NEW YORK – And the winner is ... Netflix.

Netflix Inc. awarded a $1 million prize last week to a seven-member international research group as part of a three-year, intensely waged contest to help the online movie rental company predict more accurately what movies its customers will like.

What Netflix gained from the experience is likely worth more than $1 million, and the company’s launch of a second $1 million contest shows it is well aware of that. In fact, when the contest launched in 2006, the first entrants took just three weeks to improve on what Netflix’s internal team had been able to do on its own.

By identifying ways that Netflix could improve its movie recommendations by at least 10 percent, the winning team – BellKor’s Pragmatic Chaos – is actually letting Netflix make picks that are twice as good as they are now.

That’s important if Netflix wants to retain subscribers and keep them from exhausting their list of movies to watch. With more than 100,000 films and TV shows available, it’s not enough to simply list them alphabetically, or even by genre.

Netflix’s recommendation system is the software version of the video store clerk who’s a movie buff, except the Netflix computer knows your personal tastes and doesn’t pass judgment if “Runaway Bride” is your top pick.

Loved “Monty Python and the Holy Grail”? How about “Clerks”? If you rated both flicks at least four out of five stars, Netflix’s system will likely suggest “Shaun of the Dead,” a British zombie comedy from 2004. It is in Netflix’s best interest that you like it – after all, no one keeps going back for movies they hate.

But Netflix sometimes recommends duds, and incorporating BellKor’s improvements will reduce the chances of that happening and double Netflix’s chances of giving you the right pick.

The Netflix Prize contest was a close call. BellKor’s Pragmatic Chaos and a rival group called the Ensemble each showed a 10.06 percent improvement in movie picks over Netflix’s own Cinematch system.

BellKor was declared winner at an awards ceremony last week because it submitted its final entry just a few minutes ahead of Ensemble. Like in a good cliffhanger, both came dangerously close to the deadline.

For those more excited by algorithms than touchdowns, following the Netflix Prize has been like the Super Bowl. And the winning method could have implications well beyond Netflix recommendations; any business that uses people’s preferences to sell products could learn from the exercise.

Under the rules of the contest, the winning entry will be published at the University of California, Irvine’s Machine Learning Repository, and Netflix will be able to use BellKor’s work without paying royalties. The team is free to license it to other companies, too.

Tens of thousands of people have pored over the problem since the contest began in October 2006, using a database of 100 million real-life movie ratings from Netflix customers, with personal identities removed.

More than 51,000 contestants from 186 countries participated.

“I was stunned at how the Netflix Prize created its own economy of researchers competing and collaborating,” Netflix CEO Reed Hastings said.

He called the contest a “bona fide race right to the very end.” As the race grew tighter, teams began to realize they would get better results if they combined their efforts. In the end, one-time rivals joined forces to form the two remaining powerhouses, BellKor and the 30-some-member Ensemble.

Netflix is now planning a second contest – a sequel, if you will.

The first required contestants to improve predictions for subscribers who regularly provide ratings on movies they’ve watched, 50 movies or more, on average. The second will involve those who don’t rate movies often or at all; that’s about half of the company’s 10 million-plus subscribers.