< Predicting Oscar Winners


Friday, February 22, 2013


BOB GARFIELD:  And now, a little movie talk. David Rothschild is a Microsoft researcher who accurately predicted the results of the 2012 presidential race in every state but Florida. Now he’s using his powers of prediction on another major event, the Academy Awards. His predictions are data-driven, having nothing to do with whether or not the chase scene at the end of “Argo” was completely implausible or if “Silver Linings Playbook” devolved from a quirky character study to a pat romantic comedy.

DAVID ROTHSCHILD:  I've seen, actually, very few of these movies and it really needs to be that way because once you start throwing your gut in, you’re into a world of trouble because, in the long run, data beats gut every time.

BOB GARFIELD:  Some of the data points that you follow are, are a little surprising. For example, first week box office not such a big indicator, but the actual window of the release is.

DAVID ROTHSCHILD:  A movie that's released in the fall is much more likely to get nominated than another given movie released in the spring. But if a movie that is released in the spring ends up getting nominated, there’s somethin’ special there. So it's a very weird game thinking about what things kind of correlate with the likelihood of getting nominated but then, conditional on being nominated, what correlates with outcomes during the Oscars?

Whether or not a movie was gaining or losing gross between the fourth and fifth week is a really key indicator. If a movie was rated high by the general population versus critics, this was much more important, completely dwarfed out the critical ratings.

If you have been nominated many times before but haven’t won, that’s a good chance that that’s gonna be your year. But then if you’ve been nominated a bunch of times and have won, that kind of offsets that entire bump.

BOB GARFIELD:  Now, a few moments ago – we’re speaking on Thursday - I got some extremely bad news. My wife walked into my office and told me that she had purchased ballet tickets for [LAUGHS] Oscar night. Who’s gonna win?

DAVID ROTHSCHILD:  Who’s likely to win? Let’s put it that way. You know, it’s looking increasingly like “Argo” for Best Picture. “Argo” has won a lot of the previous awards. Right off the bat, “Lincoln” looked really strong by the fact that Steven Spielberg was nominated for Best Director and his counterpart, Ben Affleck, was not nominated from “Argo” for Best Director. But as the award season progressed, “Argo” has gained steam. Interestingly, Steven Spielberg seems even more likely now to win Best Director because his biggest competition, Mr. Affleck, is not in the race.

Daniel Day-Lewis is just a runaway favorite for Best Actor. And then one of the more interesting categories, Best Actress, we have Jennifer Lawrence as a very likely favorite, at about 70%. Jessica Chastain from “Zero Dark 30” was a much closer second place earlier on but has really fallen down.

The least exciting race was Anne Hathaway in Best Supporting Actress, which is a runaway at this point. Helen Hunt as an extremely distant second. By far, of the big six, the tightest is gonna be the Best Supporting Actor category. Tommy Lee Jones is the most likely leading the pack here in the mid-forties at percentage chance over Christoph Waltz of “Django Unchained” at about 40% and then Robert De Niro as a kind of distant but non-negligible third from “Silver Linings Playbook.”

BOB GARFIELD:  I don’t mean this quite the way it may sound, but who cares? I mean, it's an interesting parlor game. What's the greater significance of being able to put together this kind of prediction model?

DAVID ROTHSCHILD:  That's a very reasonable question. [LAUGHS]  And, and I actually, to tell you the truth, I don’t care that much. I care because I think about the science of forecasting. It's fun and exciting and it pulls people in to study popular events. And there’s also a lot of data there, so it’s a good place to start.

But the real effort down the line is to build methods that are extremely domain-independent, so the same sort of methods can be used for politics and for entertainment, for sports, for economic indicators or business decisions that companies could be making. All these things that could be adding efficiency in the world, that’s really why we’re in this game.

BOB GARFIELD:  David, thank you so much.

DAVID ROTHSCHILD:  Thank you very much for having me. It was a real pleasure.

BOB GARFIELD:  David Rothschild is a researcher for Microsoft. 


David Rothschild

Hosted by:

Bob Garfield