Box-Office Science

Imagine a world where Hollywood producers could predict, with scientific precision, the box office revenue a movie will generate just by reading the screenplay. A new forecasting model devised by a trio of marketing professors from Wharton and NYU promises to deliver something like that. Among their findings: action movies with multidimensional conflicts are the most surefire investments, and horror films the riskiest. In a paper describing their model, they write that “a higher movie budget tend[s] to increase the movie’s box office, but at a diminishing rate. Thus, throwing money at a script does not always generate a blockbuster movie.” Here’s looking at you, Waterworld. [%comments]

Drill-Baby-Drill Drill Team

Ironically our Science Space Program is going in the Dumpster, but our Sci-Fi Movies are having a Renaissance: Bigger, Faster, and More Imaginative Block Buster Summer Films.

A Baby born today is more likely to Star as a Captain of a Sci Fi Space Ship than to be an Astronaut and go on an actual interplanetary voyage.

Be William Shatner instead of Neil Armstrong. I'll be Luke Skywalker. A light saber slices a slide rule into halves.


Just what the art of cinema and our cultural landscape needs... not!


The book Hollywood Economics: How Extreme Uncertainty Shapes the Film Industry, which Taleb discusses in his book, is a great read on the topic of Movie success predictions. In it, Dr. De Vany quotes William Goldman, who stated that "nobody knows anything" about what happens to movies once they are released to the theatres.


The problem is, once they discover this rule/formula/equation, every movie studios will move to adopt it, thereby rendering it invalid.


My faith in this paper is limited. It seems to be a case of "dazzle with statistical methodology". Why use a BART model? Much simpler (wider known) methods could have been used to establish the findings. This is a tendency in many disciplines.

I would propose the following mandatory rules for publishing papers:
1) Use simple models first and SHOW the results of these models. SHOW us the basic stuff - simple regressions with scatterplots
2) Motivate the use of more complex methodology in view of these results. A simple "XY" has been shown to outperform "classical methods" is NOT enough.
3) Provide not only the DATA set, but also instructions and all necessary programs, batch files to replicate the results.
- Martin

Kevin C.

Didn't Waterworld actually make a pretty sizable profit, despite underperforming expectations at the box office?

Surely there's a better example to use...


Isn't the use of VaR models combined with extrapolating from a small data set how quants crashed the financial system?


My sister works in hollywood doing film production, and has been for a number of years. If you name an actor, she can give you a good guess on their approximate cost, and return on investment.

I know when I asked her a few years ago, Will Smith was the best return on investment. That did not make him cheap.

However, I've also heard that he did a more subjective analysis of top films to figure out that fun sci-fi special effects fests were the films that tended to do the best.


Sounds dubious. In my opinion the best (and most successful) movies are those who do something radical different, satisfy the need for something new and then change the taste of the audience.
Also I'm not really convinced by the results. They only used scripts for released movies, so the data was biased. And the successful genres are the easiest; family movies and comedy. These movies have lower budgets and a straight pattern, so it's easier to filter bad ones and harder to make mistakes in the production phase.

Drift Marlo

Totally impossible and also about the dumbest thing I've ever heard. Considering the words in a screenplay are, maybe, a third of the experience (and the most transient in the process) while hearing it and actually seeing it (OMG, what!?) make up the bulk, this proposal it utter nonsense. If this metric was at all relevant it wouldn't matter who was in a film, who directed it, who edited, scored or art directed - IMHO said film would be unwatchable thereby invalidating the whole point. A lot of great movies would be scrapped in favor of a lot of useless ones (we'd lose the 2001s in favor of Titanics) because there's simply a lot more to a film than a screenplay. Hell, some great films never really had a screenplay (La Jetee).

Andreas Kitzing

Already heard about a pretty precise formula to predict box office revenue 2 years ago in my media management course at the University of Hamburg. It mainly checked for variables like star power, marketing budget, genre, target group, etc...
If you want to know more about that, you should contact your colleague Michel Clement at the University of Hamburg.

Alexander Bender

I am doing a project quite similar to this. I am creating an econometrics model that attempts to predict the number of Hires of a DVD at the local rental store. I have data from the business, and I am using quite a few similar variables. Will be good to reference some of this in mine I think :D

Joe Young

Waterworld did OK and while no cinematic classic is not a bad film. It was killed by bad industry buzz before it was even released.