Audience note

Tell us what went wrong

Bugs, confusing picks, missing filters—plain language is perfect.

Anonymous · goes to our inbox, never shown on the site

How We Pick

Every time you press the button, the site runs the same few steps: decide which titles are allowed in, count them, and pick five at random. This page explains each step, including the parts we got wrong and the trade-offs we chose on purpose.

Where the library comes from

Films start from Wikidata, the open database behind Wikipedia. It lists a little over 200,000 films with a title, a year and genres, and anyone can check or correct an entry. We use Wikidata as the permanent backbone because it is free to reuse and does not disappear if a commercial service changes its terms.

Posters, synopses, runtimes, audience ratings and “where to watch” come from TMDB. The streaming information on TMDB is supplied by JustWatch, which is why the “Stream on” links open a JustWatch search. Those links are not affiliate links, and we earn nothing when you use them.

TV shows come from TMDB alone. Wikidata’s coverage of series is too patchy to build on, so the TV library is the 17,627 shows that pass the rules further down this page.

Why 13,000 films and not 200,000

A truly random film from 200,000 is almost always something nobody has heard of, can’t find anywhere and has no poster. That is not useful on a Tuesday night.

So we rank every film by TMDB’s popularity score and keep the top 15,000 in the default pool. After removing films that have not been released yet and films with no genre at all, that leaves about 13,500. It covers everything from the obvious classics to plenty of films most people have not seen, while still giving each draw a decent chance of being watchable tonight.

A few pages widen the pool to the top 40,000, because they already have a strong filter of their own:

  • Streaming pages (Netflix, Disney+): being on the service is already a high bar, and the stricter cut would hide films that are clearly there.
  • The Christmas page: only a few hundred films carry the Christmas tag, and the stricter cut left the pool too thin.
  • The “Released” filter: new films have not had time to build up popularity, so the stricter cut would leave the last month nearly empty.

The ranking is not perfect. Our popularity cut misfiles some well-known films. Home Alone sits just outside the pool, for example, and the Christmas page says so openly instead of pretending otherwise.

How the draw works

When you press the button, the server counts how many titles match your filters, picks five different random positions in that list, and returns those five. Every matching title has exactly the same chance of being chosen. Nothing is promoted, sponsored or weighted toward films we like.

Within one visit, the picker remembers the last 50 titles it has shown you (ten rounds) and will not deal them again. Reload the page and that memory is gone. We do not keep a viewing history, and there is no account.

We deal five at a time because one pick feels like a verdict you want to argue with, and forty is just scrolling again. Five is a short list you can actually read and choose from in a minute.

How ratings are treated

Audience ratings come from TMDB. A score means little if only three people voted, so each rating filter also needs a minimum number of votes: 6+ needs at least 50, 7+ at least 100 and 8+ at least 200.

Without that floor, the 8+ filter fills up with obscure films that two enthusiastic fans rated 10.

TV shows work differently. Many good shows have few votes on TMDB, so instead of a score filter the TV picker has a “How well known” filter (10+ or 100+ votes). A show with no votes at all says No rating yet on its card instead of showing a misleading zero.

What we take out

Before anything reaches the draw:

  • Unreleased titles. Films dated later than the current year are excluded, and so are TV shows that have not aired.
  • Titles with no genre. About one in ten of the most popular films has none on Wikidata, and the card would be missing half its information.
  • Adult content. TMDB’s own adult and “video” flags are excluded at source. For TV, where those flags miss more, we add three more checks: the adult flag on each show’s details, keyword tags such as “softcore” and “hentai”, and words in the title like “erotica” or “uncensored”. A short hand-kept blocklist catches what still gets through.
  • News, talk and reality TV when that is all a show is. They are rarely what someone means by “pick me a show”.
  • TV shows with no poster. For TV we also require at least one genre.

How often the data changes

Every Monday an automatic job adds newly released films and shows, and refetches the oldest TMDB details: about 8,000 films and up to 3,000 shows a week, oldest first. Streaming availability changes faster than that, so treat the “where to watch” row as a starting point and check JustWatch for your country before you settle in.

The picks written by people

Several pages include short lists such as “Twelve to start with” or a viewing order for a director. Those are written and checked by a person, not generated, and each one says how the films were chosen. They sit next to the random draw and do not change it: a film being on a list does not make it more likely to come up.

Live activity on the page

The floating messages on some pages show real activity: which titles people opened, reactions like 😱 or 🍿, and filter combinations other visitors used. We do not add any of our own, and the site does not accept written messages from visitors. Feedback sent through the Feedback button is different: it goes privately to our inbox and is never shown on the site.

What we don’t do

  • We do not personalise picks or build a profile of you. See the privacy policy for exactly what is stored.
  • We do not take payment to include or promote a film.
  • We do not host or stream any film or show.

Spotted a film in the wrong place, a missing poster or a show that should not be there? Tell us. That is how most of the fixes on this page happened.