Davidson students crunch data, attempt to beat odds of March Madness ‘bracketology‘
Thursday, Mar. 13, 2014

Davidson students crunch data, attempt to beat odds of March Madness ‘bracketology‘

Tim Chartier, associate professor of mathematics and computer science, and his students have out-guessed millions in ESPN’s NCAA basketball bracket contest. In 2010, Chartier’s best student bracket beat 99 percent of 5 million entries. In 2013, the best bracket beat more than 96 percent of 8 million bracket entries.

For the last five years, Davidson College students have consistently predicted the outcome of the men’s NCAA college basketball tournament with a great deal of accuracy.

In short, their guess is better than yours. Not only that, they continue to fine-tune their prediction methods, and this year there’s added incentive to strive for perfection.

For a contest announced in January, Warren Buffett’s Berkshire Hathaway and Dan Gilbert’s Quicken Loans will award $1 billion – yes, with a “B” – to anyone who fills out the 2014 men’s NCAA tournament bracket perfectly. The deadline to enter is March 19.

Tim Chartier, associate professor in the department of mathematics and computer science, and his students have out-guessed millions of people using math. Predicting the outcome perfectly has never happened, but it appears mathematicians have an upper hand.

In 2009, their best bracket beat 97 percent of 4 million entries to ESPN’s annual online contest. In 2010, their best bracket beat 99 percent of 5 million entries. In 2013, their best bracket beat more than 96 percent of 8 million bracket entries. Other years, ratings have slumped to below 90 percent, but not one sports analyst did better than their best entry, said Chartier.

In 2010 Alex Herman, an autistic teenager from Illinois, claimed to have a guessed perfectly through the first two rounds and predicted Purdue to win – but Duke won that year. The odds of getting that far were 13 million to 1.

Last year, none of the 8.5 million brackets submitted to ESPN had a perfect bracket after the first round, said Chartier. Everyone got at least one wrong in the 32 games played.

Beating the odds

There are 263 different options (9.2 quintillion) for brackets, said Chartier, meaning if someone could create 1 billion brackets a second, it would take almost 300 years to create that many – roughly 9 quintillion.

“The math helps find information in the mass of game data,” said Chartier. “It helps lessen the information and overwhelming combinations.”

Chartier, who has been involved with sports-ranking research for about six years, said students adapt two mathematically based methods used by the Bowl Championship Series: the Colley method and the Massey method. Crunching data from 5,000 games into a computer, students create 350 equations, each with 350 unknowns, and put the computer to work.

“Those methods do better than a lot as it not only looks at your schedule but the strength of your schedule,” he said. “We apply the rankings to all Division I teams, which is about 350 teams. We rank them all. We use the results of every game in the season. We don’t confine ourselves to ranking only the 64 teams in the tournament.”

In 2009, students submitted brackets to test new research ideas that integrated into the rankings a chance for upsets and a team’s momentum, among other factors.

“We had a few other ranking ideas and those performed very poorly,” said Chartier. “Those other methods, however, even in the first year, did very well.”

More than luck

As the guesses get more educated, buzz continues to grow on campus.

“Now, I teach the ideas to (math) majors and (non-math) majors,” said Chartier. “I’ve even had friends learn to do it and that’s one of my favorite parts of this. It can give people a real sense of meaning to doing something mathematical by, in a sense, taking some of the madness out of March Madness.”

With the inherent stress and the variability of sports comes unpredictability.

“Sometimes, I think it really is an upset, the stars essentially aligned,” said Chartier. “If you have an unfair coin that lands on heads 99 percent of the time and tails 1 percent, you still, on occasion, get tails.”

In Chartier’s book, “Math Bytes,” which comes out in April, he tries to show that luck without intuition is dangerous. Chartier has created brackets by flipping a coin and, on average, those brackets beat only 1.7 percent of the 8 million submitted to ESPN last year.

“So, it takes more than just luck,” he said.

Learn from the pros

Senior Jane Gribble, a cheerleader and a math major, had the highest performing bracket last year, landing in the 96.3 percentile of ESPN’s online tournament.

“Many of the games are so close that they could go either way,” she said. “So in that sense, there was a lot of luck in the success of my bracket. But, the logic used to create the bracket was the main reason I was so successful.”

Gribble chose the Massey method for her prediction method, because it has had better success in previous years, but she tweaked it a bit.

“I weighted home games to count as half a win to take into consideration home-court advantage,” she said. “Similarly, I ranked away games as 1.5 games and neutral games as one game. I also divided the season into six equal sections, increasing the weight of the games for each section, because I believe games at the end of the season are a better reflection of a team’s ability than the games at the beginning of the season.”

She plans on using the same method this year to test how the predictions model changes from year to year.

Junior Bealela Donnelly learned how to make a bracket in her class, finite math. This project was part of the reason she wanted to work with Chartier.

“On one of the first days of class, (he) mentioned that people in his class with less basketball knowledge tend to have more successful brackets,” she recalled. “As a basketball fan, there is no way I would’ve picked Florida Gulf Coast over San Diego State or even Georgetown last year, but maybe someone using math would have given them a chance.”

A bracket submitted by senior Seth Kindig, who did research with Chartier last summer, also did well last year, but the odds are stacked against anyone who tries to outsmart fate.

“The odds of winning the Mega Millions full prize are 1 in 175,711,536,” said Kindig. “The odds of correctly guessing the bracket are approximately 1 in 12 trillion.”

Johnson: 704-786-2185

The Charlotte Observer welcomes your comments on news of the day. The more voices engaged in conversation, the better for us all, but do keep it civil. Please refrain from profanity, obscenity, spam, name-calling or attacking others for their views.

Have a news tip? You can send it to a local news editor; email local@charlotteobserver.com to send us your tip - or - consider joining the Public Insight Network and become a source for The Charlotte Observer.

  Read more