Can Software Predict A Hit Movie?
Apparently yes. I found an old article by Malcolm Gladwell that looks into how computer software can accurately predict if a movie is going to be a hit based solely on the screenplay. This article is long, but very interesting. http://www.fliquesosman.com/?p=676
My impression upon reading it: the neural network is only as good as the data you feed it, and the network seems to reflect the tastes of the two or three guys that review and grade/code each script, for good or ill. When you change out the readers, the expert system or neural net or whatever you call it, will start to break down or at least change significantly in how it interprets the samples given. This is why I liked watching Siskel and Ebert: though I am almost 97 percent alike in tastes with Ebert, I would find that the movies I liked best were one where both reviewers strongly agreed, and sometimes I didn't like a movie where they strongly diverged, even if one thought it was a a potental Oscar-winner. Wtachign any other reviewer, I couldn't match as well with their individual tastes. A Kenneth Turan, for example, often seems not to "get" things I saw in a movie and liked. he certainly "scoers' them differently.
I am sure that screenwriters and producers in development apply this kind of formula thinking, and have for a long time. They say there's only what, ten basic plots in all of film making, and the art is in how the pieces are arranged and executed. A script is easy to quantify in terms of things like word count, grade level of the dialog, even the "beats" can be counted and you can get a sense of how many pages and minutes it takes to deliver each beat and each peak and valley of a plot. Writers have, to more or less of an extent, tried to use such formulae in their script construction for decades. I mean, you can buy at least twenty books today that promise you a magic formula for writing a guaranteed screen hit. They all work along similar lines.
I think I still agree more with Goldman. Nobody knows anything, not for sure. You can try to hedge your bets by playing the odds, and using tools like this neural net *might* help, but they are only very rough tools that are themselves making educated guesses based on pasdt performances. They don't have the intuitive spark that can appreciate the oddball idea out of left field that goes on to become an unlikely blockbuster. As an executive, you can use these tools to play your development line like a low-risk investment banker, or you can swing for the fences.
The last thig I think about on this is: if these neral nets are sucessful, it really kills film making, even as it becomes more profitable. It's the cinematic version of the free will/determinism debate: why should I care *what* I do, why care to make any effort in a world where everything is already locked-in as destinty and God overrules my every decision or desire? If you really can reduce film making to a soulless process of "quants", you've killed the art of it, and moer to the point, killed off any chances for art to get greenlit in the first place.
"Nobody knows anything" is just the way we like it. Because the potential is limitless that way.
[Mark Suszko] "they say there's only what, ten basic plots in all of film making, and the art is in how the pieces are arranged and executed."
Couldn't agree with you more, a formula can get you only so far.