The 2015 March Madness tournament was a triumphant year for Bing Predicts, the predictive algorithms that Microsoft uses to pick the outcomes of everything from reality TVs to sporting events. For 2016, Microsoft has launched March Madness even earlier, with a site predicting which schools will receive an invitation to the NCAA tournament.
The premise behind the new Arewein.net site is a simple one: Pick a university, and Bing will attempt to predict whether a given school is in or out, the overall chance they’ll make the tournament, and even their seeding in the overall bracket rankings.
ESPN analyst and (ugh) Duke alumni Jay Williams said Microsoft has built its own power index that the company adjusts daily, based on “factors ranging from each team’s strength of schedule, opponents’ win/lose record, and even detailed statistical analysis regarding their on-court tendencies such as ball control, rebounding and field-goal percentage.” Social sentiments are also factored in.
Why this matters: Microsoft’s Bing prediction engine has moved more and more into the spotlight, offering up picks from the Oscars to the World Cup. It’s a smart branding strategy, combining many of what Microsoft hopes are its signature technologies—search, the cloud, business analytical tools—in a way to help solve real-world problems. The new site simply amps up March Madness earlier. Oh, and in case you missed it: Bing predicts that all of the home teams will win the NFL divisional rounds, with New England repeating as Super Bowl champs.
Some odd results
That all results in some seeming anomalies, however. For example, at press time, the California Golden Bears were given a 23 percent chance of making the tournament—Bing decided that that was good enough to extend Cal an invitation. But the Rhode Island Rams were given a 27 percent chance to make the tournament—not good enough, in Bing’s view; they’re out.
This can be explained through the vagaries of the tournament selection committee, according to Microsoft. Microsoft explained that teams in smaller conferences who are expected to finish second may have high probabilities of making the tournament even if they are not as strong as the best at-large qualifier. In other words, the main idea is that the better team isn’t always guaranteed the highest odds of making it in due to the NCAA’s process of inviting automatic qualifiers.
On Feb. 8, Microsoft added a detailed post that explains some of the rankings oddities away. Basically, it boils down to the tournament committee offering at-large bids as well as guaranteed invitations. According to Microsoft, the probability of making the tournament is the sum of the "chance to win conference tournament" plus the "chance of getting an at-large bid assuming not winning the conference tournament."
Last year, Microsoft’s Bing went the conservative route, picking Kentucky to win it all. The company Microsoft itself estimated that it finished in the top 30 percent of all brackets — and topped both Google and Facebook’s own tournament predictions in doing so.
But the most interesting part of the process was that Microsoft offered its picks for every single game of the tournament, offering some of its justifications along the way. That allowed players to let Bing “auto-fill” a bracket for them, or diverge with their own predictions.
And the most noteworthy player that did that was... Microsoft’s own Satya Nadella, who destroyed Bing’s NCAA predictions—and many more noteworthy pundits to boot, with his own shockingly accurate bracket picks. Can he do it again this year?
This story was updated on Feb. 8 with additional comment from Microsoft, regarding the way in which the company calculates its probabilities.