Netflix Prize: Was The Napoleon Dynamite Problem Solved? | Hacker News Reader

Netflix Prize: Was The Napoleon Dynamite Problem Solved?

whimsley.typepad.com

I draw a slightly different conclusion than the author.

>I noticed they were all similar in some way to “Napoleon Dynamite” — culturally or politically polarizing and hard to classify, including “I Heart Huckabees,” “Lost in Translation,” “Fahrenheit 9/11,” “The Life Aquatic With Steve Zissou,” “Kill Bill: Volume 1” and “Sideways.”

To me, what all these movies share in common is that they are so original that they are less accessible to the first-time viewer (or, in the case of Farenheit 9/11, completely tied to one's political leanings).

For example, I walked out of Napolean Dynamite and was bored to tears with Lost in Translation. The second time I watched those movies, I laughed my butt off at Napolean Dynamite and I found Lost in Translation to be one of the deepest movies I'd ever seen. Just the thought of Lost in Translation now gives me a yearning heartache. If I used Netflix to rate these movies after my first viewings, I probably would have given them 1 or 2 stars. Now I give them both 5 stars.

(As an aside, I highly recommend all of the above mentioned movies. I'd love to see an expanded list, so if anyone has one, please share. Going by my theory, I bet The Big Lebowski is in there. :)

I want to have an inverse Netflix graph. I want to be able to meet and connect with other cinema lovers based on their ratings of movies I like. In that way, I can find people that probably won't piss me off too much and a way to find other stuff that I think is cool.

So, for example, I would exclude anyone who didn't like The Big Lebowski on the principle of them being nilhists.

Seems like movie ratings would be a great way to find a perfect match. Do dating sites use this data?

Yes, anytime they ask you to list your favorite movies/books/music in your profile. Of course the matching algorithms are only in the heads of your prospective dates.

flickchart.com must have a lot of data on that. It would be interesting to see some social network or dating service arise out of that. Lots of potential, and more practical than time-filler sites like hot-or-not (which flickchart kind of is right now).

Lists of the most loved, most hated, & most contentious movies -

http://www.netflixprize.com/community/viewtopic.php?id=164

Most hated movie = "The Stepford Wives"? surprising. I really liked it. Would be interesting to see if it's a man/woman divide at work. imdb rating is 5.1/10 :/ I can think of far worse movies. Try "The Talented Mr. Ripley".

"Lost in translation" is like a good parfume or wine: you need some time to pass to get the best of it, be it true scent or fine aftertaste.

I'm not sure I agree with that. If they were less accessible to first time viewers, then they would likely have a low rating overall. If thats right, then any decent prediction engine would generally give low ratings and not rack up a huge error.

Perhaps these hard-to-predict movies simply have pointy bits sticking out. A rose is great until you get poked, then you irrationally throw it away. Napoleon Dynamite has lots of parts that make people uncomfortable - different parts for different people. Never mind its a great movie - it will get irrationally rated by a great many folks. So come up with a Beta for a movie, or cluster the folks that like it anyway, or try to detect what kind of phobias a person has - these movies could be great litmus tests for homophobia or whatever!

Thats an interesting idea. I'd still be surprised though if that didn't correlate with some movies. If someone hates Napoleon Dynamite due to feature X, then they should dislike movies that have X as well. Or, at the least, avoid watching movies with X.

So, if your theory is correct then I see two possibilities:

1) The pointy bits are themselves strange and don't appear i other movies. This goes back to "the movie is strange" again though...

2) People who dislike these traits avoid movies that have them and the algorithms don't take into account what movies people haven't seen (or that taking it into account is very tough since most people haven't seen most movies).

Not sure which is more likely.

One thought I haven't seen mentioned is the emotional aspect of several of these movies. In my opinion, Lost in Translation was bad, but I made it through it, I heart huckabees & Sideways were so terrible I stopped watching in the middle. All three movies were good in the sense that I could get drawn into the stories, but the characters/storyline just made me want to scream "Get over it you freaking cry-baby".

They weren't bad in the way that a normal bad movie is bad. They were bad in the way that it took talented writing, acting & directing to make characters and situations that were so pathetic and annoying that it was unwatchable to a guy like me.

I can easily see how someone else (who presumably could relate to the emotions of the characters) would really like the films.

They still didn't answer the question as to WHY is it so difficult to predict? I think that Napoleon Dynamite has two types of fans: The people that watch it for the purely (seemingly) primitive comedy and the people who find a deeper meaning in it and who 'get it'. The first type will also like stupid screwball Adam Sandler comedies, while the second type will go for more indie, artistic fare. Thus the ambiguity when trying to predict per user.

Reply on news.ycombinator.com