(This article was published Saturday, March 17, 2007 in the Macon Telegraph.)
By Phillip Ramati
TELEGRAPH STAFF WRITER
Allen Lynch said one of his longtime goals was to make it on to ESPN.
The 39-year-old father of three knows he won't make it as a sports star, what with a herniated disk in his back and a dearth of athletic ability. But joining the ranks of ESPN's experts isn't out of the realm of possibility.
Lynch, a professor of economics at Mercer University's Stetson School of Business, has spent the past decade working on various "March Madness" models that not only predict which teams will receive at-large bids to the men's NCAA basketball tournament, but also who will win. Lynch developed the system with University of North Florida professor Jay Coleman.
"Our goal is to get on ESPN," Lynch said. "We've been on 'On the Monday' on CNBC. Two years ago, I was on CNN and Jay was on CNN Headline News. ... We're both big sports fans, but we're not very athletic. We've reached the point where our athletic arena is limited to academics."
Lynch, who did his undergraduate work at North Florida before getting his Ph.D. at Florida State, said Coleman once approached him about working on a model that would predict which teams would make it into the NCAA tournament - now known in sports circles as "bracketology."
Looking at many different kinds of variables, the pair narrowed down the key predictors for a college's NCAA chances to three factors: the Ratings Percentage Index, or RPI rating, that measures a school's relative strength; the RPI rating of that school's overall conference; and how the school did during its conference schedule.
The formula works pretty well - Lynch said they usually get about 94 percent of the picks right. This year, Lynch and Coleman missed out on four schools: They had Georgia Tech and Indiana just missing the field and predicted several other schools that were more worthy than Stanford and Texas Tech, both of which lost their opening round games Thursday night.
"Those two picks by the selection committee were just bizarre," Lynch said. "I was glad to see both got pummelled in the first round." Despite a lot of basketball pundits saying that Florida State deserved to be in the field of 65, Lynch said he couldn't find a way to justify his alma mater's inclusion.
"I tried to massage the model in so many ways," Lynch said with a laugh. "I couldn't make them fit."
With the success of what they call the Dance Card, Lynch and Coleman expanded their methodology to predict who would win the games. Using their model, they usually get about 74 percent of the games right on an annual basis.
So, Lynch must be rolling in all the money he wins from winning the NCAA pool every year, right?
Not so much.
"I've never won the pool we have (at Mercer)," he said. "I like to pick crazy upsets, and I tend to weight schools I like a little bit more, like Florida State and Syracuse. We usually enter our model in the ESPN (online) brackets. We usually finish in the top 15 or 20 percent. But the model never picks crazy upsets."
Thus, their model, called the Score Card, never saw George Mason's Cinderella run to the Final Four last year.
Lynch said this is the first year he has strictly gone by the model and not let any of his emotion factor in, and it paid off the first day when he got 14 of the 16 games on Thursday correct. The two he missed - Virginia Commonwealth beating Duke and Xavier's defeat of Brigham Young University - were both statistical toss-ups, Lynch said. According to their system, Duke only had a 60 percent shot at winning, while BYU had just 53 percent chance.
"Anything in that 45-55 percent range, they're all toss-ups," he said. "Usually, they turn out to be very close, entertaining games."
The model predicts a Final Four this year of Ohio State, Florida, North Carolina and UCLA, with the Ohio State beating Florida in the title game.
Lynch said the model they use is not set in stone. He and Coleman are constantly examining new variables that might change what factors they use to weigh a team's chances. For example, with so many of the marquee schools losing their talent after a year or two to the NBA, the smaller schools that keep their players around for four or five years will have a better chance with their veterans.
"There is a lot more parity in the game," Lynch said. "The internal process is always changing. These days, players leave (college) early. A model that was based on schools in 1980 may not have much relevance to today. Thankfully, we'll always have work."
Lynch has expanded his sports models to beyond college basketball, and has done research with the NBA and high school football recruiting.
But as with any statistical formula, Lynch said the models should be a factor in research, not a substitute. The pair's research can be found online at www.unf.edu/
"It should be used as a decision aide, not as replacement analysis," he said.
To contact Phillip Ramati,
call 744-4334 or e-mail firstname.lastname@example.org.