I realize I'm a bit LTTP but I just finished reading Malcolm Gladwell's article in the October 16th New Yorker. "The Formula" relates the story of a man obsessed with films who turned his passion into a company called Epagogix which will predict, by amassing variables and analyzing patterns, how much a script can make at the box office.
What's crazy to me is not that it works, but that, near the end of the article, it comes out that the analysis still depends on real people reading and grading scripts (based, I assume, on their own metrics) - that essentially, it still depends in some ways on their taste, before it gets fed into the neural network computer. Hence it's important that one founder prefers Alien while another likes Yojimbo. I'm not saying taste should be unimportant - I guess I'm saying that after all, it is still important.
But let's leave that aside for a moment and think about what it would take to build a hit predictor machine for videogames. What factors would you start including? Can the Epagogix model - whatever it is, since we don't know for sure - be applied to interactive entertainment?