This Sunday's Academy Awards promise to be a hell of a show—quite literally.
A series of unfortunate events shadowed the 91st edition of the Oscars night in multiple occasions, from the mutiny by prominent Academy members over the producers’ plans to give out four awards during commercial breaks to the selection of Kevin Hart as a host, who had to withdraw after a dispute over homophobic tweets.
Celluloid or real life, everyone loves drama. Apparently, DataRobot doesn't—the Boston-based predictive modeling company decided to take some thrill out of the competition for Best Movie by unleashing its machine learning algorithms to predict the winner.
The result? A list of calculated probabilities of each nomination winning the golden statue. According to DataRobot, the Oscar goes to...
- Roma– 17.8% chance of winning
- A Star is Born – 16.2%
- The Favourite – 15.5%
- Black Panther – 15.1%
- Bohemian Rhapsody – 12.2%
- BlacKkKlansman – 11.3%
- Vice – 9%
- Green Book – 6.9%
DataRobot is not new at this kind of ruin-all-the-fun analysis on popular culture subjects. Last year, the company tried to predict who would die in the next series of Game of Thrones; more recently, they successfully predicted this year’s Grammy "Song of the Year" winner.
"Most of the times, we're doing predictions for places like banks, insurance companies, healthcare, [which] don't want to share back the results," DataRobot data scientist Rajiv Shah explained. "Here, we can share that out."
For this study, Shah grabbed data on previous Oscars winners since the 1960s, as well as the full list of nominations. Then, he added information about the movies, such as release date, actors, genre and duration; finally, he combined the dataset with the number of nominations each movie got, the critics' evaluations and popular rating scores. DataRobot’s machine learning platform did the rest—coming up with over 100 models for this particular problem.
"For Roma, the biggest factor was the critics' score; critics really liked Roma, and DataRobot's model picked up on that," Shah said.
Directed by Alfonso Cuarón, Roma is a drama about the life of a family—another factor taken into account by the algorithms, as movies that end up winning Best Movie often have the word 'life' in their plots, Shah added.
The algorithm may be positive about the outcome, but it didn't watch any of the movies, nor can it spot more intangible assets of a story. Shah, for example, would love to see a superhero movie win the award, so he leans towards Black Panther; as for me, I am part of the large fan base of Green Book, the comedy-drama film starring Viggo Mortensen and Mahershala Ali, which scored low on DataRobot's ranking.
"I worry a little bit, because I know [Green Book] is a favorite of people," Shah admitted. "If Green Book does win, next year we're going to have to figure out a model that figures out the factors that led Green Book to win."
As it always happens, the truth will be revealed on Sunday by opening the envelope—and maybe not even then.
