What people played in our browser arcade last quarter

If you are deciding whether to put a browser build in front of players, you are mostly deciding without evidence. Play numbers in this corner of the industry are private, or they are stated in a way nobody outside the company can check.

So we published ours. The Q3 2026 arcade play report is the first one, it carries the method it was built with, and the numbers behind every bar are downloadable as a CSV. This post is what we think it says.

The direct answer

Between 1 July and 30 September 2026, in Bounty Board Arcade:

1. 1,229 plays from 433 players, across 87 hours of play time.
2. 46 games were played, out of 72 in the arcade when the report was generated.
3. Puzzle games took 42 percent of play time: 537 plays, 253 players, 11 games.
4. Action games took 36 percent: 364 plays, 160 players, 12 games.
5. Everything else is one row at 22 percent: five genres that did not clear the row floor, plus the games outside our genre list, together 328 plays and 23 games.

The average play ran about 4.3 minutes.

What a play and a player mean here

A number is only worth citing if you can see how it was counted, so the report states its own rules and this is the short version.

A play is one session started in our own player. Preview plays by a studio testing its own game are excluded, the same as in every other arcade count. A player is counted once per account, and once per browser for a guest, which means a signed out player on a phone and the same person on a laptop are two. Play time is measured while the game is open and added in steps of at most 45 seconds, so it is time present rather than time claimed. Quarters are UTC and a quarter is only published after it ends.

There are floors underneath all of it. A quarter is published at 500 plays and 100 players, and gets no page at all below that. A genre earns its own row at 150 plays, 40 players and 3 games, and otherwise folds into the everything else row, which is why the shares still add to 100.

Those floors are stored with the quarter rather than read fresh, so if we raise them next year, this page keeps claiming what it claimed on the day you cited it.

Puzzle games reached more people, action games held them longer

The two numbers worth sitting with are not the shares. They are reach and hold, and they point in different directions.

Puzzle reached the most people per game: 23 players per puzzle game against 9 across the whole arcade. Eleven puzzle games pulled 253 of the quarter's 433 players. If your goal is the largest number of distinct people trying the thing, that is the shape of it.

Action held a player longer once they started. About 5.1 minutes per play against 4.1 for puzzle. That gap is real, and it is also small: it is a 1.24 times difference, which sits just under the 1.25 times bar the report uses before it will call something a finding. So the report does not claim it, and neither will we. We are telling you because the restraint is the point of publishing a method at all.

The one thing the quarter does claim is the reach gap, because 23 against 9 clears the bar comfortably.

What this quarter cannot tell you

One arcade, one quarter, 72 games. That is a small catalog and an early one, and the report says so on its face.

It also cannot tell you why. We know puzzle games reached more people per game; we do not know whether that is the genre, the particular eleven games, where they were linked from, or the order they happened to sit in on the wall. A quarter of play data is a measurement, not an explanation, and anyone who hands you a causal story from numbers this size is selling something.

What it can do is give you a public, dated figure to argue with. If your own browser game holds a player past five minutes, you are above the arcade's average play, and you now know by how much.

If you are shipping a browser game

Three things we would take from this quarter.

A puzzle game is the cheaper way to reach people, not to keep them. Reach per game was more than double the arcade average, and the time each play held was the lower of the two rows. Those are different goals and they want different builds.

Five minutes is a real bar. The average play across every genre ran 4.3 minutes. If your opening takes two of those to explain itself, you spent half your evidence before the game started.

Most of the catalog is not the story. 46 of 72 games were played at all in the quarter. Being listed is not distribution, in our arcade or anyone's, and the games that got played had something pulling people to them.

You can put a game in the arcade while any other submission is still in review, and the Arcade SDK gives you leaderboards, cloud saves and the play tracking these numbers come from. Eligible Arcade studios keep 90% of finalized ad revenue attributed to their game, settled monthly via Stripe once the $50 minimum is reached. Terms apply.

The next report covers Q4 2026 and goes up after the quarter ends. Every published quarter stays at its own address with its numbers unchanged, so a link to one keeps saying what it said when you made it.