It’s certainly an admirable goal for Netflix to try to improve its movie recommendations, and an even more admirable PR move, as megamark sagely points out, but I do have to wonder how much headway can actually be made. In my last entry, I pointed to Alexander McCabe’s blog Flaunt It, in which he analyzes the ratings system and Cinematch’s accuracy:
The ratings are between 1 and 5 stars. Their current system ‘Cinematch’ doesn’t do too well. On average it’s out by just under 1 star for each rating - so typically Cinematch will predict that a viewer will watch a movie and score it 4 stars - the viewer will actually score it 3 or 5 stars. You could probably get close to that level of competence by guessing that the viewer will rate every movie at 3 stars. Many times, you’d be right, and many times, you’d only be one star away. The vast majority of movies that I’ve seen are 2, 3 or 4 stars.
The first problem, as I see it, is that we are not dealing with a particularly accurate ratings system to begin with. A margin of error of 1 star doesn’t sound like much, but 1 star does constitute 20% of the ratings graph, and could be the difference between “I really enjoyed it” (4 stars), or “I wasn’t overly aware of the posterior pain of sitting in a chair for two and a half hours” (3 stars), or from the latter to “I was kind of bored after I ran out of popcorn” (2 stars).
So, naturally, there is room for improvement in terms of ratings predictions. But the ratings themselves are given by human beings, which I must confess, despite my fondness for them, are not renowned for consistency. I could rate a movie down by a star if I were feeling grumpy either the day that I watched it or the day that I rated it, or rate it up by a star if I watched the movie at a get-together with friends, forgetting, perhaps, that most of the witty dialogue was supplied for the people on-screen by my friends. And I do not think I am markedly more capricious than the average movie-viewer.
And ’stars’ are about the most subjective rating system after ‘thumbs up/down/horizontal.’ Perhaps 2, 3 and 4 stars have an entirely different meaning for you than for me, and you would call a 3 star movie “I was kind of bored after a ran out of popcorn” and a 2 star movie “I did escape from the experience without gouging out my eyeballs.” What does it mean to be a four star movie? A three star movie? How long will we ponder this question before it starts to become slightly existential?
All this is to say that Netflix knows me about as well as I know myself, and I’m not exactly certain how they can provide a much more accurate prediction. An article in the Times-Herald Record jubilantly opens, “Does it sometimes seem as if Netflix knows your tastes better than your friends, your spouse, even you? ‘You loved it,’ the DVD rental site will remind you, offering a movie for sale that yes, by golly, you did love.”
The news item here is that, while Netflix bides its time waiting for a recommendation engine to tell you how much you loved it with an even greater degree of accuracy, the company has also introduced a new feature in which you can find out what other people loved, too. Just put in a zip code, and you can find out what people in New York, San Fransisco, Portland, Maine, or Portland, Oregon, or, for that matter, Hindustan, Indiana (actual town) are watching and enjoying from Netflix.
An important distinction on the service is that it does not display the movies rented most often, but the titles ordered “much more than other Netflix members” in other or comparable areas. This reveals that regional specialty tastes frequently have to do with (surprise) the region itself. For example, “Northern Exposure,” about the little town of Cecily, Alaska, tops the list in Fairbanks and Juneau, “Midnight in the Garden of Good and Evil” is number two in Savannah, Georgia, where the movie (and book) is set, “Wyatt Earp: Special Edition” is particularly loved by those in Phoenix, near Tombstone, Arizona where it was filmed.
New Yorkers love Ric burns’ historical series “New York,” “New York Stories,” and Woody Allen’s “Manhattan,” but they also display a particular weakness for French films, as opposed to, say, all the Cuban films that are rented in Miami. And regional favourites are not a given: “Fargo” is not on the list for Fargo, North Dakota, nor is “Paris, Texas” on the list for its namesake.
The Times-Herald Record also noted that regional placement made little difference in their own research of “top rentals” rather than “top unique rentals.” For example, “Crash,” though set in L.A., also hit No. 1 in Boston, New York, Miami, Albany, NY, and Billings, Montana. There were a few interesting variations amongst the cities they researched, which you can read about in the full article, but not ones that reflect an easily-definable trend.
After all, no matter how many stars they give it, the capricious human being does have remarkably similar affinities for certain kinds of stories, regardless of where they live. “I think taste is uniform across the country — and across the world, and the millennia,” said Richard Walter, a film professor at the University of California, Los Angeles. “People everywhere respond to the same kind of drama, the same characters, the same conflict.”
So, as Greg Gershman wittily suggests on his blog, gregword, you could probably please everyone by suggesting ‘The Shawshank Redemption,’ regardless of which movie they’d just rated, or how they rated it.