Lights. Camera. ACTION!
On Sunday, March 4th, the nominees for the 90th Academy Awards will gather at the Dolby Theatre in Hollywood, patiently waiting to see who will take home an Oscar. In anticipation of 2018 Awards season’s biggest night, McCann Minneapolis put their brains to work. Taking into account multiple components including previous awards won, number of nominations, and age, the McCann Minneapolis analytics team has predicted the ultimate winners for four specific Oscar categories.
An In Depth Training of the Models
Sam and Charlie, two data analysts at McCann, have built a model that includes three machine-learning methods that are used on previous Oscar nominations and wins. Those three methods are Gradient Boosted, Random Forest, and Spline. All three of these models learn from existing dataset dating back to when the Oscars first premiered. Other factors, like date of movie release, age of actor/actress, number of past Oscar wins, and other award ceremonies are included in the dataset. Each descriptive statistic is converted to a 1 or a 0 and fed into a predictive model. Using this information, each model type chooses the best fitting model to previous years.
The models all used the same input to learn and make predictions, but sometimes they disagree with the importance of a variable. For example, the Best Director predictive models have been correct for the last 13 years with only one disagreement in 2013 when the Boosted Model disagreed with Ang Lee winning Best Director with Life of Pi. This year, all of the models list Guillermo Del Toro as the most likely to Best Director category. Likewise, all three models suggest that winning the Director’s Guild Award is the most important variable in the equation. However, all three models disagree on what is the second most important variable. The Boosted Model and the Random Forest model both agree that the total number of Oscar nominations for the movie is the second most important factor. However, the Spline model shows that being nominated for Best Costume Design is the second most important variable.
In mentioning a disagreement between models, the Best Actress predictions were split between McDormand in Three Billboards Outside Ebbing, Missouri and Saoirse Ronan in Lady Bird. The Boosted Model and the Spline model both agreed that McDormand was the stronger pick, but the Random Forest tipped toward Ronan. It is worth noting that the Best Actress category has been accurate for the last three years. However, in 2014 when this category was split, the Boosted Model was the method that was accurate by choosing Cate Blanchett in Blue Jasmine. With the confidence levels of the Boosted and Spline being at 74% and 90%, and the Random Forest being at 59%, McDormand will most likely be the winner.
A few interesting quirks from the model, such as age of performer, also came up as a key factor. The average age for the last five Best Actress winners was 34, while Best Actor was 44. 2018 has a wide age range of nominees in both categories, starting with 23-year-old Saoirse Ronan to 68-year-Old Meryl Streep and for men, 22-year-old Timothée Chalamet through 63-year-old Denzel Washington. It is important to note that the model does not specify if old or young is better, but simply that age plays a stronger role in its decision. Will this bode well for Saoirse Ronan or Margot Robbie? Will Daniel Day Lewis close out his career with yet another win?
These models have helped us predict the winners for the last four years. In addition, to add a less analytical, more personal spin, we have included results from an internal poll of our agency. While the opinions of our own team, ranging from those who are avid moviegoers and those who wait for the rental, suggest a toss-up for Best Picture, the other three categories have more distinct leaders.
Analytics Model The Shape of Water
McCann MPLS The Shape of Water and Three Billboards Outside of Ebbing, Missouri (TIE)
This could be as competitive as 2017, which saw La La Land and Moonlight duke it out for this category. Hopefully when it comes time for the big reveal, the right card will be in the hand of the presenter! USA Today’s poll from February 26 says that the public calls for The Shape of Water – other major media outlets assert the same result. While Three Billboards Outside of Ebbing, Missouri picked up most of the more prestigious early awards, nothing is a given at this point.
Analytics Model Gary Oldman in The Darkest Hour
McCann MPLS Gary Oldman in The Darkest Hour possible upset by Timothée Chalamet in Call Me By Your Name
With Gary Oldman racking up all of the major acting awards, coupled with being a previous nominee, all lends itself to his being the category’s likely winner. Although the Academy has been generous with Daniel Day Lewis, he is not likely to overcome the pluses in Oldman’s column. However, it should also be noted that three major publications have given Timothée Chalamet the edge (The Guardian, The Washington Post, and The New York Post) due to his extraordinary work at such a young age and his performance in the closing credits of Call Me By Your Name.
Analytics Model Frances McDormand in Three Billboards Outside of Ebbing, Missouri
McCann MPLS Frances McDormand in Three Billboards Outside of Ebbing, Missouri possible upset by Saoirse Ronan in Lady Bird
Frances McDormand has the edge in earning her 2nd Oscar for Best Actress, while Sally Hawkins (The Shape of Water) garnered interest, all signs point to McDormand picking up the prize. Nevertheless, do not count Saoirse Ronan out, given that she once again serves up an extraordinary performance in Lady Bird at the young age of 23.
Analytics Model Guillermo Del Toro’s The Shape of Water
McCann MPLS Guillermo Del Toro’s The Shape of Water
With Del Toro’s momentum in the category, and his success with the Director’s Guild Awards, it is not likely to see any upsets in this category. Christopher Nolan may have some sentimental votes due to his long standing in the industry, but the likelihood of a win is lessened as the film Dunkirk did not prevail during awards season. If anyone else in the category could upset Del Toro, Greta Gerwig’s work on Lady Bird is seen as the most likely challenger.
The Final Verdict
All signs lead to the Analytics team’s model as being spot on this year. However, these trends do not include a pulse of how actual Academy voters – all 7,258 of them – experienced these films. Did any media prognostications affect opinions? Did any of the millions of dollars spent to sway their votes in the days prior to the close of voting make a difference? How does the current state of social issues impact preferences of Academy voters? If recent history is an indication, there will definitely be a few surprises come Sunday night.
Check out our predictions from 2017’s Academy Awards to see how well our methods work: http://www.mccannmpls.com/blog-1/2017/11/30/predicting-the-oscars-2017
Pardoe, I. and Simonton D. K. (2007) Applying Discrete Choice Models to Predict Academy Award Winners, J. R. Statist. Soc. A(2008), 171, Part 2, pp. 375-394