View Full Version : Can computer models help select better movie scripts?

12-02-2006, 11:06 AM
An article discusses this here (http://knowledge.wharton.upenn.edu/article.cfm?articleid=1609).

12-02-2006, 12:23 PM
Sure, but can a computer style Barbara Streisand's hair?

12-05-2006, 11:20 AM
As much as the idea of reducing story and art to computer models reviles me, I can see this being potentially useful.

I say potentially, because in all likelihood if this software was of any quality, it would be used in the wrong way- by becoming the primary decision point on whether or not to greenlight a movie. Whereas I could see it being used as an analytical tool. And not necessarily only greenlighting high-scoring movies, because you would have to compare against similar movies, and then compare box office results, etc.

The other thing is, if the software actually was used, people would do everything they could to 'game' the software- basically shoving in everything the software model is looking for, even at the expense of the story. Kind of like trying to game search engines or trying to outwit SPAM email filters...

12-05-2006, 12:36 PM
Kelly LeBrock was a computer model in Weird Science. She sold it for me. :D

12-06-2006, 09:59 AM
My problem with this is the program analyzes scripts against past successful films.

But what about groundbreaking stories which some exec took a risk on even though it had no similar predecessor?

And does it factor in trends. Some films do better because they're released during a trend toward that particular genre - released a couple years later, the same film may not have done as well.

And - as the article says - it doesn't factor in stars. Perhaps a film did well because the hot-at-the-time star drew BO in spite of a thin story.

I think this program is bad news for everyone on the creative end of the film-making process.

But I can see how the marketing dept. would love it.

Pull Back Reveal
12-06-2006, 03:35 PM
sc - I think the reason the program doesn't account for the elements you cite is because they don't matter -- not for the purposes of the program. That's why it's super-secret, very expensive and produces surprising results. If it told everyone what they already knew, would it be worth a king's ransom?

It's a radical way of trying to get to the DNA of films (or pop songs) strictly to predict success in the marketplace. It's not a tool for measuring aesthetic success, or for predicting success with agents, producers, studios or attachments -- all the things we're probably more concerned about when writing a script.

What's remarkable about the program's conclusions is that broad strokes, basic set-ups are what matter -- the die is cast at the concept and outline phase, not when you're worrying over the wording of a 30-second dialogue exchange. (Of course, those details will matter to everyone else considering your script.)

sc, your points -- I think the program's authors would suggest there really aren't any groundbreaking stories (we hear this again and again from gurus going back to Heraclitus), but let's examine a movie considered groundbreaking at the time. When SEVEN was made, it was made with difficulty because Walker's script was considered too dark, too disturbing. But the DNA of the script might have predicted its break-out success: strong sense of place, black-and-white partners, life-and-death conflict, a romance at the heart of the story, everything pushed to the limits. (Those are points mentioned in the article that spell success.)

I'd be curious what groundbreaking stories you're considering here. MEMENTO, BEING JOHN MALKOVICH were huge failures with the public (what this program measures). Again, it doesn't say anything about aesthetic success or stature among your peers or landing future jobs.

Trends? Hundreds of movies try to jump on (or fortuitously benefit from) trends, but more fail than succeed -- is there something about the story that determines which catch hold? And would the successful members of the pack have been successful anyway? I imagine the authors would concede a trend-follower might get a brief, first-weekend bump depending on the environment but that the ultimate success lies in the story, and that this is born out in any example you can think of.

Stars? The dvd remainder bins are full of failed star efforts. It's a truism that movies make stars, not vica versa, and this program seems to back that up. Does that have anything to do with whether your script gets made? Obviously it's easier if a star is attached -- again, not something the program cares about.

madmax -- I think the whole point is to try to game the software, keeping in mind, if it actually works, you're gaming the box office. How's that any different than trying to guess what works without any quantitative way of figuring out if you're right? We're already trying to game the unwritten software that runs Hollywood decision-making, and all along the way, everyone's trying to "game" the results.

One thing to keep in mind, according to the description of this program, different things work, it's not trying to make every movie into the same movie (studio tentpole strategies to the contrary). What it suggests is, different things work better than others, so why not chose the elements that lead to success rather than the ones that guarantee failure? This is, after all, a business, not an art-school project.

It would be useful to find out which studios and producers are using this service, so it'll make more sense when crazy notes start arriving from left field. Or maybe not.

Laura Reyna
12-07-2006, 01:13 AM
Here's a similar article by M Gladwell.


Is this a different program, or what?

So is every nerd developing his own program? What ever happened to wanting to beat the house in Vegas? :cool: