Coilr report·Player activity·
Steam has one rush hour, and it is Sunday afternoon
Coilr records how many people are inside a game every few minutes. Over the last 28 days that is 1,158,405 readings across 504 games — enough to ask a question the individual charts cannot answer: does Steam as a whole have a rhythm, and is it the same rhythm for everyone?
It does, and very nearly. The average game carries 2.10 times as many players in the busiest hour of the week as in the quietest, the peak lands on Sunday 15:00 UTC, and 128 of 504 games reach their own high point in that same hour of the day. The exceptions are the interesting part, and they are exceptions of geography rather than of taste.
2.10x
between the busiest and quietest hour of the week
15:00
UTC on Sunday — the week's peak
24%
more activity on Sunday than on Wednesday
The shape of the week
Every game is normalised against its own weekly average before anything is added up, so this is the shape the average game shares rather than the shape of the biggest one. A darker cell is a busier hour.
The peak, Sunday 15:00 UTC, is 08:00 Sun in Los Angeles, 11:00 Sun in New York, 16:00 Sun in London, 17:00 Sun in Berlin, 00:00 Mon in Tokyo — a weekend afternoon across Europe and a weekend morning across America, the one stretch of the week when both are awake and neither is at work. The darkest column, 06:00 UTC, is the mirror of that: Europe has not started, America has stopped, and Asia, awake and at its desk, does not make up the difference.
The week changes the volume, not the schedule
Read the chart down a column rather than across a row and the real surprise appears. Every one of the seven days peaks in the same hour — 15:00 UTC — and every one bottoms out at 06:00 UTC. A Wednesday is not a differently shaped day from a Sunday. It is the same day, quieter. The weekend does not move when people play. It moves how many of them do, and how long the busy stretch lasts.
Hours of each day in which the average game is above its own weekly average. The busy stretch on a Sunday lasts 15 hours; on a Wednesday, 9.
Weekends are not one thing
The weekend lifts everything, but not evenly, and the spread says more about a game than its genre does. The games at the top are the ones you sit down for; the games at the bottom are habits, dailies and tools — things people open on a Tuesday because it is Tuesday.
Most weekend-shaped
Flattest across the week
The games on another clock
A handful of games peak between midnight and 10:00 UTC — the middle of the European night. They are not being played by insomniacs. They are being played on the other side of the world, and their peak is somebody else’s evening.
Method
Coilr polls Steam’s official player-count API on a fixed schedule and stores every reading. This report uses the 28 days ending August 22, 2026: 1,158,405 readings across the 504 games sampled densely enough to have a real daily curve. Games polled only occasionally are excluded entirely rather than averaged in, because their quietest hour would only be an hour nobody happened to measure. The per-game figures cover all 504 of them; the tables name only the 334 averaging at least 500 concurrent players, because a ratio built on a few dozen is noise, and noise is what rises to the top of a list sorted by extremes.
For each game, the readings are averaged into the 168 hours of the week and then divided by that game’s own weekly average. Averaging that across games is what makes the figure a shape rather than a headcount — without it, the single largest game would decide every bucket on the chart. An index of 1.00 means an ordinary hour for that game; 1.47 means 47% above its own norm.
As a check on whether the pattern is real or an artefact of one busy month, the window was split into two halves and the 168-hour curve computed independently in each. The two agree at r = 0.96, with the peak landing in the same hour and the swing within 0.1x. Times are UTC throughout, which is also how they are stored; nothing here is adjusted for daylight saving, so the boundaries of the bright band drift by an hour across the year.
Player counts are concurrent players, not unique players, and Steam’s API reports them per game rather than per account — so a person in two games at once is two readings. No individual account is involved in any figure on this page.
Figures free to quote with attribution to Coilr. The underlying per-game curves are visible on every game page, and if you would rather use this than read it, the teammates board matches people on exactly these hours.