Every year thousands of people gleefully draw on months of college basketball knowledge and deeply rooted sports loyalties to draw up (what they believe) will be the million dollar March Madness bracket.
Around the same time, I think back on all the college basketball I didn’t watch, realize my alma mater once again did not make any tournaments (we’ll always have 1963, Ramblers), and consider my general apathy toward sports. Then I fill out a bracket based purely on whims and the motivation to win cash from my coworkers.
But not this year.
I’m choosing a bracket through BracketOdds, a program from the University of Illinois Urbana-Champaign’s computer science department led by professor Sheldon Jacobson. It samples from the 9,223,372,036,854,775,808 possible bracket combinations to create predictions using principles from probability theory and operations research. It’s based off the idea that seeding is the most likely predictor of who moves on in the tournament: number one seeds have won the national championship over 60 percent of the tournaments since 1985, the BracketOdds team explains. So many of the brackets generated by the tool have number one seeds winning.
It randomly generates a bracket, and if you don’t like it, you just refresh the page and get a new one. If you have a pretty good idea of how you want to fill out your bracket, but are interested in using probability to guide a few choices, they also have tips on likelihood that a team will move forward based on their seed, and projected seeds for this year’s tournament.
“We just sample across the 9+ quintillions possible brackets.”
In a time of the year literally named for “madness,” it’s a blissfully logical take on guessing a winner.
“We do not attempt to predict who will win, like other websites or sports pundits; we just sample across the 9+ quintillions possible brackets,” the team says on their site.
Will computer science prevail in the face of sports insanity? We’ll find out. Check out my bracket here (spoiler: I have Oregon winning because I went to Portland for the first time last fall and it was awesome) and I’ll update this post with the results.
Update 3/21: After one weekend of March Madness, it looks like my algorithm might have a few bugs. After two rounds, my PCT is 29% and my bracket has only correctly predicted 28 out of the 40 games played so far. It did correctly guess the Arkansas- Little Rock upset, but also guessed Iona and South Dakota State would upset (they didn’t). At least Oregon is still around!
Update 4/6: While Oregon didn’t stick around as long as I’d hoped, when all was said and done, my algorithm-generated bracket came in right in the middle of the Streetwise bracket competition: 11th place out of 22. My final PCT was 62%, and the bracket guessed 34 out of 63 games correctly, about a 53 percent accuracy rate. Considering I did better than over 60 percent of the people who played, and likely picked their own bracket, I just might try BracketOdds again next year.
Image credit: Screenshot via YouTube