Sentiment Spotlight: The Revenant Claws Its Way to the Oscar’s Best Picture
February 26, 2016
This year’s race for the Best Picture Oscar is one of the most divided in recent history. Both the buzz and the usual award-season predictors are all over the map. The Directors Guild Award and the Producers Guild Award, two awards that usually go to the eventual Oscar winner, went to two different movies: The Revenant and The Big Short respectively. Spotlight, another critically favored drama, won the SAG awards for best ensemble.
Further complicating matters is the nature of the voting. The Oscar’s voting system makes the Iowa Caucuses seem straightforward. The Academy uses a ballot that allows voters to put each movie into first, second, or third place, and then weighs how many of each type of vote each movie gets when compared to the others. Given the mixed results of the previous awards, this means that this year’s Best Picture may come down to how many second- and third-place votes each nominee gets.
With the prior awards less predictive than usual this awards season, we decided to dig deep into historical Oscars trends and use sentiment and text analytics to build a predictive model.
After examining best picture nominees from 2000 to 2015 we found several key patterns:
- Best picture winners tend to have a predominately male viewership, even more than best picture nominees as a group
- Emotions like anger tend to be predictive of winners, and these are seen more often in males
- NPS on IMDB reviews is even more predictive than either RT Critic or Audience Scores
- Sentiment and number of comments on themes like Direction, Cinematography, and especially number of mentions of “Oscar” are incredibly important
- Box Office totals for Oscar winners are larger
As we concluded in our last piece describing our initial findings, films with demographics that resemble the Oscars’ voting population overall tend to preform better as Best Picture candidates.
We used all of these features and more, to build a random forests based statistical model.
Our model overwhelmingly predicts the Revenant to win Best Picture, with a 64% likelihood of winning.
Surprisingly, the second place finisher in this model is Mad Max: Fury Road, listed by some as a spoiler. The comments about The Big Short lack emotive responsiveness and a lack of Oscar mentions in reviews hurt its chances here, as evidenced by its low odds in this model.
Why The Revenant?
The film tracks well on several categories:
- 85% male audience from IMDB
- 39 NPS, one of the highest scores of all Oscars movies, second highest this year
- Very high box office gross, at $361 million
- A high portion of Oscar “worthiness” comments, at 128% more often than the average 2016 best picture nominee
- A high portion of cinematography comments, 281% more common than the average 2016 movie
- 52% more comments about editing than average and at a higher sentiment
- 26% more mentions of direction than average and higher sentiment
Of course, there are certain aspects we cannot model against. Will Oscar voters give Alejandro Inarritu a second straight best picture win? Will the momentum of The Big Short carry over to Oscar’s voters? Will The Revenant relatively late theatrical debut (December 2015) hurt its chances? These all-too-human biases are what make predictive decisions so difficult.
However, if the patterns in the data hold true, expect The Revenant to echo its story by surviving this grueling challenge to raise the Best Picture Oscar this Sunday.
Dheepan Ramanan is a data scientist at Clarabridge. Follow him on Twitter @DheepanRamanan.