How To Predict The Oscars' Winners With 'Oscarmetrics' Math
TURNING MOVIES INTO MATH
Ben Zauzmer's spent the last eight years using data analysis to predict who will win the Oscars, and he has a high rate of success -- his model's even managed to top FiveThirtyEight data master Nate Silver.
By day, he's manager of baseball analytics for the Dodgers -- but when he was a Harvard freshman, he wanted to know if that same approach could be applied to predicting the Academy Awards. He'd wanted to know who the math said would win that year's Oscars, but Googling didn't turn anything up.
So he decided to apply Moneyball-style principles to the movies. He spent a month in the library gathering data from websites and press releases, building his own Oscars dataset. Then he started building statistical models using information like the past winners of different awards shows, and how they correlated with who won at the Academy Awards.
He's gone on to write about what that analysis predicts for the Hollywood Reporter, as well as writing a recent book on the subject, Oscarmetrics.
He has a lot less data to work with than the incredibly detailed numbers drawn from thousands of Major League Baseball games, but told us his Oscar predictions are constantly being improved. Zauzmer adds new data sources and tweaks his model to make sure his own biases don't affect what the model spits out as the probabilities.
WHAT THE MATH PREDICTS
It's an interesting year for making predictions, according to Zauzmer, with four movies receiving 10 or more nominations each for the first time ever. That could make determining the winners that much more difficult, without one film leading as clearly. Joker has 11, with 1917, the Irishman, and Once Upon A Time In Hollywood with 10 nominations each.
Looking back, one of the most surprising predictions was in 2012, Zauzmer said. The buzz was that Viola Davis would likely win Best Actress for The Help -- but the math showed Meryl Streep as a slight mathematical favorite.
Streep won for The Iron Lady.
Another example: Ang Lee's Best Director win for Life of Pi over Steven Spielberg and Lincoln.
The math isn't always right -- sometimes the pundits manage to outscore what the hard numbers say.
"And that's to be expected, because the math is essentially establishing probabilities, rather than making predictions. And sometimes things with high probability happen and sometimes they don't," Zauzmer said.
WHAT THE PAST GOT WRONG
One of the questions he tackles in his book: which movie should have won the Best Picture Oscar each year.
Using stats like the American Film Institute's rankings and the IMDb user votes, he breaks down what modern critics and experts believe were the best choices each year.
He wrote about all the years that the math says the Academy got it wrong.
"I challenged myself to watch every single one of these movies -- not just the ones that actually won Best Picture, but also the ones that should have won Best Picture," Zauzmer said.
Being who he is, Zauzmer shared his own personal work stats, and broke down how he spends his time -- 95 percent on statistical work for his day job, the other 5 percent on awards.
"I get to do the Oscars work for this one month of the year, but then the rest of the year, I get to hang out at Dodger Stadium," Zauzmer said. "If that split were able to continue for years in the future, I would be extremely happy."