Coleman

Later this month when the NCAA Tournament Committee selects the 64 Division I college basketball teams to participate in the "Big Dance" no one will be more interested than Dr. Jay Coleman.

Coleman, chair of the Department of Management, Marketing and Logistics, and Dr. Allen Lynch, a former UNF faculty member who has since moved to Mercer University, have developed a math model that predicts the college basketball teams to be chosen for the NCAA tournament with up to 97 percent accuracy. In 2001, the model referred to as the "dance card" missed only one selection. The committee chose Missouri, while the researchers went with Richmond.

This year, Coleman is hoping for 100 percent.

He and Lynch have already gotten the media's attention with their model which was built by analyzing a variety of statistics over a six-year period. What they've found in their analysis is the NCAA committee, which meets in secret each year to select the teams, is actually fairly predictable in its decisions, Coleman says. "Despite the criticism they occasionally received from fans and the media, they (the committee) seem to apply a consistent standard in selection," he said.

Since about 30 teams get automatic invitations to the tournament each year because of winning their respective conference championships, the model concentrates on the at-large invitations. The NCAA committee is required to consider a variety of team-performance statistics in making its selection. Every year an NCAA "nitty gritty" report summarizes relevant information for all teams combined. The most important statistic to the committee is the ratings percentage index (RPI) of each team. The RPI statistics are provided only to the committee. Over the final weekend of the regular season, the committee comes up with its decisions.

Coleman and Lynch have taken the NCAA reports of the six college basketball seasons from 1994 through '99 along with the information on which teams made the tournament in those years to model the selection criteria of the committee. To approximate the RPI information, the two use a similar index developed for CBS Sportsline by Jerry P. Palm and posted on his web page. With all of this information, the model correctly selected more than 98 percent of all at-large teams in Division I men's basketball between 1994 and 1999.

So other than showing that the NCAA committees have been remarkable consistent in its selections, what is the value of the model?

Coleman says the model could help the NCAA committee by indicating how previous committees tended to make decisions. "Our objective was to create a model that could be universally applied in the future regardless of who might sit on the committee," Coleman said.

There may also be benefits for college teams. "The model can be used by teams to measure their performance compared to the competition and examine what factors are given more weight in the selection process," Coleman says. Ultimately, the model could also be of use by teams when developing their schedules. The number of wins against top-ranked teams, for example, weighs heavily in the model.

But despite all the mathematical calculations, Coleman admits there is no way to predict the "human element" in the committee. Since members change from year to year, there is always the possibility that the selection process will also change especially when considering the "bubble" teams.

The reception in the NCAA to Coleman's model has been somewhat cool. Wire stories have quoted the director of the NCAA tournament as saying that while the model is interesting it is not particularly important to the process. Unspoken, of course, is the potential problem facing the NCAA selection committee if it picks a team which obviously shouldn't be at the "Big Dance" according to a tested model while overlooking a more qualified team. The committee may have a lot more explaining to do in the future if it varies significantly from the "Dance Card" analysis.

Stayed tuned to the March 10th committee selections. For further information about Coleman's selection method go to www.unf.edu/~jcoleman/dance