Skip to main content
/technology
  • E-mail
  • Save
  • Print

Netflix offers $1M coding prize

  • Story Highlights
  • More than 27,000 contestants from more than 160 countries have entered contest
  • At least $50,000 a year will go to whoever comes closest to 10 percent improvement
  • First "progress prize" goes next week to AT&T team
  • Some of AT&T team's suggestions to be incorporated into Netflix software next year
  • Next Article in Technology »
Decrease font Decrease font
Enlarge font Enlarge font

SAN FRANCISCO, California (AP) -- Netflix Inc.'s system for recommending movies to its subscribers is proving tough to beat, as thousands of engineers and statisticians have discovered after long hours of research.

art.netflix.afp.gi.jpg

Netflix leads the online DVD rental market.

Hoping to widen its lead in the online DVD rental market, Netflix last year dangled a $1 million reward to anyone who could improve upon the Los Gatos-based company's current movie-recommendation software by at least 10 percent.

The prize remains on the table, Netflix said Tuesday, for the more than 27,000 contestants from more than 160 countries who have entered the quirky coding contest so far.

To keep the programmers motivated, Netflix is handing out at least $50,000 annually to whoever has come closest to the 10 percent improvement.

Netflix will give its first "progress prize" next week to three researchers at an AT&T Inc. laboratory in Florham Park, New Jersey.

The team, consisting of Yehuda Koren, Robert Bell and Chris Volinsky, spent more than 2,000 combined hours poring through data to develop a method that improved upon Netflix's movie recommendations by about 8.5 percent.

The judging was done using a program that quantifies how well the recommendation systems predict which movies will be liked or disliked by a profiled consumer.

Because the AT&T trio did much of their Netflix work as part of their job assignment, the $50,000 award is being paid to their company, which plans to donate the money to charity.

Although his team is near the 10 percent target, Koren said he isn't sure the goal is reachable.

"Getting the first 5 percent (of improvement) is a lot easier than the last 5 percent," Koren said.

Even if they don't continue their work on the Netflix project, the AT&T research team plans to release their findings next week. Their insights conceivably could help other competitors in the contest to win the $1 million prize.

Netflix plans to blend some of the AT&T research team's suggested improvements into its movie-recommendation software early next year, said Jim Bennett, a vice president who oversees the system.

The current system draws upon about 2 billion movie ratings collected from the more than 10 million consumers who have subscribed to Netflix's service at some point during the past eight years.

Netflix ended September with about 7 million subscribers, giving the company a substantial lead over its nearest rival, Blockbuster Inc., which has 3.1 million online customers. E-mail to a friend E-mail to a friend

Copyright 2007 The Associated Press. All rights reserved.This material may not be published, broadcast, rewritten, or redistributed.

All About Netflix Inc.

  • E-mail
  • Save
  • Print
Today's Featured Product:
Fujitsu Lifebook T1010
 6.6 out of 10
Recent Product Reviews:
Griffin iTrip Universal
 5.3 out of 10
Philips NP1100/37
 7.0 out of 10
Fujitsu LifeBook A6120 (2.26GHz Intel Core 2 Duo P8400)
 7.1 out of 10
Home  |  World  |  U.S.  |  Politics  |  Crime  |  Entertainment  |  Health  |  Tech  |  Travel  |  Living  |  Business  |  Sports  |  Time.com
© 2008 Cable News Network. Turner Broadcasting System, Inc. All Rights Reserved.