The math of 'March Madness'

In bracket picks, statistics compete with superstition

You might pick your favorite colors. Or most unusual mascot (lumberjacks, anyone?). Hottest point guard. Hottest cheerleaders. Best uniforms. Your girlfriend’s mom’s cousin’s son’s alma mater.
Or maybe you actually like numbers — wins and losses, margins of victory and defeat, RPI, Elo and all that classic (sports) nerd stuff.

Or maybe your picking strategy is no strategy at all. That’s cool, too. Your chances are probably still as good as anybody’s to fill out the ideal NCAA basketball tournament bracket correctly.

In fact, given a 50-50 probability of any team winning any game in the tournament, statisticians say your chances of constructing the perfect bracket are about 1 in 9.2 quintillion. That’s one out of 9.2 billion times 1 billion. So, good luck. (But for what it’s worth, a perfect bracket could earn you a $1 billion check from Quicken Loans and Warren Buffett.)

The odds may not be exactly in your favor — the same could be said for any of the 132 Division I men’s and women’s teams vying for their respective tournament championships starting this week. But that’s no reason not to play the game, said Brian Habing, a statistics professor at USC.

“You can’t win if you don’t try,” he said. “Play for the dream.”
Sports statisticians use a number of mathematical formulas — with myriad combinations of variables such as a team’s win-loss record, average opponent ranking, strength of schedule and home-court advantage — to rate teams and predict the probability of one team beating another.

Among the stats-based tournament prediction tools is the FiveThirtyEight forecasting model, developed by acclaimed statistician Nate Silver, who is famous for his accuracy in predicting political elections. The model calculates each team’s chances of advancing to each round based on a composite of power rankings, preseason rankings, tournament seeding order, player injuries and geography.

Statistically speaking, the odds are ugly for the two men’s teams from South Carolina competing in this year’s tournament field.
Silver’s model gives Wofford, the No. 15 seed in the bracket’s Midwest region, only a 5 percent chance of advancing to the third round over No. 2 Michigan. Coastal Carolina, the No. 16 seed in the East, stands an even slimmer chance over No. 1 Virginia, with a 4 percent probability of advancing to the third round. The model predicts both teams have less than a 1 percent chance of making it to the Sweet Sixteen, much less becoming national champions.

For that matter, no team’s statistical probability of winning it all is greater than 15 percent. That means that Louisville, the team surest to go the furthest, according to the model, has an 85 percent chance of not being the nation’s best team after all.

“I guess if it was easy, everyone would win every year,” said Habing, who noted that intangible variables like a team’s style of play and injuries to significant players make statistics an imperfect predictor.

He doesn’t let the numbers be his sole guide when filling out a bracket.

“Being a statistician does not save you from being superstitious,” he said.

That’s why he’ll let his gut guide his picks for close statistical matchups, and he’ll probably choose too many Big 10 teams just because he likes them more, he joked.

“The thing about the tournament is if you’re using your brain at all, you’ve probably got about the same chance of winning as anyone else,” Habing said. “So part of it should be the fun.”



Comments powered by Disqus

Please note All comments are eligible for publication in The Daily Gamecock.