Harvard's Oscar Guru

Ben T. Zauzmer '15, an Applied Math concentrator in Mather House, gained Internet fame last year for his strikingly accurate predictions of who would win the 2012 Oscars. This year, he's at it again.

Using results from similar awards shows, critics' picks, and other available data, Zauzmer runs statistical analyses to determine the probability that nominated movies, directors, or actresses will win in their given categories this Sunday. Flyby sat down with Zauzmer to talk movies and math.

Flyby: How long have you been interested in the Oscars?

BZ: My interest in the Oscars goes back many years, although last year was the first time I tried to predict the Oscars with math. Last year went well. [I got] all 8 of the major categories [and] 15 out of 20 overall.

Flyby: What are your predictions for this year?

BZ: For Best Picture right now it's "Argo" at 60%. I don't use personal predictions, just math. For Best Director it's Ang Lee for "Life of Pi," and then the acting awards are: Daniel Day Lewis for "Lincoln," and Jennifer Lawrence from "Silver Linings Playbook." For Best Supporting Actor, it's Harvard's own Tommy Lee Jones '69. That's going to be a little closer, with Christoph Waltz from "Django Unchained" being a close second. And then Anne Hathaway is the runaway favorite, mathematically, for Best Supporting Actress [in "Les Misérables"].

Flyby: How do you get the data for these predictions?

BZ: A lot of time spent on the Internet, on MetaCritic, Rotten Tomatoes, IMDB, Wikipedia, the websites of different guild awards, the Oscars' website, etc., gathering it all, throwing it in Excel, then using Excel and Mathematica and Stata to actually come up with the formulas. Much of it this year was done over winter break. I had a wonderful break gathering data about movies and watching movies.

Flyby: How has it been having your predictions garner a lot of press?

BZ: It's been a lot of fun, people from all over the world have been interested. I think a lot of people are on there because they're filling out their own Oscar polls this weekend, they want to give it their best shot. I know last year at Harvard, some kid entered his freshman dorm's Oscar pool. He won after using my picks, and he gave me a bar of chocolate that he had won. So that's always fun.

Flyby: As far as you know, are you the only person who has done statistical Oscar predictions based on data instead of just general consensus?

BZ: Last year, I was the only person, because I scoured the Internet and did not find a single other [Oscar prediction website] as far as I know. This year, there are a couple others that have just started. I think it's all just for fun, and it's great to see other people getting in on the mix. I don't necessarily agree with their models mathematically. I think a lot of other people are using data that they expect to be predictive [but is not]. A good example is box office data, for Best Picture. You think, "Oh, a movie is more popular with viewers, it's more likely to win Best Picture." Unfortunately, the math doesn't hold. The Huffington Post Model heavily relies on that. But, at the end of the day, I might prove more right than them, they might prove more right than me, just by random chance.

Flyby: In your findings, what is the biggest predictor of Oscar winnings?

BZ: It depends on the category. For Best Picture, right now, I believe it is the win for the Directors' Guild. BAFTAs are huge in a lot of categories. If I had to pick one that was overall the closest mirror to the Oscars, it's easily the BAFTAs. There is no polling from the Academy. If there were, that would make my job so much easier, and the percentages could be a lot higher.

Flyby: Are there any awards shows that seem to not be a good predictor for Oscar wins at all?

BZ: Oh, sure, a good example is the Golden Globes. I have all the data in my spreadsheet, and so last year I checked it, this year I checked it again, the answer is still no. When you do the multiple regression, you come out with a negative coefficient. The Golden Globes are famous, historically, for not being a great predictor.

Flyby: Have you ever thought of applying your methods to other awards shows, like the Emmys?

BZ: Yes, I have, and right now there's just not enough data for the Emmys, the Grammys, the Tonys. What you need is a lot of other people, award shows, critic scores, and it just doesn't exist for everything else I've seen.

Flyby: What are your plans for the Oscars this weekend?

BZ: I will be watching them in Mather. Last year I hosted an Oscar party for the freshman class [and] I got about 70 people, that was a lot of fun. I don't know how well this will go, there's an enormous amount of chance, part art and part science. Science at best could predict the majority of the Oscars, but I fully expect that there's still a lot of reason to be nervous Sunday night. I think that's why they're so much fun.

This interview has been edited for concision.