Letterboxd Upsets
Learning about the Letterboxd user base from what films they prefer over the general public, vs what films they dislike
Methodology
I scraped 60,000+ movies and their review data from IMDb and Letterboxd -- all movies that had at least 500 user ratings on IMDb. I used a Bayesian star rating to calculate how Letterboxd users rated a movie out of 10, and compared this to the weighted average for user ratings given by IMDb's dataset. Ratings were calculated within a 95% confidence value.
For each film, I compared the difference in its ranking, sorted by how many places a film jumped from both sources. The percentage is how far up/down the list a film moved between sources.