by Mike Sharkey
Staff Writer
Maybe Dr. Jay Coleman will learn one day to quit leaving Missouri out of his NCAA men’s basketball pre-tournament “dance card.”
The 64-team tournament started Thursday with 16 games at 16 venues. And, once again, Coleman — who is chair of the Department of Management, Marketing and Logistics at the University of North Florida — and his model for picking the 64-team (it’s actually a 65-team field with a play-in game) field has been remarkably accurate.
“We got 63 of 65 teams,” said Coleman. “We missed Missouri and Wyoming, who both got in. Last year, we got 64 of 65 and only missed on Missouri.”
Coleman’s model is based on several elements: a team’s RPI ranking, the conference RPI ranking, a team’s number of wins against top 25 RPI teams, their conference record, their record against teams with 26-50 RPI rankings and their record against teams with 51-100 RPI rankings. Like most computer-generated ranking systems for any sport, teams are awarded points for quality wins and penalized for losses, especially if the loss came against a weaker opponent.
Since Coleman devised his dance card in 1994, he has successfully predicted 289 of the tournament’s 308 at-large bids, good enough for an accuracy rating of 93.8 percent.
“We have never missed more than three at-large teams in nine years,” said Coleman, who bases his model on the past performance of the NCAA Selection Committee, which every year draws the ire of fans and sportswriters all over the country. Coleman’s dance card web site is quick to point out that his model is only as accurate as the Selection Committee is consistent. “The model is a model of the Committee’s behavior, right or wrong.”
With 31 automatic bids going to conference champions, about half the work for Coleman is already done. The tricky part, he says, is trying to figure out whether the Selection Committee will go with a 25-5 Butler (they didn’t this year) or opt for a middle-of-the-road team from a powerful conference (like Wisconsin, who went 18-12 in the Big 10).
That very dilemma is what causes Coleman and those who help him with the model each year a great deal of anxiety as they try to mirror the Selection Committee’s last few bids.
“I was a little nervous this time,” said Coleman. “In my opinion, it’s kind of an odd year. You have teams like Butler and Kent State who had good records, but low RPI rankings. Butler didn’t get in, but Kent State did. We had them ranked as 13th seed, but the Committee gave them a 10 seed.”
Next year will mark Coleman’s 10th year of creating his own dance card — a term used because the tournament is informally referred to as “The Big Dance.” Looking ahead, Coleman thinks it may time to tweak his model and expand his forecast beyond just who he thinks will get into the three-week tournament.
“We may start to look at predicting first round results,” said Coleman, indicating that anything beyond that would be difficult for a model to foresee, mainly because the model will go with the favorites and upsets are random and unpredictable, even for a computer. “And, even though the data is far less interesting to a lot of people, but we start doing a model for the women’s tournament. I think we’ll run a model with the past few year’s women’s data and how well it would have predicted the women’s bids. My guess is that it would be about the same.”
Surprisingly, Coleman said he isn’t involved in a tournament pool. He said by the time he and his comrades pore over mountains of data and statistics for several days, he’s a bit tired of the whole thing. Still, he can’t help but offer up an opinion as to who may make it to the Final Four in Atlanta (the title game is April 1). Coleman likes Cincinnati, Duke, Kansas and Maryland (all No. 1 seeds, by the way) to make it to Atlanta.
“I’m a Clemson grad and an ACC [Atlantic Coast Conference] fan, so I’ll probably pull for Duke,” he said.