x
Breaking News
More () »

Mathematicians create computer algorithm to dominate March Madness

Looking for a way to dominate March Madness?
Credit: Maddie Meyer/Getty Images
A detailed view of a Wilson Basketball with a NCAA March Madness logo on it during the first half between the Wisconsin Badgers and the Florida Gators during the 2017 NCAA Men's Basketball Tournament East Regional at Madison Square Garden.

Looking for a way to dominate March Madness?

The answer may not be in watching the games or knowing the first thing about basketball. Instead, it may be about plugging in the numbers. At least that's the theory for a pair of mathematics PHD candidates, from Ohio State University, who have created a trio of computer algorithms, to predict the upsets in the first round. Their results are at the bottom of the article.

These algorithms looked at various statistics, such as shooting percentage, point differential, and travel time, to find the matchups that are statistically most likely to have an upset. To create this algorithm, they gathered March Madness statistics from the last 17 years, dating back to 2001.

"Essentially what you do is you get a bunch of old data," said Matthew Osbourne. "And you feed it through these algorithms. You can kind of treat it like a black box. You set dials and knobs on this black box. And that allows your machine to essentially learn from the past."

Why Humans Can't Be Trusted:

While we all probably feel that our strategy will give us the perfect bracket, often that is far from the case. Too often, we allow the emotion to get in the way.

"Humans are really bad at letting people effect their psychology essentially," Osborne said. "So they'll hear something like 'oh, this person had a really tragic upbringing, and he's making it to March Madness. He's going to make it big.' And they'll allow this to kind of tug at their heart strings and influence their decision."

And our bias goes beyond emotional attachment to players. Often fans will select their hometown team, against better logic. We also fall victim to the "Recency Bias," which means we're more likely to select a team that has done well recently.

"What our models do," said Osborne. "Is they ignore all that. Because they're just a computer. So they look at the data, and look at past trends, and try to identify common things that happened in these upsets."

So, will their bracket be a perfect predictor? Osborne admits, probably not.

"We're never going to have a perfect model," he said. "The tournament is really unpredictable. You're in or out on one game. And a lot can happen in one game."

Results:

The graphic below breaks down the likelihood of an upset for each of the three models. The percentage under Model A and Model B indicates the likelihood that there will be an upset in the first round. Model C just indicates whether or not there will be an upset.

This model indicates that some likely upsets include Texas (10) over Nevada (7), Florida State (9) over Missouri (8), Davidson (12) over Kentucky (5), Buffalo (13) over Arizona (4), and Oklahoma (10) over Rhode Island (7).

What do you think about their predictions?

Before You Leave, Check This Out