Can computers predict which movies will flop?

[Brown Jenkin]

For one of them, it's entirely explicable. Penguins man! Penguins! They just charm the socks off of people. Don't be surprised if Hollywood catches on, and instead of having big special effects budgets, they start having big penguin budgets.

I have to agree with you. There is the story of Marry Poppins where Walt Disney was running some test screenings. One of the replies he got from the kids in the screening was "Needs Penguins." Walt then ordered the movie back and Penguins were added. I'm not saying this is what made the movie a success, but Walt thought it would help.

 

[Thornir Alekeg]

Looking at the 2002 paper, it does not contain a whole lot of surprises. Number of screens released, month of release and "star power" have the largest influences on the total box office numbers.

Here are his ranges for numbers (in millions):

< 1 (Flop)
1- 10
10-20
20-40
40-65
65-100
100-150
150-200
> 200 (Blockbuster)

The paper does not say how many movies fall into each category, but I'm willing to bet that it would look like a pretty classic bell curve. In that case, falling only one category away 75% of the time isn't all that impressive if the majority of those 75% come in the center of the curve. The true value of this system would be in its ability to accurately predict the results at the extreme ends of the range. In the 2002 paper at least, this is not addressed.

My conclusion is that anybody with knowledge of the movie industry can probably predict which box office total category (+/- 1 category) listed above any movie will fall into at about the same success rate, especially if you keep picking the center of the range.