Computers predict good movies

StaringatComputerScreen.jpgYes it’s really come to that, this either means that the Execs who did this before in Studios are idiots, or that Computers are now so powerful that they can predict movies which will be a success. Of course, there’s another option. It’s rubbish and doesn’t really work.

From New Scientist – which yes I do read regularly and even buy a paper copy of! – comes the story of a man who has developed a Neural Network (a computer system capable of mimmicking the brain’s operation) that he claims can predict successful movies.

The idea comes from Ramesh Sharda, an information scientist at Oklahoma State University in Stillwater, who has trained an artificial neural network to recognise what makes a successful movie…

…Using data on 834 movies released between 1998 and 2002, Sharda found that the neural network can judge a film based on seven key parameters: the “star value” of the cast, the movie’s age rating, the time of release against that of competitive movies, the film’s genre, the degree of special effects used, whether it is a sequel or not, and the number of screens it is expected to open in. This allowed it to place a movie in one of nine categories, ranging from “flop” (total takings less than $1 million) to “blockbuster” (over $200 million).

The system cannot take into account the intricacies of the plot, but Sharda says it can nonetheless get the revenue category spot-on 37 per cent of the time, and correct to within one category either side 75 per cent of the time. This is enough to make the system a “powerful decision aid”, Sharda says.

Oooh right. So it’s “spot-on” 37% of the time, and close 75%….ermm, is that me or does that add up to 112%? Wait though, looking at the categories involved it’s analysing cast, special effects, screens, rating, release, genre..okay, apply that to the Matrix trilogy and we have a surefire hit franchise! The previous Batman movies? Anymore that just makes this look a bit limp?

The idea is now to make it into a website, add DVD sales and presumably sell it to the studios…geez, they’ll make anything, perhaps they’ll buy anything!

Comment with Facebook

5 thoughts on “Computers predict good movies

  1. They have already been using technology like this in the music industry for years. They use complex algorithms to analyze melody, rhythm, hook etc.. It works wonderfully for a few producers, but ruins most music, making it all sound the same.

    They’ve also been using it informally in movies and TVs with the studio execs using focus groups, screening previews and giving creators notes like “Add a wacky neighbour, our marketing research shows that people love them”. It’s inevitable that technology would replace this process.

    Certain big hits like Seinfeld’s TV show can never be predicted as they go against every rule in the sitcom book. Never fear, certain networks will want to be seen as creative and ignore such technology.

  2. “a powerful decision aid”

    On what? What you should and/or should not see?

    By the way, Ramesh Sharda’s idea isn’t a new one. It is also, like Viacom’s failed “reverese engineered” movie hypothosis back in ’96, extremely flawed. You recall Viacom’s ‘study’ don’t you? No? Well, it was, based on marketing surveys, what determines scripts and stars.

    In both cases, it simply does not work, because the biggest problem in entertainment marketing is that people generally don’t they want until they see it. In addition, if Sharda admits it is ‘spot on 37%’ of the time, well, I’m sure that folks would line up around the block for ‘Star Wars’ or ‘King Kong’. People will go see Tom Cruise. But it’s still 37%- with the rest up in the air for a margin of error. If the studios were to buy into this, if they thought the summer of 2005 was bad, they haven’t seen nothing yet.

    Let’s say the major studio execs thought this was a killer idea. Well, nearly everyone uses the system. Then they are shocked when thier films are blown out of the water by “that” film which counters that marketing ploy. Consider one of the biggest hits of last summer and how well this “prediction” meter would have “rated” it. It had no actors, no special effects,it wasn’t a sequel, and by demographics, the kids should go for the cartoons instead and it’s a documentary. “March Of The Penguins” woud not even be on this systems’s radar.

    How about last year’s Oscar winner “Million Dollar Baby”? Would it rate high, you think. You’re tempted to say yes- but bear in mind Clint Eastwood had to go round Warners for financing as well as pushing a late 2004 release date. For that matter, how about Eastwood’s “Unforgiven”? The age demographics suggest ‘nobody should be interested in Eastwood anymore; look at Harrison Ford’. So, here you have something of an anomoly:Movies made not based on factors on second guessing the audience, movies made where the talent stands up for the project and refuses to cave in to political correctness, or worse, the studio mental state.

    Last year, we had “Open Water”. “Passion” and Micheal Moore’s film. Under this marketing guesswork, telling audiences what we should and should not see, why did we go to see these films? Are these pictures in that other ‘percentage’?

    “The system cannot take into account the intricacies of the plot”<—really?

  3. I agree, sounds hideous. But I think to be fair, “correct to within one category either side 75 per cent of the time” would include the spot-on correct 37%. Now, how do I do a nerd smiley?

  4. Wow isn’t technology great. Someone has invented the virtual fuckwit, sorry studio exec.

    Oh yes this bodes well.

    Wouldn’t it be cheaper and quicker to train chimps to do this, then again judging by Stealth this has been tried already.

    Looks like in Oklahoma Stillwater does not run deep! (apologies, I just couldn’t resist that one)

Leave a Reply