Predicting Movie Hits

Does anyone really know anything?

There is a famous line from the screenwriter William Goldman: nobody knows anything. People repeat it constantly. Almost nobody checks it.

For forty years, economists, marketing professors, statisticians and computer scientists have tried to predict how much a film will earn, using budgets, stars, scripts, trailers, tweets, search data and, lately, large language models. This guide collects that work in one place. 

For each study you get the people, the data, the method and the result, all the technical detail kept in but explained in plain English, with a short, honest verdict on whether it actually works.

Inside, in eight parts:

1. Why nobody can predict a single film. The mathematics of hit-driven markets, and why the usual forecasting tools break.
2. How studios win without picking winners. How Hollywood has ruled world cinema for a century despite every film being a coin toss.
3. The first attempts to forecast box office. The founding regression studies of the 1980s and 1990s, and what they genuinely found.
4. Predicting a film from its script. What story analysis can and cannot see before a film is made.
5. When machine learning had a go. Neural networks, headline accuracy claims, and what survives scrutiny.
6. Reading the buzz before release. Tweets, blogs, reviews and search data as crystal balls.
7. Judging a film by its trailer and poster. The newest wave: feeding the film's own raw material to the machines.
8. Why European films struggle to travel. The thread that connects prediction research to the wider story of European cinema.

Every study is fully cited, and every verdict is written in plain English.