Microsoft, Big Data Pick Oscar Winners - And They Are...

Ten days before Hollywood hands out its Oscar statuettes, a pair of studies - one by Microsoft's in-house "Nate Silver," and another measuring social influence - have already picked the winners.

Microsoft Research's David Rothschild, who, like Silver, used early polls to correctly predict the outcome of the presidential elections in all but Florida, has used the predictive nature of the early awards shows to place his bets on who will be winning the various Academy Awards. Meanwhile, an English analyst firm, Brandwatch, has attempted to slice social media data in a couple of new ways (by both critical reaction and popular acclaim) to anticipate the winners.

And who are those winners? The envelopes, please...

  • Best Picture: Argo
  • Best Director: Steven Spielberg (Rothschild, Brandwatch popular) / David O. Russell (Brandwatch critics)
  • Best Actor: Daniel-Day Lewis
  • Best Actress: Jennifer Lawrence (Rothschild, Brandwatch popular) / Jessica Chastain (Brandwatch critics)
  • Best Supporting Actor: Tommy Lee Jones (Rothschild) / Christoph Waltz (Brandwatch popular) / Robert de Niro (Brandwatch critics)
  • Best Supporting Actress: Anne Hathaway
  • Best Animated Film: Brave
  • Best Original Song: Adele's "Skyfall"

The remaining categories, including best makeup, screenplay, documentary shorts, and others, can be found on the respective sites: Predictwise for Rothschild's predictions, and the Brandwatch Oscars site.

Why Will They Win?

In November, Rothschild used the same methodologies employed by quant hero Nate Silver to determine the outcome of the 2012 Presidential election: examining polling information collected before the election to determine the outcome. In February 2012, Rothschild wrote that Obama would win, well before election season got underway. As it turned out, of course, he was right.

(See also Why Nate Silver Won, And Why It Matters and Nate Silver's Model A Stunning Portrait Of Logic Over Punditry)

 “I approach forecasting the Oscars the same way I approach forecasting anything, including politics,” Rothschild said in a blog post. “I look for the most efficient data, and I create statistically significant models without any regard for the outcomes in any particular year. All models are tested and calibrated on historical data, with great pains taken to ensure that the model is robust to 'out-of-sample' outcomes, not just what has happened in the past. The models predict the future, not just the past.

“Thus, the science is identical, but there are differences in which data prove most useful,” Rothschild wrote.

The predictive models that Rothschild could tap into are the ones that most people are now using to handicap Oscar races: previous awards shows like the BAFTA awards, Screen Actors Guild (SAG) awards, and the Golden Globes. Some data he tossed out: For elections, fundamental data such as past election results and economic indicators can be used as predictive tools. But in movies, box-office figures and even ratings are not statistically effective, he said.

"I focus even more heavily on prediction markets, which are very robust, but I also include some user-generated data that helps me learn more about correlations within movies and between categories, such as, ‘How many categories will Lincoln win?" Rothschild added.

Finally, he updates his results in real time. Naturally, there's a way to tap into these results yourself: the Oscars Ballot Predictor app for Microsoft Excel, one of the few apps to provide real-time data for Microsoft's Office suite. The app allows users to vote, and includes the real-time, up-to-date Oscar predictions.

What could be Rothschild's next step? "Sports is something we're looking at," a Microsoft spokeswoman said via email.

Whose Opinion Matters: Critics, Or Audiences?

Brandwatch has taken a more "traditional" approach: pull together mentions of each actor, director, movie, or other category across a broad swath of social media to look for positive, relevant references that can indicate a good chance of winning.

Brandwatch taps into the Twitter firehose, and to date relevant Oscar mentions have totaled 304,550 mentions, with about 1,400 to 1,600 per day being added at the end of January. Naturally, that number will go up. Twitter makes up about 40% of the data that Brandwatch samples, according to a FAQ provided to ReadWrite.

One surprise, it found, was that Lincoln was the early odds-on favorite to win Best Picture. But sentiment flipped after Argo started winning the title at the Producers Guild of America, British Academy Film Awards, LA Film Critics Association, the Golden Globes and others.

What Brandwatch tries to do - differently, it says, from other studies - is pull together the volume of positive predictions. There are two variables: the number of mentions, as well as the sentiment behind them. This tries to ensure that a large number of comments on Helen Hunt's red-carpet dresses, for example, won't be factored in any more than a smaller number of positive comments for rival Jessica Chastain's performance.

However, the study also breaks down the projected winners by two categories: Critics, both "professionals" at major papers, plus semi-pro bloggers at enthusiast sites, and the general public. The skews show both sides of the acting industry, who aren't paid critics but know their business presumably more than the average joe.

Brandwatch was hired to perform the study by the Motion Picture Association of America (MPAA), most known for its aggressive stance on copyright and attacks on file-sharing networks. 

But Brandwatch suggests another use for the data: "The findings hold wider implications for the film industries. If winners diverge from viewer favorites, this could indicate a greater need to relate to target audiences. Further qualitative analysis can uncover why film titles are recommended online: vital information for gaining endorsement and boosting box office takings. Key actors and directors can be correlated with film titles: To what extent does an established cast boost online reputation (and by extension sales)?"

It's not quite clear why Brandwatch's critic/public split is a better gauge than Rothschild's single number. But if you're running a betting pool, the smart money is on Argo, Daniel Day-Lewis and Anne Hathaway going home with Oscar on their arm.

Image source: flickr/ebbandflowphotography.