Math Madness: Statistics Might Help Choose NCAA Brackets

Every March, the NCAA tournament allows committed college hoops fans, as well as casual observers, to predict the outcome of 63 tournament games.

This year, thanks to online pools and Facebook applications, more people are entering than ever before.

There are myriad methods for determining which team will beat another: mascot matchups, school rivalries, team colors, or more-complicated superstitions.

But for those looking for a more scientific method to winning their pool, there are plenty of guides and strategies to improve your chances of bracket success.

The odds

	Considering a team's pre-tournament ranking, or seed, can help improve the accuracy of a bracket, like the one seen above.

If you assumed that each game was a toss-up, like a coin flip, your chances of picking every game correctly would be .5 to the 63rd power - one in 9 million trillion, explained Wall Street Journal columnist Carl Bialik.

No one has ever filled out a perfect bracket, according to his research.

However, game outcomes aren't completely random, and your odds can greatly improve. By figuring in the seeds of teams, which is a pre-tournament ranking of each team on a scale of one through 16 for four different regions, among other mathematic techniques, Bialik improved his chances of the perfect bracket to one in 150 million.

Using research

	By compiling data from past tournaments, you can improve the accuracy of your picks.

Joseph Chatlos, who teaches statistics at Upper Arlington High School in Columbus, Ohio, said that because game outcomes and team selection for the tournament are not random, you can use statistics to your advantage.

"In some ways statisticians always deal with subjective cases where you don't know what will happen," Chatlos said. "We are never very sure of anything in statistics. It's a matter of weighing the odds and figuring out what's going to happen."

Studying the outcome of past tournaments is one way to improve your bracket.

Pete Tiernan, who runs bracketscience.com, uses a database of statistics from the past 23 years of the 64-team tournament to help his customers outperform the average bracket.

Tiernan compiles coaching records, each seed number's past performance and school histories on his old Macintosh computer. He analyzes individual odds for different factors without the use of sophisticated math programs: how does a team do in the tournament if it has won six of its last 10 games? What about seven of the last 10?

"Some of the more clever kids in stats class ask if my sample size is large enough. And sometimes, admittedly it's not," Tiernan said. "There's been 23 tournaments in the modern era since [the] field expanded to 64 and sometimes our sample size is limited."

Numbers versus instinct

	UCLA is a top seed in the 2008 tournament after making it to the final four in last year's contest.

By using data from the past tournaments, you can keep your bracket from placing lower than 55 percent of the entrants, Tiernan said. However, the rest is chance.

To win a large pool, you must go out on a limb and pick some upsets, he said.

"I think that's what people like to do, be the clever person who picks that upset. You also risk being the dumbest person in your pool," he explained.