model derived by Coleman and Lynch to predict which teams "on the bubble" will get at-large tournament bids and which ones will have their bubble burst.

Using the decisions of the NCAA Tournament Selection Committee from 1994 through 1999, and 42 pieces of information (e.g., RPI rankings, number of wins and losses, conference records, etc.) for all teams that were candidates for at-large selections in those years, Coleman and Lynch devised the Dance Card as an estimate of which pieces of information were most important to the selection committee, and the weights that the committee placed on those pieces of information. The Dance Card model suggests that only six pieces of information are highly important in determining whether a team gets an at-large bid:

• RPI rank.
• Conference RPI rank.
• Number of wins against teams ranked from 1-25 in RPI.
• Difference in number of wins and losses in the conference.
• Difference in number of wins and losses against teams ranked 26-50 in RPI.
• Difference in number of wins and losses against teams ranked 51-100 in RPI.

Since SAS began working with Coleman and Lynch, we have been pleased but not surprised at their results, which are typical of those SAS helps businesses achieve every day. With the power of true analytics, businesses can turn their decision making from a jump ball to a slam dunk.