Netflix's $1 million contest was launched almost three years ago in October 2006. The aim of the game was to develop a recommendation system that was more accurate than the Cinematch system Netflix already uses. As you can imagine, $1 million isn't exactly pocket change and, while there was probably a lot of folks who dreamed of taking the prize home, it took a team of scientists several years to get to the finish line.
NYT reports that Netflix formally announced the winners yesterday afternoon. Dubbed BellKor's Pragmatic Chaos, the group of seven statisticians, machine-learning experts and engineers hail from the U.S., Austria, Canada and Israel. The announcement isn't without drama, as there were actually two winners who, once they crossed the finish line submitted even, tied results.
A couple of months back, Netflix announced that BellKor had developed software that was the required 10 percent more accurate than Cinematch. As per the contest rules, there was then a 30-day period in which other contestants could try to beat BellKor and one other team, Ensemble, came incredibly close.
While Netflix publicly said the difference between the two was too close to call, NYT reports that it privately told BellKor it had won. Yesterday Netflix CEO, Reed Hastings, explained exactly what had happened.
The contest rules state that in the event of a tie, the team who presents its submission first, wins. The BellKor team members submitted theirs just 20 minutes before the deadline but at the last second, Ensemble presented its own submission.
“That 20 minutes was worth $1 million,” Mr. Hastings said at a press conference.
Read the full story here.