Using data to forecast the Oscars Best Picture winner

Patrick Flynn

20 January 2023

Using data to predict the outcome of the Oscars is not a new concept - Ben Zauzmer’s book ‘Oscarmetrics’ relies on this very idea.

However, while Zauzmer does not reveal behind the curtain and therefore keeps his methodology hidden, I’ve created a model for some of the key races (loosely based on this FiveThirtyEight piece from 2018) and will reveal how it works. This article focuses on Best Picture.

The idea behind the model is a simple one: compile a database of comparable film accolades and weight their results based on how closely they match the results of the Oscars in a particular category.

These accolades are divided into three groups: Awards (comparable prizes like the BAFTAs, Golden Globes and AFI Awards that are voted on by a mixed membership of critics, journalists, filmmakers, academics and industry professionals), Industry (the Producers/Writers/Directors/Screen Actors Guild Awards) and Critics (from large prizes like the Critics Choice Awards, to the favourites of regional and demographically-based groups of critics like the Phoenix Film Critics Society and the Alliance of Women Film Journalists).

In forecasting, there is always a balance to be struck between volume and relevance of data. While we have access to over 90 years of Oscars history, data from the 1940s is not really relevant to the 2020s. Furthermore, since the #OscarsSoWhite controversy in 2016, there has been a push from the Academy to diversify the pool of voters and, naturally, that will have some effect on the films and individuals who get nominated and ultimately win each award.

From 2008 to 2013, the BAFTAs and Oscars awarded Best Picture to the same film on all six occasions. However, in the eight years since then, the two have shared the same preference just once. It’s clear that different accolades can diverge or converge from each other over time, so it’s important to weight the data more heavily towards recent years.

The model allocates points to each award based on the weighted percentage of how often they have matched the Oscars Best Picture winner in the last 15 years. Using the above example of the BAFTAs, while the raw figure is 7/15 (47%), the weighted version is 37%. I then squared that figure and multiplied it by 1000, so a film that wins the BAFTA for Best Picture will receive 135 points.

Finally, I made some adjustments to weight each group of accolades (Awards, Industry and Critics), to make sure the 25 prizes in the Critics category don’t overpower the Industry category, which only has four.

Before we take a look at 2023, what do the most recent years look like? As you can see below, CODA was an outsider whose success in Industry prizes seemingly carried it to a win, despite being ranked a lowly seventh in the Awards category and fifth by Critics. On the other hand, in 2020, Parasite was ranked fourth in the Awards group, but its support among Critics in particular enabled it to see off 1917.

In the 2023 race, Everything Everywhere All At Once is the frontrunner so far on 6,000 points, ahead of The Banshees of Inisherin on 3,400. In third is The Fabelmans and fourth is Tár, a film that hasn’t been discussed much as a contender and could be a half-decent outside bet.

There are still a lot of points left to be allocated, including the vast majority in the Industry category, so I’d be wary of making any hard-and-fast predictions. A win for The Banshees of Inisherin in the prestigious Producers Guild of America awards, for example, would see the film overtake Everything Everywhere All At Once.

Right now, I’d be looking at backing Everything Everywhere All At Once at better than evens, and consider having an outside bet on Tár at the right odds. Top Gun: Maverick, in particular, ought to be avoided.