Posted by It’s that time of the year again when all eyes will turn to the Dolby Theatre in Hollywood for the 95th Academy Awards as movie fans speculate on who will receive one of those coveted statuettes, who will be snubbed, and perhaps more importantly, who will be slapped while on stage! Well, since Will Smith has been banned from all academy events for ten years after last year’s unpleasant incident, Jimmy Kimmel, the third-time host of this year’s ceremony can take a deep breather.
The 2023 nominees are led by a genre classification-defying Everything Everywhere All at Once with 11 nods, followed by the dark comedy Banshees of Inisherin and the 1930 World War I epic remake All Quiet on the Western Front both having secured 9 nominations.
The Data and Models
As we have done since 2018, the BigML team has been hard at work gathering the relevant data points about this year’s nominees and updating our Machine Learning models. Thankfully, this makes it possible for anyone out there to join the fray to make their own ML-powered predictions by simply cloning the Movies 2000-2022 public dataset from the BigML gallery and using it in conjunction with the wide variety of modeling resources of the BigML platform. Don’t have an account yet? Get a free one here in less than a minute!
This latest version of our dataset covers 1,452 movies nominated for various awards from 2000 to 2022 with 300+ features including:
- Film characteristics such as synopsis, duration, budget, and genre.
- Film critic and audience reception measures like ratings and box office performance.
- Nominations and winners for key industry awards such as Golden Globes, BAFTA, Screen Actors Guild, and Critics Choice.
Given that our OptiML models outperformed the 50/50 blend of OptiML and Deepnets predictions last year, we decided to stick with OptiML this year. This eternally popular AutoML feature on BigML automatically finds the best-performing supervised models for you. When the individual OptiML model search executions finished for each award category, our engineering team built Fusions, by combining the top 20 best-performing models from each OptiML search. This means different modeling techniques partake in the final predictions which in turn can make the resulting predictions more robust and arguably less noisy due to equal weight averaging.
Once our Fusion models were created, we made Batch Predictions against the movies produced in 2022 that we had split aside into a separate dataset from our main dataset mentioned earlier.
As usual, given BigML’s emphasis on white-box models with full explainability, we can dig deeper into models and predictions for added introspection as needed. For example, you can navigate to any Fusion model’s partial dependence plot to see how various data fields interplay in determining whether a given movie or individual nominee will win the Oscar.
The 2023 Predictions
We can hear you say “Enough techno-babble already…” so here’s what you probably came here for. Below, we predict the most likely winner along with other nominees in the same category sorted by decreasing scores. Keep in mind that these scores are not supposed to add up to 100. The models are telling us how a movie/artist with a given set of characteristics will likely do in a given award category based on 20+ years of historical data on that particular award. That assessment is made independent of the other competing nominees for the same award this year. In other words, a high score can be interpreted as that nominee’s overall profile looking quite like the old winners of that category.
- Everything Everywhere All at Once looks to be a strong win candidate here.
- This prediction naturally goes hand-in-hand with the previous one of Best Picture.
- One of the most anticipated showdowns of the night will be for the Best Actress Oscar. Let’s see who will prevail.
- Brendan Fraser seems to have the edge here but don’t sleep on Austin Butler or even Colin Farrell.
- Jamie Lee Curtis may have this one in the bag.
- Looks like a pretty comfortable cushion in favor of the Everything Everywhere All at Once star.
- This St. Patrick’s Day week may bring good luck to the Banshees of Inisherin entry with strong Irish roots.
- Women Talking possibly receiving the only Oscar of the night for that production.
Last year we added 11 more categories that are more technical in nature aside from the high-profile categories listed above. We almost had a clean sweep of these inaugural categories missing only a single one. So why not keep predicting them? Here, we followed the same approach of generating Fusions for each award category and arrived at the following top picks for you. Unlike last year, these categories are likely to be handed out to eight different entries perhaps pointing out to a highly competitive crop in terms of technical merit.
- Best Cinematography: ALL QUIET ON THE WESTERN FRONT
- Best Costume Design: BLACK PANTHER: WAKANDA FOREVER
- Best Film Editing: EVERYTHING EVERYWHERE ALL AT ONCE (NOTE: TOP GUN: MAVERICK is a close 2nd here.)
- Best Sound: ALL QUIET ON THE WESTERN FRONT
- Best Visual Effects: AVATAR: THE WAY OF WATER (NOTE: TOP GUN: MAVERICK is a close 2nd here.)
- Best Makeup and Hairstyling: THE BATMAN
- Best Music, Original Song: EVERYTHING EVERYWHERE ALL AT ONCE (NOTE: BLACK PANTHER: WAKANDA FOREVER is a close 2nd here, and TOP GUN: MAVERICK a close 3rd.)
- Best Music, Original Score: BABYLON
- Best Production Design: BABYLON
- Best International Feature Film: ALL QUIET ON THE WESTERN FRONT (See footnote:
ARGENTINA, 1985) - Best Animated Feature Film: GUILLERMO DEL TORO’S PINOCCHIO
This concludes our 2023 Oscars predictions. Now it’s your turn to play the all-knowing movie critic either by printing out our predictions, or better yet, coming up with your own by the time Sunday’s festivities begin. After the winners are revealed, we’ll follow up with a post early next week that grades how we did this time around. Till then, good luck to all the Oscars hopefuls!
NOTE: In the original version of this post, the Best International Feature Film prediction was mistakenly listed as ARGENTINA, 1985 (our second pick) instead of ALL QUIET ON THE WESTERN FRONT due to a data sorting mishap.
The Official Blog of BigML.com 