Talk about a photo finish... of course, nothing is decided yet, it will depend on the performance on the final test. I think its possible that one of the team overfitted by submitting tests every day and adjusting accordingly; it might perform worse on the final test for this reason.
I hope they'll simply restart the contest with the new record holder as the benchmark and another 10% improvement (or 11% to keep it linear and to avoid getting into deminishing returns on investment territory).
Are you suggesting that the next contest should improve the winner of this contest by 10%? That seems like you're already well into diminishing returns territory.
Maybe not, if they offered a more descriptive dataset next time. I think the only data offered was (customer_rating, movie_id, rating_value, date). If they included things like the customer's age, how long they've used Netflix, and other such data that's on file, I'm guess that the RMSE could be improved by a fairly good amount (at least 5%, if not 10%).
11% of the 90% they've got now is 10% of the original (or close enough for government work).
According to Yehuda Koren, BelKor et al. won the prize after all. We'll have to see tomorrow. Surely there will be some sort of press release.
So the team called "Grand Prize Team" didn't win the grand prize? Hilarious.
"The Ensemble was created by a merger between the two teams that appear on the leaderboard as "Grand Prize Team" and "Opera Solutions and Vandelay United"."
And no one has officially won as the scores on the test set are not public.