Matthew Stewart first became enamored with the Academy Awards while watching the infamous 2006 ceremony, as Crash beat Brokeback Mountain in one of the most controversial best-picture results in Oscars history. But it wasn’t the final award that drew Stewart’s attention so much as something that happened much earlier in the broadcast: Brokeback Mountain star Jake Gyllenhaal, the film’s co-lead, was competing against Syriana costar George Clooney in the best-supporting-actor race, even though Stewart thought he rightfully belonged in the best-actor category. He’d been submitted as a supporting actor, though, in an apparent attempt to up his chances of winning.
“I kind of felt that was not quite right,” Stewart told Vanity Fair of seeing Gyllenhaal in the category. “I knew his performance was long and also a lead performance. I didn’t see why that move was made.”
The Brokeback gambit didn’t work; Clooney wound up prevailing. Still, an invaluable research tool for awards prognosticators and aficionados was born. Starting in 2007, Stewart began watching and timing nearly every Oscar-nominated performance in the history of the Academy Awards, with the goal of trying to determine whether category fraud—that awards-season political mainstay of jockeying actors into potentially less-competitive categories—had occurred.
“The first one I timed was Fargo,” Stewart said. He found that Frances McDormand, who won best actress for the film, appeared in 27.01% of the Joel and Ethan Coen classic. William H. Macy, who ran in the best-supporting-actor category and lost, appeared in 27.66% of the same film.
“It was alarming to me that one was in the lead [category] and one was supporting,” Stewart said.
While his original plan was to simply focus on all the winners—a goal that took eight years to compete—Stewart branched out to focus on the rest of the nominees as well. His database, available at the website Screen Time Central, is only missing five performances, from films that are not readily available to modern audiences: Emil Jannings (a best-actor winner for two films in one year, including 1927’s The Way of All Flesh), Betty Compson (a best-actress nominee for 1928’s The Barker), Richard Barthelmess (a best-actor nominee for 1928’s The Noose), Lewis Stone (a best-actor nominee for 1928’s The Patriot), and Lawrence Tibbett (a best-actor nominee for 1930’s The Rogue Song).
But while it’s a valuable metric, Stewart said that counting total screen time can’t be the only way of measuring which cases are true instances of category fraud.
“Screen time helps the argument, but I don’t use it as a defining factor, because I don’t think it can be. There can’t really be a formula,” he said. “There’s just no way to plug every performance into a formula and come up with a category. You have to use screen time as a tool to help after you’ve looked at everything else. It should be the least-important thing.”
To that end, what he’s seen in the last decade is a great many lead performances landing in the supporting categories.
“The biggest ones are Emma Stone, and Rachel Weisz too, in The Favourite,” he said. “I think Emma Stone is inexcusable. It’s not explainable at all; she’s just the lead.”
For what it’s worth, the numbers back that up: Stone, who was a supporting-actress nominee for The Favourite, appeared in 48.03% of the Yorgos Lanthimos costume comedy, far ahead of fellow supporting-actress nominee Weisz (35.9%) as well as eventual best-actress winner Olivia Colman (41.72%). Beyond that, Stewart argued, the film is centered on Stone’s character’s perspective: she is the audience’s entry point into the story, and her descent is where the film concludes.