JoinGG

Methodology

How we measure

Every number on this site comes from asking the server itself, over the protocol its own players connect with. Nothing here is estimated, scraped from another directory, or supplied by the server’s owner.

The short version

A poller queries every one of the 11,567 approved listings every 15 minutes, speaking each game’s wire protocol directly. It keeps one reading per server per hour for 7 days. Uptime is the share of those readings in which the server answered. That is the whole method.

There are 1,645,168 readings in the database as this page renders. Where a figure cannot be supported by enough of them, the site shows nothing rather than a guess — which is why some listings have an uptime percentage and others do not.

What we ask, and how

Each game has a query protocol that its own client uses to populate a server browser. We speak those directly, with no third-party API in between and no dependency doing it for us:

  • Minecraft — Server List Ping, preceded by the same SRV lookup a real client performs, because the address a player types is often not the host the server runs on.
  • Source engine (Counter-Strike 2, Garry’s Mod, Rust and the rest) — A2S_INFO over UDP, including Valve’s challenge–response step.
  • FiveM — the HTTP info and players endpoints the server publishes itself.

A server that does not answer is recorded as offline for that reading. It is not removed, and one silent cycle is not treated as a server being gone.

Why one reading an hour, not one every 15 minutes

The status on a listing page — players online now, current map — is refreshed on every 15-minute cycle. The history is written once an hour, because the activity chart buckets by hour and nothing on the site renders a finer resolution. Storing four readings an hour would mean writing three rows that get averaged away before anybody sees them.

Readings older than 7 days are deleted on each cycle. This site does not keep a permanent record of what any server was doing.

How uptime is calculated

Uptime is the number of readings in which the server answered, divided by the total number of readings held for it, to one decimal place. It is not weighted, smoothed, or adjusted.

It is shown only when there are at least 24 readings spanning at least 3 days. Below that the figure is withheld entirely rather than shown with a caveat: a percentage computed from four readings taken over one afternoon looks exactly like one computed from a week, and there is no honest way to print the first.

The consequence is worth stating plainly, because it looks like a bug and is not: a newly listed server has no uptime figure for its first few days.

Peak hours

The busiest hour is the hour of the day with the highest average player count across the readings we hold. It is stored in UTC and converted to your own clock in your browser — the UTC reading stays available in the tooltip.

A peak hour is an average over days, so for a reader in a time zone that observes daylight saving it can be an hour out for part of the year. For “when is this server busiest” that is well inside the noise; for anything needing precision, read the UTC value.

What the numbers cannot tell you

Some of what a server reports is free text that its operator sets, and we publish it as what it is rather than dressing it up:

  • Version is a string the operator chooses. Plenty of servers put something else there — a status message, a plugin name, an error. We show it only when it actually contains a version number, and show nothing otherwise.
  • Map is whatever the server was running at the moment we asked. On a rotating server it will be different an hour later, and a map hub is a snapshot rather than a roster.
  • Player counts are what the server reported. On the Source protocol, bots are included in that figure by the server itself, and we do not attempt to subtract them.
  • Country is derived from the server’s address, so it locates the machine, not the community using it.

What we do not do

  • We do not interpolate. An hour nobody polled is an hour with no reading.
  • We do not accept owner-supplied statistics. An owner can write a description for their listing; they cannot set its player count or its uptime.
  • We do not take numbers from other directories. Everything is measured here or it is absent.
  • Paid placement changes where a listing appears and is labelled where it does. It does not change a listing’s measurements or its uptime.

How complete this directory is

It is worth being direct about this, because it is the thing a visitor is most likely to discover the hard way. The catalogue names 147 games, and 32 of them currently have at least one listed server. The remaining 115 have none.

Our guides cover many of those games anyway, because how to join a server is worth knowing whether or not this particular directory lists one. Where a guide is about a game we have no servers for, it says so at the top rather than sending you to an empty page.

A hub with fewer than five servers is marked so, and is not submitted to search engines — a thin page presented as a full one is the thing this site is trying not to be.

When something here is wrong

This page describes what the code does. If a figure on the site disagrees with what is written here, that is a bug worth reporting, and the contact page reaches someone who can fix it. The same goes for a listing whose measurements look wrong — a server behind a proxy, or one answering on a port other than the one listed, will produce readings that are honest about what we asked and wrong about what you would connect to.