What is deliberately wrong with movie-ratings

What is deliberately wrong with movie-ratings.

9,000 rows, 7 columns, one row per user-film rating. Nine thousand user ratings over 260 films, with genres, release years, and the long-tail popularity distribution that makes naive recommenders look better than they are.

Ratings are missing-not-at-random: people mostly rate films they chose to watch, which is why an average rating is a popularity measure as much as a quality measure. Twelve films have fewer than five ratings — those are where a "top rated" leaderboard goes wrong without a shrinkage prior.

Something to try: Compare item-item cosine similarity against a plain popularity baseline. The SQL playground has it loaded already, and the data explorer will profile every column in one pass if you would rather look before you query.

select *
from movie_ratings
limit 20;

The file is generated from a fixed seed, so the CSV you download today is byte-identical to the one from last year. That matters more than it sounds: it means a lesson can say "row 412 is the interesting one" and still be right in eighteen months.

Download it, break it, keep it. CC0 — no attribution required, no account, no email.