The Academy of Motion Picture Arts and Sciences announced its nominees for 85th Academy Awards and the big story so far this awards season is Lincoln, with 12 nominations. In our initial likelihoods of victory for the big six categories, Lincoln is our most likely winner in three: best actor (Daniel Day-Lewis) at near certainty, best picture at 94 percent, and best director (Steven Spielberg) at 70 percent. For best supporting actor, Lincoln’s Tommy Lee Jones at 46 percent is in a tight race with a The Master’s Phillip Seymour Hoffman at 48 percent. Silver Linings Playbook’s Jennifer Lawrence is the favorite for best actress at 56 percent and Les Miserables’ Anne Hathaway is the favorite for best supporting actress at 95 percent. After best supporting actor, best actress is our next most competitive category with Zero Dark Thirty’s Jessica Chastain just behind Jennifer Lawrence. Two other key movies to consider are Life of Pi with 11 nominations and Argo, which, while a long-shot, is our second most likely best picture behind Lincoln.
The likelihoods that I will describe in this column over the next six weeks, through the February 24th awards ceremony, utilize three key sources: prediction markets, fundamental data, and, eventually, user generated data. Today’s predictions rely heavily on prediction markets, including Intrade, but will expand as the data becomes more available in the coming days. The full methodology is the same as I utilized in predicting elections, just adapted for the Oscars. Of course, one key difficulty relative to politics is that a much smaller, exclusive group of voters decide this election and we have no polling data, which proved so prescient in the 2012 general election. As with my political commentary, I will leave the pontificating to others; this column will focus solely on what the data is saying about the likely outcomes.
Prediction markets are markets where knowledgeable users can back up their convictions over upcoming events with either real money or other meaningful rewards. Contracts on the outcomes of events are worth either a dollar if outcome occurs or nothing if does not occur; thus, the price of the contract is a strong indicator of the probability of it occurring. If the price for a contract that could be worth a dollar crosses ninety cents, users are very confident the outcome will occur, but if the same contract trades for pennies, users are very confident it will not occur. Users trade on many types of information and the markets are an efficient way of aggregating this dispersed information among dispersed users. Many academic papers have confirmed the value of prediction markets in forecasting upcoming events, especially political election ranging from the late 19th and early 20th century elections through the 2008 election. Further, prediction markets did very well in predicting the 2012 primary and general election and, more relevantly, both the 2011 and 2012 Oscars.
Fundamental data for movies includes categories such as: budget, release date, genre, gross revenue, average gross revenue per screen, ratings, etc. This is similar to the models in politics that focus on: past election results, economic indicators, ideological indicators, and biographical information about the candidates. Fundamental models serve two purposes. First, they provide baseline predictions of upcoming events. Second, they provide insight into which variables we should follow. For example, should we follow weekly, average, or aggregate gross revenue? My work shows that there is particular predictive power into the how much revenue the movies grosses in week 5 versus week 4. Fundamental models did very well in predicting the 2012 general elections.
User generated data will launch later this month to include input from people like you. First, I want to give users the chance to add to the data by pushing my predictions up or down. If the crowd consensus is in a unified direction, I will consider that a valuable indicator. Second, I want to expand my reach into all of the different categories; in particular those categories where the traditional data sources are very thin.
We expect our predictions to change in the coming weeks as events unfold. First, there are awards shows to happen: Golden Globes, Screen Actors Guild, etc. Second, there is data to still compile as some of the top movies this year have only been in release for a few weeks. Third, there are rumors that have yet to circulate and last minute pushes that have yet to play out. Stay tuned!
This column syndicates with the HuffingtonPost