The hour game servers fill up, game by game

Game servers on JoinGG fill up at 17:00 or 18:00 UTC. In a week of hourly readings, eight of the nine games with enough data had their busiest hour at one of those two times, and Half-Life followed at 19:00. The quiet hour varies more, from 03:00 UTC for Counter-Strike 2 to 08:00 for Half-Life, and so does the distance between the two.

The readings come from JoinGG's own poller. It queries every listed server every 15 minutes over the game's own protocol (Valve's A2S_INFO for the Source and GoldSrc games, the server list ping for Minecraft) and stores one reading per server per hour. The window is 21 September 2026 10:00 UTC to 28 September 10:00 UTC. Only servers with a reading in every hour of the day were used, so a gap in one hour cannot pass for a dip. Minecraft listings with no server behind them were dropped, and so were the 11 Minecraft networks averaging over 1,000 players, so the curve is a typical server's rather than one network's. Six hours in which the poller itself recorded far fewer readings across every game at once were left out. Player counts from Valve's games include bots, because A2S reports bots inside the total. The methodology page has the rest.

18:00 UTC is a European evening

Converted, 17:00 to 18:00 UTC is 19:00 to 20:00 in Central European Summer Time, 20:00 to 21:00 in Moscow, 14:00 to 15:00 in Brasília and 13:00 to 14:00 on the US East Coast. That is the start of the evening across Europe and early afternoon in New York. A peak placed there says as much about who plays on these servers as about the games: it is what you would expect if a large part of the population lives in or near European time zones. Why a server feels empty at 3am looks at the same effect from one server's side.

Central European Summer Time ends on 25 October 2026. If the peak is set by European evenings, it should arrive an hour later in UTC after that, at 18:00 or 19:00. That is a prediction, and this week's data cannot show it.

How far each game swings

The last column divides the players in the busiest hour by the players in the quietest. Counter-Strike 2 has almost five times as many at 18:00 as at 03:00, the shape of a game people sit down to play in the evening and leave at night. Garry's Mod (2.91) and Left 4 Dead 2 (2.36) swing hard too, with legacy CS:GO not far behind at 2.16. Minecraft (1.80) and Team Fortress 2 (1.67) sit in the middle.

The older Counter-Strikes and Half-Life are much flatter. Counter-Strike 1.6's busiest hour is only 1.49 times its quietest, and it has no weekend lift at all: its daily means stayed between 9,970 and 10,072 players on all six full days of the week. Because A2S counts bots as players, a population that barely moves by hour or by day is what bots and always-on servers look like. The data is consistent with that. It does not prove it, and it cannot say which servers are responsible. What a player count actually tells you goes through the ways a number on a list can mislead, and reading a player graph honestly applies the same caution to a single server.

Weekends

The weekend figures use every server, big networks included, and compare the average of Saturday's and Sunday's daily means with the Tuesday-to-Friday average. Minecraft gains the most, 53%, and its Saturday alone ran about 72% above the weekday average, with Sunday about 34%. Counter-Strike 2 gains 29%, Garry's Mod 27%, Left 4 Dead 2 22%, Team Fortress 2 19%, legacy CS:GO 18%, Half-Life and FiveM 10% each, Counter-Strike: Source 4%, and Counter-Strike 1.6 nothing.

Minecraft's Saturday is the biggest single-day lift in these numbers, and the obvious day for anything on a Minecraft server that needs a crowd. Planning an event people come back for covers what to run once you have picked the slot.

CS2's bad Wednesday night

On Wednesday 23 September, between 00:00 and 07:00 UTC, the number of CS2 servers answering at all fell from a typical 757 or so to as low as 396. Wednesday finished as CS2's quietest day of the week, with a daily mean of 1,836 players against 2,286 to 3,081 on the other days. The cause is not in the data. Those hours include CS2's quietest hour, so part of its swing may come from that one night, and a single week cannot say how much. The live CS2 figures will show whether it happens again.

Why FiveM is not in the table

FiveM is left out of the hourly table on purpose. Only 15 of 1,854 FiveM servers had a reading in every hour of the day that week, so an hour-by-hour comparison would be built on gaps. The weekend comparison works from daily means, which is why FiveM appears there and not above.

When to look for a full server

For a player the rule is short: convert 17:00 to 18:00 UTC into local time and look then, and on Minecraft look on a Saturday. In CS2, Garry's Mod and Left 4 Dead 2 a server that looks empty at 04:00 UTC can be busy at 18:00, so an empty browser at the wrong hour proves little. Individual servers vary around these averages, and one server's own peak hours are what to check before giving up on it.

Where to put the restart

For an owner the useful column is the quiet one. Restarts and updates hurt least when the fewest people are on, and across these servers that hour was 03:00 UTC for CS2, 04:00 for CS 1.6, 05:00 for CS:S and Garry's Mod, 06:00 for Minecraft and legacy CS:GO, 07:00 for Team Fortress 2 and Left 4 Dead 2, and 08:00 for Half-Life.

Those are averages over hundreds of servers. A community based in Brazil or on the US West Coast will probably have its trough somewhere else, so check your own graph before moving a scheduled job. Keep big updates out of the two hours before the peak, when regulars are arriving and would find the server down. If the job runs on local time in Europe, it moves by an hour against UTC when the clocks go back on 25 October. Whether a scheduled restart fixes anything is a separate argument; if you run one, the quiet hour is the place for it.