The bots in the player count

The daily average of players JoinGG counted across about 1,500 Counter-Strike 1.6 servers stayed between 9,970 and 10,072 on all six full days of the week to 28 September 2026, weekend included. Human populations rarely hold that still. It fits a count that is partly software: Valve's query protocol reports bots as players. Some of those bots are part of the game, and some are there to be counted.

A week with no weekend

The rest of that week's readings, taken from 21 to 28 September, point the same way. On those Counter-Strike 1.6 servers the busiest hour of the day was only 1.49 times as busy as the quietest, and Saturday and Sunday were no busier than the weekdays. Counter-Strike 2, measured the same way over the same week, looked like a game people play after work and at weekends: its busiest hour was 4.8 times its quietest, and its weekend was 29% busier than its weekdays.

That comparison proves nothing about any single server, and there are human explanations for part of it. A player base spread across many time zones flattens the daily curve, and a core of regulars who play every evening whatever the day flattens the week. Neither explains an entirely missing weekend very well. What would produce that shape is a large number of bots and always-occupied servers: a floor that never moves, with a thinner layer of people on top. JoinGG's figures include bots because the protocol does, as the methodology page says.

What the protocol counts

A2S, the query that Valve's games answer, returns a player total and, separately, a count of bots. The Steamworks header that defines how Steam's server browser stores that total describes the player field as the total number of players on the server and adds, in capitals, that it includes bots. Valve's own server API calls a bot a fake user, one that shows up in the list of people playing. Lists then differ on whether to subtract the bot field, which is one of the reasons two trackers can disagree about the same server. JoinGG does not subtract it.

Subtraction only works if the server declares its bots honestly, and a server can simply not. When Valve ran Quickplay for Team Fortress 2, a community server that wanted to be included had to report its player count and its bots accurately, and could not dress bots up as people, for instance by showing a ping or a custom avatar for them on the scoreboard. A patch on 29 October 2013 took a number of servers that had been forging their player counts out of Quickplay and the server browser. What a forged count looks like to a poller is covered in how server lists see your server.

Bots that are the game

Some bots are the point of the mode. In Left 4 Dead 2 any survivor without a human player is controlled by the computer, so a campaign server with one person on it is running three bots by design. Team Fortress 2's Mann vs. Machine sets up to six players against waves of AI-controlled robots, and the dedicated server refuses to start one of its maps with fewer than 32 player slots, which leaves room for the robots. Counter-Strike servers fill teams with bots in quiet hours and during warm-up, and the better setups drop a bot each time a person joins. Counter-Strike 1.6 servers can run third-party bots such as the open-source YaPB.

Any list that reports raw counts shows all of these, and it should, since a Left 4 Dead 2 server with one player and three bots is a playable game.

Bots that are for show

Then there are bots whose only job is to be counted. Lists sort by population, and an empty server stays empty because nobody wants to be the first one in. A handful of bots that never leave, often with ordinary-looking names, lifts a server up any list sorted by players, and some of the people who join because of them stay long enough to make the number partly true. That is why the practice survives: it works a little, and the cost falls on someone else.

The someone else is the server that does not do it. Its eight real players sit below somebody's twenty decorative ones on every sorted list, and the newcomers who might have become its regulars never scroll that far down. Lists have a dilemma of their own. Subtract every bot and a Left 4 Dead 2 server with a full team looks emptier than it plays; subtract none and a decorated server outranks a real one. Showing both numbers is the honest answer, and it still leaves the reader to judge.

Reading a count that might be decorative

A player count is a reading of whatever the server chose to report, so the useful checks are about shape more than size: a line that does not move across a whole week, a server at its maximum at four in the morning, a player list whose names never score. A server that shows as full at every hour of the day is telling you something, though not what it hopes. For owners who run bots for good reasons, the fix is simple. Declare them as bots, and let the bot field say what they are.