Oscar predictions: Comparing three mathematical models

Published February 26, 2015   |   

Ben Zauzmer, a Harvard statistics whiz who has a 75 per cent success rate in predicting the winners of Oscar Awards every year, has correctly predicted 18 of 21 winners in 2015 Academy Awards with a success rate of 86 percent.
Using his mathematical model, Zauzmer successfully picked all the high-profile winners, including Best Picture (“Birdman”), Best Actor (Eddie Redmayne), Best Actress (Julianne Moore), Best Director (Alejandro G. Iñárritu), Best Supporting Actor (J.K. Simmons), and Best Supporting Actress (Patricia Arquette). The only categories he missed were Animated Feature, Film Editing, and Original Score.
How does he do it? “I have gathered thousands of data points on Oscar ceremonies over the past two decades – such as categories movies are nominated in, other awards show results, and aggregate critic scores – and I use statistics to calculate how good a predictor each of those metrics is in each Oscar category. Then, I plug in the numbers from this year’s awards season, and that gives me the percentage chance that each film will win in each category,” Zauzmer told the Boston Globe.
“For example, in the best actor category, the British equivalent of the Oscars (BAFTA), the Screen Actors Guild (SAG), and a host of other organizations pick their winners. Often, these groups don’t all agree. So, I use math to determine, based upon previous years’ results, how much we should listen to each group. I have set up the formulas such that the math automatically gives higher weights to those predictors that have done a better job of forecasting the Oscars in the past,” he said.
Popular Science has examined the predictions of three mathematical models by Ben Zauzmer, FiveThirtyEight and PredictWise and published a table of comparison, in which they found that one prediction out of 10 popular Oscar categories goes wrong. You can know more about the comparison here.