Predicting who will win the Oscars is hard.
The winners are chosen by a group of about 7,000 people we know very little about — the members of the Academy of Motion Picture Arts and Sciences — so our usual forecasting methods don’t work as well.
If we want to predict an election, we need polling data. We can’t poll Oscar voters if we don’t know who they are, and they probably wouldn’t talk to us anyway if we did. If we want to predict the outcome of a sports season, we need data on the past performances of the competitors. We can’t get that here either, because every film in the Oscar race is in it for the first time and previous wins for performers and directors are far from guarantees of future victories. If we want to predict what will happen in the economy, there are many leading indicators with strong historical success that can help analysts develop a forecast. Similar data is harder to obtain for the Academy Awards, an event with an annual sample size of one.
The best we’ve been able to do when it comes to predicting the Oscars is to look at preceding events that have strong correlations to historical success at the Academy Awards.
We’ve developed a model that we’re pretty proud of. We assign weights to earlier award shows based on their historical predictive power, sum the weights for the winners, and figure out who’s ahead. While it’s not perfect by any means, it’s an attempt to quantify the institutional support for different major categories. If we can’t directly poll the directors, we’ll look at the awards shows that poll directors and see who wins in order to infer the state of the race. If we can’t run a survey to figure out support for actors among Oscar voters, maybe we can find out who the consensus candidates are based on who wins other awards.
But maybe this is not the best way.
We’re committed to helping push the ball forward on Academy Award prediction, and we want to highlight people who think they can predict the Oscars better than anyone has before. That’s why we’d like to talk to you! We want to find up-and-comers who think they have found the approach that cracks the Academy Awards wide open.
We’re calling for submissions from people who have developed ways to predict the Oscars using data. From the submission pool, we’ll talk to a few modelers who attack Oscar prediction from different angles and then highlight them over the course of our awards season coverage. Creativity is absolutely a plus: Unique models that take a new approach to predicting the winners are preferred to models that rely on traditional ways of predicting the awards, such as betting markets.
The ideal submissions will have a well-planned and transparent strategy, with the overall objective of ascertaining which of the announced nominees is most likely to win the prize. Ideally, they will have compiled most or all of the data needed to justify the approach. But most importantly, they will show they’ve done their homework. Modelers should be willing to stick with it through the season and be willing to maintain the model and provide regular updates on how it’s going.
Throughout the awards season, we’ll check back in with the modelers to see which films and nominees are doing well and to further discuss their models and approaches. We’ll also be hunting for folks who could contribute to our coverage on the day of the big shindig.
Since one thing we’re big on here is that you need more than one successful prediction to demonstrate a model’s validity, we’re not going to crown a winner, per se. But we will use this year’s Oscar race as a chance to look under the hood and see what goes into making a model and to open up some of our resources to help people fuel their work.
If you’d like to submit an entry, please email the following information to walter.hickey@fivethirtyeight.com by Dec. 31 with the subject line “Oscar Model Pitch”:
- A one-paragraph description of your approach
- A one-paragraph bio of you or your team
- A list of the data that you have assembled for your model so far
- Data you will be able to collect throughout the process
- Data you may need help obtaining before or during the process
If we like your approach, we may email you asking for more details. Either way, stay tuned to this space throughout the season for our regular Academy Awards coverage, as well as some new stabs from fresh faces trying to do a very hard thing.