THE STUDY APRIL 3, 2012
March Madness has finally come to an end, and the Kentucky Wildcats are the national champions. It wasn’t exactly a surprising end to the tournament—Kentucky was one of the season’s most dominant teams—but it still meant lighter wallets for everyone who chose poorly in their office pools. Choosing one winning team from 64 tournament hopefuls isn’t easy, but have statisticians come up with a method that beats picking the higher seeds and making a few lucky guesses?
It doesn’t appear so, according to a 2001 study. The authors set out to “show how to efficiently calculate the mean and the variance of the number of correctly predicted wins.” Their method didn’t fare much better than models based on regular season performance, team rankings, or odds from Las Vegas—and, with a correct prediction rate of 58 percent, it was hardly an improvement over simply picking the higher seed, at least in simple scoring systems where the only object is to correctly choose the winning team. That may be frustrating for people who labor over a carefully-considered bracket every spring, but for the rest of us, it’s a refreshingly egalitarian finding (and it helps soothe the egos of TNR’s best and brightest, who lost the office pool to a web intern).