So what predicts a successful movie? At the bottom of the expanded view is a square blue leaf node. Of the movies that reach this node, 86% will make more than $2 million in their opening weekend. Given that the default rate is just over 50%, this leaf node represents a very successful subpopulation of movies.

What characteristics cause a movie to reach this particular leaf node? Moving up from the bottom, we see that 90% of the actors associated with the movie have an HSX score greater than 11. The HSX is the Hollywood Stock Exchange, a virtual stock exchange where users trade in the stocks for actors and movies. A high HSX rating for an actor indicates popularity, so successful movies have a high proportion of actors with moderate to high HSX scores. Moving up through the other levels, we find that successful movies are generally not documentaries, that at least one actor has an HSX score of more than 36 (a very high score), and that the movie is not a drama (perhaps comedies and action-adventure movies do better at the box office).

It is important to note that this tree was constructed entirely automatically, based on data for which we know the actual box office receipts. Given this tree, we can now apply it to predict the box office performance of new movies, given that we know a few things such as the movie's genre and the HSX ratings of its actors.