Can a computer fill out a March Madness bracket better than humans? So far this year, it doesn’t look like it.
As the Sweet 16 kicks off Thursday with four NCAA men’s basketball tournament games, we took a look at Microsoft Bing’s bracket performance thus far.
Bing, which used “intelligent machine-learning technology to analyze social and search signals” to make its bracket predictions, is 34-14 so far through 48 games played. It missed eight picks in the first round, and another six in the second round. Bing also only has two Final Four teams remaining — it picked Louisville and Villanova (its champion pick) and both teams were upset in the second round.
Compared to ESPN’s “People Bracket,” which calculates the most popular picks from each of the 18.8 million brackets submitted to its Tournament Challenge, Bing actually is doing worse. The “People’s Bracket” is 38-10 so far; it missed six picks in the first round and four in the second round. It also has three Final Four teams remaining.
Bing’s performance this year through two rounds matches last year, when it also fell behind the “People’s Bracket.”
“Bing Predicts may be the magic ingredient for making all your bracket dreams happen,” Bing wrote before the tournament started. “We use our intelligent machine-learning technology to analyze social and search signals, plus more than a decade of college hoops statistical data to bring you insights on every single one of your picks.”
We’ll see how the rest of the tournament shakes out, but with two Final Four teams out already, it’s not looking good for Bing to beat the human crowd.
Predicting March Madness outcomes is typically extremely difficult — there are already no perfect brackets remaining this year — and it doesn’t look like “intelligent machine-learning technology” offers much of an advantage.