In October 2006, Netflix announced it would give a cool seven figures to whoever created a movie-recommending algorithm 10 percent better than its own. Within two weeks, the DVD rental company had received 169 submissions, including three that were slightly superior to Cinematch, Netflix's recommendation software. After a month, more than a thousand programs had been entered, and the top scorers were almost halfway to the goal.
We've entered the long twilight struggle of the Netflix Prize. "The last 1.5 percent is going to be harder than the first 8.5 percent," Potter tells me. In the past three months, BellKor's score has barely budged and now stands at 8.57 percent. Potter, meanwhile, is at 8.07 percent, and his pace has slowed, too. It's entirely possible that neither will ever make it to 10 percent. After all, there's a certain inherent variability to human choices that even the savviest computer can't predict.
