Sports fans' fancies are once again turning toward the annual NCAA basketball tournament.
By submitting your email address, you agree to receive emails regarding relevant topic offers from TechTarget and its partners. You can withdraw your consent at any time. Contact TechTarget at 275 Grove Street, Newton, MA.
The office pool is almost as much a staple of spring planning as first-quarter financial results. Two university professors are applying predictive analytics to narrow the field of potential contenders for the so-called Big Dance.
The "Dance Card" began when Jay Coleman, professor of operations management at the University of North Florida in Jacksonville, Fla., was surfing the Internet and came across a site listing the attributes of tournament teams and those that nearly qualified over a six-year period. Armed with that historical data, Coleman decided to model the future.
Software vendors' increased focus on predictive analytics has made it possible to make forecasts -- whether they be the likelihood that a customer will respond to a marketing campaign or the odds that your alma mater will make the tournament, Coleman said.
"Any time you've got a set of historical data, you've got an opportunity to apply these sets of analytics," Coleman said. "Bank loan analysis, projecting marketing successes, or [selecting] where a product would be best located on a shelf can all be predicted."
Four years ago, Coleman and Allen Lynch, associate professor of economics and qualitative analysis at Mercer University in Macon, Ga., created a model to determine the colleges in the country that would qualify for the tournament's 34 at-large bids. They used the same applications that many businesses rely on: predictive analytics software from SAS Institute Inc., in Cary, N.C.
When the duo applied their formula retroactively, they correctly predicted 94% of the at-large teams from the past 10 years. This year, they have added a prediction for the women's tournament as well.
In the 65-team men's tournament, 31 teams are awarded berths by virtue of winning their conference in the regular season or through a conference tournament. The remaining teams are chosen by the NCAA Selection Committee, which considers a range of factors, including its own Ratings Percentage Index (RPI), for measuring a team's performance throughout the season.
"What this model does is synthesize the data," Lynch said. "We've developed one mathematical model to hone in on the important stuff."
Who's in and who's out
The NCAA considers 29 metrics in awarding bids to teams. Lynch and Coleman have narrowed those to six: a team's wins against top 25 teams, its conference record, RPI rank, win-loss differential against teams ranked from 26 to 50, win-loss differential against teams ranked 51 to 100 and the overall RPI of the team's conference.
Conference champions and automatic bids are still to be determined, but the professors' "Dance Card" is already filling up with forecasts.
Fans of the University of Utah and the University of Colorado can rest easy. The professors predict that they'll be the last two schools to make the cut. Xavier University and University of Texas El Paso fans should worry about their teams getting a chance to play for the big prize, Coleman said.
The University of Missouri, which twice got an at-large bid when the professors predicted it would be left out, are poised to foil the Dance Card again this year. The school has a questionable record but has played some tough opponents.
Applications like SAS 8.02 have made this type of predictive modeling far easier to conduct than ever before, Lynch said.
"When I started graduate school, SAS language was only available on mainframes, and it took a lot of coding," Lynch said. "We're at a point where SAS is evolving. Most coding has been taken care of behind the scenes and most programs allow for users to quickly perform analysis with a comfortable [graphical user interface]."
Still, some need statistical training to apply predictive analytics, but Coleman said that advances have broadened the number of possible users. In fact, the biggest obstacle is the data -- finding clean, accurate information.
Click here to see the "Dance Card."
FOR MORE INFORMATION:
Best Web Links: Business intelligence and data analysis