Here at Goodfilms, we’re all about doing interesting things with film ratings data. We’ve developed a new way of rating movies that lets you express your opinions in a much more meaningful way, and we’d like to introduce how we’re using it.
We’re calling this project InGen after the corporation from Jurassic Park (one of our favourite films here at Goodfilm HQ). We admire the innovation and ambition of InGen from the film, and think that as long as we don’t breed velociraptors, we’ll be ok.
Ranking your friends
The first project to come out of InGen is about putting you and your friends’ tastes head-to-head, to see how similar you are. And, by extension, figuring out who is the most and least similar to you, so you know who to go to the cinemas with, and who to make fun of.
Because all of the ratings on Goodfilms are based on two separate criteria (quality and rewatchability), we can plot them on a graph (see goodfil.ms/graph).
We realised that placing the films on a graph makes it really clear where two films sit in relation to each other, so we decided to classify every pair of films in one of 9 ways, which we call a Vector Victor (again, after one of our favourite films ). For example, Goodfilms users think the film Anchorman is more rewatchable but of lower quality than something like Black Swan. These are the 9 possible Vector Victors (hover for an example):
|More rewatchable but lesser quality.||Same quality but more rewatchable.||Better in both respects.|
|Same rewatchability, lower quality.||Same in both dimensions.||Higher quality but same rewatchability.|
|Worse in both dimensions.||Same quality but less rewatchable.||Higher quality but less rewatchable.|
This gives us quite a meaningful comparison of two films, but when we apply it to an individual’s ratings, it gets even more interesting.
Judging your friends
We realised that we have a nice way of comparing you and your friends:
For a pair of films, do you agree on their Vector Victor?
If you think that Black Swan is not higher quality than Anchorman, you and I don’t agree. And if you would enjoy rewatching a film about a ballet dancer’s descent into psychosis more than a light-hearted Will Ferrell movie, we’re not going to have a fun movie night.
What’s neat about this approach is that it the exact ratings don’t have to match for us to agree. Which works quite nicely because it doesn’t matter how harshly you or your friend rate films, as long as you agree on their relativity.
Also, since both quality and rewatchability are being considered, you are making quite a precise statement about how two films compare. The nine possible Vector Victors all represent something fairly specific, so the fact that you and a friend agree is quite meaningful.
This lists your friends in order of how close you align. The numbers on the right represent how many films you’ve both rated, and, more interestingly, how many films are in both your queues. Click on this and you have a great way to pick a film to watch with someone:
We hope you all find out something interesting about your friends with this. If you’re already a user, hop on over to goodfil.ms/people/friends and check it out. If not, head straight to goodfil.ms and sign up!
- This is a work in progress. Things will change as we refine it.
- This is not the only signal we are using to compare you and your friends, but it’s the most significant and the most original.
- Two ratings (in either dimension) don’t need to be identical for us to consider them ‘the same’. We use a continuous rating system, so we account for a little noise in people’s ratings by using a certain tolerance in our calculations. At the moment, that tolerance is roughly a half-star or half-smiley wide.
- We’d like to thank FlickChart for showing how informative it can be to directly compare two films.
- Evan Miller’s blog post and the resulting Hacker News discussion also deserve thanks for helping to explain and recommend the Wilson interval for ranking algorithms like ours.
- We love feedback. Comment here, post something on our feedback forum, or hit us on Twitter or Facebook.