By Ben Zauzmer
The Oscars are dramatic, exhilarating, entertaining, and…mathematical?
In an awards show famous for upsets, plot twists, politics, and preference, I set out to determine if numbers alone can accurately model the Academy’s voting patterns. Last year, in my first attempt, I went eight-for-eight in the major categories (Picture, Director, the four acting awards, and the two screenplay awards), and 15-for-20 overall.
This year, I’m at it again, and the predictions can be found at Ben’s Oscar Forecast. For those of you more into the movies side of this, I give a brief explanation of my methodology here. But for the more mathematically inclined, let’s dig a bit deeper.
The entire premise is that the best indicator of future behavior is past behavior. For instance, if the Golden Globes are a better predictor for Best Score than the BAFTAs (Britain’s version of the Oscars) over the past decade, then they will be a better predictor on Sunday night. It makes sense logically, since each set of awards or nominations is essentially like an early poll among people similar to the Oscar voters. Still, this is the main assumption.
From that point, there are no assumptions, only math. I gathered all of the data from the past 15 years. The exceptions are the few awards where data does not go back that far, as well as the three short film categories, for which enough data does not exist. I then plug everything into a linear multiple regression. The two key differences between my site and the new competitors: I try for every category other than the short films, and I test the variables to make sure they are actually good predictors. No multicollinearity, overfitting, or illogical “noise” allowed.
After plugging in this year’s data to the regression model, I adjust the scores to avoid boundary errors and divide by the sum to create a percentage. These numbers also allow me to calculate other fun facts about the Oscars, using some standard laws of probability.
So, what do the numbers say when all of this is finished? Right now, Argo is the favorite for Best Picture at 60 percent. Best Director is a closer race, but Ang Lee for Life of Pi comes out on top. Daniel Day-Lewis and Jennifer Lawrence are the lead acting favorites. The entire list can be found here.
But of course, these are percentages, not guarantees. Anything can happen when the curtain is drawn and the lights go on in Hollywood.