Tuesday, April 21, 2009

Georgia Tech profs put heads together, pick an NBA champ

David Liam Kyle

The Playoffs are filled with talking heads. They're NBA veterans in nice suits, or hard-nosed reporters spending their lives in the locker rooms, all spewing opinions on who the next NBA champion migt be. Once in a while, they even turn out to be right.

What if there were some guys who were never in the league, though, and who never appeared on TV, and they turned out to be better analysts than your typical talking head?

And what if they could kick your butt in a game of Jeopardy, too?

Those guys are professors Joel Sokol, 37, Paul Kvam, 46, and George Nemhauser, 71, at the Georgia Institute of Technology's H. Milton Stewart School of Industrial and Systems Engineering. Together, the professors have invented an equation called the Logistic Regression Markov Chain (LRMC) that can predict the outcome of sporting tournaments. Using this highly complex mathematical equation, they can pick the NBA champion for the 2008-09 season without watching a game.

With the LRMC by their side, these group of mathletes claim to be better prognosticators than Charles Barkley, Gary Payton or any other talking-head analyst on the planet.

"I think we could take on Gary and the guys, when it comes to predictions," says Sokol, the mastermind behind the LRMC.

"We could give this talk about the LRMC to thousands of professors at universities who would love to hear the equation," says Kvam. "But most of them would probably say 'What are Cleveland Cavaliers and why are they playing Jazz in Utah?'"

The LRMC initially was designed to decide the champion of the NCAA men's basketball tournament. The professors were inspired by a last-second shot by Tennessee that knocked Tech out of the tournament in 2002.

Since that game, the LRMC has picked the correct NCAA tournament winner three out of five times, including this year, when North Carolina won. In addition to picking the champ, in 2008 the LRMC got all three of the final rounds correct, and it even got the NIT champion right, too.

"Everyone has this intuition just by looking at head-to-heads and blowouts for who is going to win," says Sokol. "All we are doing is making a more rigorous argument."

Because the professors seemed so sure in their equation, and did pretty well in the NCAAs, we decided to challenge them to see if they could predict the outcome of the NBA Playoffs.

They scoffed. They said they could do it in 20 minutes.

Which, when you think about it, easily beats hours of sports chatter from the talking-head analysts.

The professors took 16 years worth of NBA data from every team in the Playoffs and used that to determine the strength of each team and its ranking. The professors then picked random matches between teams from the past 16 years, took the outcomes and plugged them into a logistic-reasoning equation.