Netflix Gives $1 Million Prize to BellKor’s Pragmatic Chaos

Netflix has awarded $1 million to the team BellKor's Pragmatic Chaos for a 10 percent improvement to its movie recommendation engine...and now there's another $1 million up for grabs.

Back in 2006, movie rental powerhouse Netflix announced a competition: if someone—or a team of someones—could come up with a way to improve the service’s movie recommendation system by 10 percent, they’d wein $1 million. As expected, the competition drew interest from a wide variety of programmers and statisticians—and, naturally, Netflix began to reap the benefits of recommendation improvements investigated by the competitors without having to spend a dime of its own money on additional engineering and research efforts.

However, today Netflix put its money where its mouth was in 2006, announcing that the team BellKor’s Pragmatic Chaos—a collation of three teams that had previously been competing against each other—have won the $1 million prize. The prize was awarded by Netflix CEO Reed Hastings and Chief Product Officer Neil Hunt in a ceremony in New York.

“We had a bona fide race right to the very end,” said Hastings, in a statement. “Teams that had previously battled it out independently joined forces to surpass the 10 percent barrier. New submissions arrived fast and furious in the closing hours.”

BellKor managed to submit their winning solution just 24 minutes ahead of the competition’s three-year deadline on July 26, 2009. Netflix’s panel of in-house and external judges took weeks to evaluate the effectiveness of the submitted solutions.

Given the interest in the Netflix Prize, the company has now announced Netflix Prize 2, which aims to improve the service’s movie recommendation service significantly for Netflix members who don’t rate movies often or at all; instead, the improvements to the recommendation service will have to rely on user demographic information and behavioral cues taken from their use of the Netflix site. Netflix Prize 2 doesn’t set any accuracy goals, but will award $500,000 to the submission with the most improvement after 6 months, and another $500,000 to the best solution after 18 months.

BellKor’s Pragmatic Chaos had seven members—Bob Bell, Martin Chabbert, Michael Jahrer, Yehuda Koren, Martin Piotte, Andreas Toscher and Chris Volinsky—from the Canada, Austra, Israel, and the United States. The awards ceremony was the first time all the team members had met in person. The team hasn’t announced how they plan to split or distribute the prize money.