Movies, math and money: Waterloo grad predicts box office revenues

What would it be worth to be able to predict how a new movie will do at the box office before it’s ever released?

Waterloo math grad Jack Zhang says the information is gold for the film industry so he founded a startup, called BoxOfficePrediction. His company is based on a mathematical model he created that predicts the minimum amount of revenue a movie will generate in the first week it’s released.

With BoxOfficePrediction, Zhang, a graduate of Waterloo's mathematical economics program, is entering the emerging field of box office analytics which, much like baseball and sports analytics, can be used to predict hits - and flops.

Zhang, who can’t divulge the specifics of his model, says it uses variables like marketing expenditures and natural language processing. He posts his predictions online in the comments section of the Hollywood Reporter so there’s a public record of his predictions.

Predicting box office successes

So far, the model has been correct 18 out of 20 times. For example: 

Boxtrolls

• Prediction: 85 per cent chance of earning $10 million+
• Actual First Week Revenue: $20 million

Tusk

• Prediction:  85 per cent chance of earning $450,000+
• Actual First Week Revenue: $1.1 million

Planes: Fire and Rescue

• Prediction: 85 per cent chance of earning $16 million
• Actual First Week Revenue: $25 million

Nut Job

• Prediction: 85 per cent chance of earning $22 million
• Actual First Week Revenue: $27.9 million

Planes

• Prediction:  85 per cent chance of earning at least $27 million
• Actual First Week Revenue: $32 million

That Awkward Moment

• Prediction:  85 per cent chance of earning at least $8.9 million
• Actual First Week Revenue: $11.3 million

“One of the challenging things about model theory is trying to identify which variables you use,” explains Zhang. “In general, we look at how much production companies spend on marketing and advertising. Also, there is a lot of information online that wasn’t available previously thanks to websites where people comment and share their opinions.”

Zhang developed the model by combining various bits of code from modeling software that would allow for the isolation and comparison of data.

Statistically sound methods 

“Jack is doing interesting work combining statistically sound methods with innovative ways of gathering explanatory data to develop predictive models for box office performance for new films,” said Paul Marriott, professor in Waterloo’s Faculty of Mathematics.

Zhang’s model has caught the eye of the film industry. After attending a pre-event mixer for the Toronto International Film Festival, Zhang has connected with the Canadian Film Centre accelerator program, ideaBOOST, which acts as a bootcamp for companies looking to bring new techniques and technologies into the entertainment ecosystem.

Zhang explained that while the model itself is great at predicting success, it isn’t a substitute for creativity.

“Industry professionals have given me mixed reviews about whether or not a model like this should be used. Some said it was a great way to help determine which movies should be produced. Others said that the idea of using a model to predict which movies are made can be scary,” said Zhang.

He hopes that by working with industry professionals and an accelerator program like ideaBOOST, he can ensure his model will help production companies give people what they want.

Zhang has his eyes set for bigger and brighter horizons. “I’d like to use the model for consulting – to provide competitive analysis,” said Zhang. “It would also be great to produce movies myself.”