What our co-op finder could not tell you

What was there, and what replaced it

Until this week our co-op page offered a filter: couch co-op, split screen, online, minimum player count. It listed 38 games and it worked exactly as described.

It is gone. The page now shows the award record for co-op games and sends anyone who wants to find one to the Co-op Compendium on Next Game Finder — a catalogue built for that question, with verified co-op data for an order of magnitude more games.

That site is run by the same person as this one, which you should know before clicking. It is linked because it answers the question properly, not because it is ours.

The list was not the problem

38 games is a short list, but a short list is honest if it says so. What made the filter worth removing is quieter.

This database holds 603 games. Co-op status has been checked by hand for 42 of them. The other 561 carry whatever Wikidata knows, and Wikidata records only whether a game is single-player or multiplayer. It has no field for couch play, none for split screen, none for player counts.

So the filter was reading three states and showing two. A game that nobody had examined came out the same as a game examined and found to have no co-op.

15 of the 23 census-season winners have never been checked for co-op. Their blank in that column is not a finding.

The number that looked like a trend

Here is what that does to a statistic. Take the share of each season's awards that went to a game with co-op, and read it as a trend:

Season by season, that share swings from zero to above seventy per cent and back. It looks like co-op surging and collapsing. It is nothing of the kind. In 2022 it reads high because Elden Ring won the season with 438 awards and happens to be one of the 42 games somebody checked. In 2025 it reads zero because only one game from that season was ever examined, and that one has no co-op.

The line does not track co-op. It tracks where the hand-checking happened. Any chart drawn from it would have been a picture of our own effort wearing the costume of a finding.

What is actually true, stated as a floor

The award record survives all of this, because awards are what this site actually counts. Games known to have co-op hold 2,398 awards — 26.7% of everything in the database, led by Elden Ring on 438.

At least 6 of the 23 census seasons were won outright by a game with co-op. At least, not exactly: 15 winners were never checked, and some of them certainly qualify.

Every one of those figures is a floor. Written as a floor they are useful; written as a count they would be wrong, and a filter that let you sort by them implied a precision that was never there.

The general version

A feature that answers a question badly is worse than no feature, because it does not look like an absence. It looks like an answer. Someone filtering for split-screen games on this site got a confident, sorted, fast-loading table that quietly omitted every game nobody had got around to examining.

The same test applies to everything else here, which is why totals are published as documented minimums, why seasons below census breadth are kept out of comparisons, and why the publisher win counts are split by era rather than added up. When a number cannot carry the weight of the question, the honest move is to say so and point at whoever can.

For co-op, that is Next Game Finder. For which games the press actually rewarded, this is still the place.