Jev's taste in films vs MovieLens users

Does Jev like the films that MovieLens users like, and where does it disagree most?

result3,935 questions

Jev's ranking of 3,935 films agrees closely with MovieLens users' (rank correlation 0.76). It likes cult favorites like A Nightmare on Elm Street 3: Dream Warriors and Death Becomes Her far more than they do, and the true-crime series Death on the Staircase and the stand-up film Bill Cosby, Himself far less.

How to read this: Each dot is a film. Across: its rank among MovieLens users (left is lowest). Up: its rank for Jev. Dots on the diagonal are films both rank the same; the labeled ones are the biggest disagreements.

3,935 items; 90% interval 0.75 to 0.78.

In short

Films are where Jev's taste comes closest to real audiences. Its ranking of 3,935 films lines up closely with MovieLens users' (rank correlation 0.76).
Where it departs, it leans toward crowd-pleasing genre films like Sister Act and Hocus Pocus, and away from stand-up, documentaries and slow international classics that their own audiences rate highly.

What the data shows

Sparta Leonidas meme: Jev on A Nightmare on Elm Street 3: Dream Warriors: THIS. IS. UNDERRATED. — Jev on A Nightmare on Elm Street 3: Dream Warriors: THIS. IS. UNDERRATED.

Close agreement overall: rank correlation 0.76 (90% interval 0.75 to 0.78). Films people love, Jev mostly loves; the flops, it mostly dislikes.
Jev likes some cult and comfort films much more than MovieLens does: A Nightmare on Elm Street 3: Dream Warriors, Death Becomes Her, Sister Act, Supercop 2 and The Grandmaster.
It likes some acclaimed or niche titles much less: the true-crime series Death on the Staircase (Jev 1.24 on the 0-to-4 scale, MovieLens users 2.99), Bill Cosby, Himself (1.08 vs 2.89), The Postman (1994), Paint It Yellow and Song of the Little Road.

What it means, and what it doesn't

Jev largely knows which films people enjoy, and its departures run one way: toward comfort and genre films, away from the titles a small, devoted audience prizes. Recommendations from Jev would lean mainstream.

It doesn't mean Jev has watched any of these. A ranking that tracks the audience this well could come from knowing each film's reputation. The other catalogs (books, board games) agree much less, which is part of what makes films interesting.

Caveats

Fans rate what they chose to watch. MovieLens users rate films they picked, and people pick films they expect to like, so a niche documentary gets rated mostly by its fans. Jev rates every film cold. That alone pushes some niche titles higher for the audience.
Stars squeezed into five levels. MovieLens ratings are half-stars from 0.5 to 5; they were binned onto the project's five described levels to show them next to Jev's. The bins are the project's choice, so the comparison uses ranks, not levels.
A famous-film list. Every film had at least 300 ratings, and Jev has read about most of them, including their reputations. Its rating may track what critics wrote as much as anything else.
Reputations change. MovieLens ratings span 1995 to 2023, but what's written about a film or its star keeps changing. One possible reason Jev rates Bill Cosby's stand-up film so low is what has been written about Cosby since; that can't be tested here.

Jev on this experiment

Would a person find it interesting to read?: Yes80%
Does it describe you?: No64%
Would you have predicted it?: No61%
How fair is the comparison?: The comparison is shaky
How much should a reader rely on it?: Moderately
Which caveat matters most?: Fans rate what they chose to watch72%

Why ask this

Ask a chatbot what to watch tonight and the answer carries its taste, whether or not it says so. If that taste drifts from what audiences actually enjoy, every recommendation drifts with it.

Most of what Jev says about taste can only be checked against itself. Films are different: MovieLens, a recommendation site run by the GroupLens research lab, has millions of real ratings. So here Jev's ranking of 3,935 films can be set against what the people who watched them thought.

How this was done

The people and the data

MovieLens users: people who signed up to a free film-recommendation site and rated films they'd seen, in half-stars from 0.5 to 5, between 1995 and 2023. The films are the 3,935 in the MovieLens 32M release with at least 300 ratings each, so every one has a real crowd behind it, from a few hundred raters for niche titles to tens of thousands for hits like Dumb & Dumber. For each film, the experiment uses the distribution of its ratings, binned into the same five levels Jev answers on.

What Jev was asked

Every film one at a time:

How much would you enjoy watching Hud (1963)?
You'd turn it off within the first twenty minutes · You'd finish it but forget it within a week · You'd enjoy it once and not seek it out again · You'd recommend it to a friend · You'd rewatch it and count it among your favorites

Each was also asked with the answers in reverse order, and the two averaged. Jev never saw the MovieLens ratings.

How it was measured

The analysis ranks all 3,935 films by Jev's average level and by the audience's, and compares the two orders with a rank correlation (1 means the same order, 0 means no relation). Ranks rather than levels, because Jev's described levels and people's stars aren't the same scale. The biggest disagreements are the films whose two ranks are furthest apart.

Where these questions live

3,935 questions across 1 topic of the map. Each opens on the map with every question in it.

Ratings

Film Ratings3,935

Every question

All 3,935 questions behind this result, the telling ones first: the examples the analysis points to, then the ones where Jev misses, biggest gap first.

Jev’s own answer

Showing 0 of 3,935