The NCAA Division I men's basketball tournament is big business and March Madness is exciting, but is the selection process fair?
Using the SAS Analytics-powered "Dance Card" formula developed by Jay Coleman of the University of North Florida, Mike DuMond of Charles River Associates, and Allen Lynch of Mercer University, these professors uncovered selection committee bias in favor of particular conferences, as well as bias in favor of the teams with some representation on the NCAA men’s basketball selection committee itself. Bias in the seeding process appears to be even more pronounced than bias in the at-large selections.
Although many have surmised over the years that these types of biases exist, no study before this one has comprehensively examined the range of biases in both processes. With SAS analytics, the evidence is speaking for itself.
Check out the Dance Card rankings of all NCAA Division I men's basketball teams. Applying SAS analytics software, the professors have long predicted, with great accuracy, the "at-large" teams – those teams that did not get automatic bids to the tournaments.
In classes, these three professors use these predictive models as a way to illustrate the surging business analytics trend to their business school students. Right after showing students the Dance Card application, the professors show them six or seven business applications that use the same type of model. These students are getting real world experience with the power of analytics so they can embrace, influence and prepare for what will be, what can be, and what should be.
I'm seeing this story everywhere, Faye. And I'm fascinated to hear that the professors’ Dance Card correctly predicted 33 of the 34 at-large selections for a 2010 accuracy of 97%.
Sports Illustrated says, “In NCAA tournament selection unconscious human bias exists.” http://bit.ly/bpjghl
Here's this week's NBC news story, discussing bias in the NCAA selection process. http://bit.ly/aPWQRw
That's some great accuracy at 97%
Yes, Dave. With 10 years of data (1999 through 2008), the version of the SAS-powered Dance Card currently being used would have correctly predicted 329 of the 341 available at-large NBA Tournament slots, or 96.5% accuracy. The Dance Card can only be as accurate as the Selection Committees are consistent;