Netflix names $1 million prize winner
DIGITAL: Group awarded for improving company's recommendation system
By Danny King -- Video Business, 9/21/2009
SEPT. 21 | DIGITAL: Netflix declared a group led by AT&T researchers the winner of a $1 million grand prize for creating an algorithm that improved the company’s user-generated recommendation system by at least 10%. The largest U.S. movie-rental service via mail also launched a second version of the contest, which also will involve a $1 million prize.
The winning group, called “BellKor’s Pragmatic Chaos,” beat out more than 40,000 teams from 186 countries to win the prize, which stemmed from a contest started in October 2006, Netflix said in a statement today. The seven members of the winning team, which were from the U.S., Canada, Israel and Austria, had previously been part of three competing groups before merging for the winning entry, which was submitted less than a half-hour before the contest ended July 26, Netflix said.
Netflix started the contest three years ago to improve the accuracy of its user-generated movie recommendation system. The company’s search engine processes more than 2 billion user ratings to make recommendations for a particular subscriber based on how that subscriber has rated other movies.
“Accurately predicting the movies Netflix members will love is a key component of our service,” said Dr. Neil Hunt, Netflix’s chief product officer. “This extreme level of personalization is like entering a video store with 100,000 titles and having those that are most interesting to you fly off the shelves and line up in front of you.”
Netflix paid out $50,000 in progress prizes for both 2007 and 2008. BellKor’s two members from AT&T, Chris Volinsky and Robert Bell, both were on the teams that won the progress prizes, AT&T said in a separate statement today.
The race was so close between the BellKor team and the runner-up group, called “The Ensemble,” that Netflix’s judges required “several weeks” to verify the winning entry, the company said today. Blog HackingNetflix.com declared the BellKor team the winner in late June before reporting the following month that The Ensemble edged out the BellKor entry.
The new version of the contest differs from the one just decided, which involved predicting movie preferences of people who’d submitted ratings for at least 50 movies. The current contest is challenging teams to improve the recommendation system using demographic information of Netflix subscribers who rarely rate movies.
Because a specific improvement goal was not determined, Netflix will award $500,000 to teams with the best algorithm both six months and 18 months after the contest launch, the company said today.
