Posted by Rob Knies

David Rothschild

March Madness, they call it, but David Rothschild is taking things to new extremes.

Rothschild, an economist from Microsoft Research New York City, has been making a name for himself in the past few years. In 2012, he correctly predicted 50 of 51 Electoral College outcomes in the U.S. presidential race. Last month, his models accurately forecasted the winners in 21 of the 24 categories in the 86th annual Academy Awards.

Now, for the first time, he is applying his prognosticative powers to the Big Dance, the 2014 NCAA men’s basketball tournament. With Selection Sunday complete and with bracket in hand, Rothschild will be following every step of the spectacle, beginning March 18 and continuing all the way through to the April 7 championship game.

As are so many of us, he’s filling out a bracket, based on Selection Sunday pairings and culminating with the title game. His, though, will be based on prediction markets, which tend to favor the highest-seeded teams in the tournament. But the real action, predictively speaking, will come in his round-by-round probabilities, using fundamental data that includes not only seedings, but also such information as team records and past performances.

From the Round of 64 through the regional quarterfinals, the Sweet Sixteen, the Elite Eight to the Final Four, Rothschild will be calculating the ongoing likelihood of success for each surviving team in this single-elimination tournament. Like the event itself, his fine-tuning promises to be a wild ride.

“Unlike other events I have been covering,” he says, “March Madness is unique, in that there are 67 events spread out over a period of time. It is not just making a series of predictions and seeing how they evolve in real time over the course of one event.

“In this domain, there are a series of conditional events that we do not even know about until a few days before. This adds a unique challenge to our infrastructure, and I am looking forward to testing!”

Such an endeavor has an ulterior motive, of course. By creating a model and iterating over time, Rothschild’s experimentation aims to advance the state of the art in such predictions and promises to deliver answers to bigger questions than just those from the sporting realm.

For example, his methods resulted in correct predictions of the winners of 19 of the 24 Oscars categories in 2013. This year, by refining his model, he ratcheted up that already astounding success a couple of notches higher.

“The Oscars do not have great fundamental data,” Rothschild observes. “There is no data point imaginable that will help me determine by statistics alone which movies have the best makeup or sound editing. This information is almost all idiosyncratic, so I rely on the collective wisdom of the crowd to help aggregate this data.

“Politics is a little better, where fundamentals such as past election results and incumbency can lead to very accurate forecasts early in an election cycle. But, like the Oscars, there is a lot of idiosyncratic information that flows in later, from specific traits of the candidates to potential gaffes they might commit.”

Athletic competitions, though, are simply awash in numerical information.

“Sports has great data,” he says, “and the models we can create from them are shockingly accurate. While it might be hard to quantify a particular injury or the emotion on a given play, the data is really dominant in sports.”

The goal, he adds, is to learn more about predicting the evolution of the tournament across a few weeks. Stronger predictions make for a stronger model. That’s why Rothschild, who obtained his Ph.D. from the prestigious Wharton School at the University of Pennsylvania, will be spending the next 21 days listening to sneakers squeaking across basketball courts.

“I look forward to providing real-time, accurate predictions for all games,” he says, “and success will look like well-calibrated predictions for all games. I know everyone will focus on the static brackets, but as a challenge, I look toward each and every game.