I believe that the Netflix Prize was fundamentally flawed. They have access to much better information than user rankings, especially for the online service, things like "started but not finished" or "watched all of it the day it came out" and "watched more than once".

That's much more honest feedback than user's star rankings, which can be skewed by a lot of factors: users failing to rate things, rating based on what they feel they should like rather than what they actually enjoy, the psychological anchor of the suggested ranking. It's not surprising that their existing work got most of the value out of that data.

They have much better data now, and I'm sure they're not letting anybody get so much as a glimpse.

Looking at the link. It seems that it was beaten and someone received the money. Now, this is my reason for posting my idea here : it's to get feedback on it. It seems to be rather unlikely that I would come up with something better than teams who are payed to come up with those recommendation algorithms.

However, what do you think of the proposed idea. Maybe in theory it looks better but until I A/B test it I might never know if it's really better than what's currently out there.