Skip to: Content
Skip to: Site Navigation
Skip to: Search


Can Computer Help Hollywood Pick Hits?

CLICK HERE FOR 'FLOPS'

By Daniel B. WoodStaff writer of The Christian Science Monitor / January 3, 1997



LOS ANGELES

Everyone in this town knows the Big One is coming. Ears to the ground, they can hear the rumbling, the tremors, the shaking. When will it hit?

Skip to next paragraph

Chris Lanier says he has mastered the technology to tell us. No, not about earthquakes, but rather the other major obsession of southern California: movie blockbusters.

"Being able to predict how much money a film will make - before it's released - is the Holy Grail of the motion picture industry," says Mr. Lanier, a veteran of Stanford University's Artificial Intelligence Project. His computer-based company, Motion Picture Intelligencer, claims to identify cinematic duds before moguls spend millions making and marketing them. In theory, movie studios would spend less, Lanier would become rich, and Americans would have a lot less garbage coming to theaters near them.

Here's the catch: Lanier has yet to make a sale.

In Tinseltown, the idea of using computers to determine "what is art?" leaves many studio executives and producers cold.

Because of this chorus of "Bah, humbug," Lanier three years ago decided it was time to put his assertions on the line. In a public demonstration overseen by the University of Southern California in 1993, he predicted the earnings probabilities of 81 films before they were released in the US.

After reviewing rough-cut copies and attending reviewer screenings for such films as "Demolition Man," "Beverly Hillbillies," "Pelican Brief," and "In The Line of Fire," he faxed his projections to film-industry executives, Wall Street analysts, and the press.

His accuracy rate: 100 percent, 81 out of 81.

Lanier's "category predictions are a substantial improvement over the notion that box-office success cannot be predicted rationally," says Jan de Leeuw, professor of mathematics at UCLA, whose statistical consulting program analyzed the results of the demonstration. "The MPI categories exhibit predictive validity ... sorting potential hits from likely duds."

Six jokes, one pratfall

How does Lanier do it? Because of amount of time, energy, and money he has invested for more than a decade, he is reluctant to give too many clues. Suffice it to say he screens the scripts or rough cuts himself, making extensive notes in personalized shorthand.

If he is evaluating a comedy, he might categorize the movie as one of the following: action comedy, youth comedy, romantic comedy. Jokes? (What kind? Verbal, physical, combination?) Pratfalls? (How many? What kind? Big or little? Where in the film?)

He also attempts to determine whether story-line setups in early acts reach a payoff in the final act or, instead, fizzle out.

Unlike other industry attempts to determine box-office success, Lanier does not look at advertising expenditures, the number of theaters used in a release, time of year of the release, or competition from other movies. He uses the analogy of an auto race.

"There are lots of factors that influence who will win," he says, "including the driver, the track, the weather, the competition. We only tell you what kind of car you have." Lanier crunches the numbers, comparing entertainment factors to the ticket-buying behaviors of past movies.

Studios are then left to their own strategies for deciding whether to make (or fix) a given movie, when to release it, and how much to promote it.