How we rank Minecraft survival servers
The order on this list is ours. We play the survival servers, we rank them, and we call that a judgement rather than passing it off as arithmetic. Every number on the page is measured though: player counts, 24 hour ranges and Bedrock support all come from our own status pings, and no placement is for sale in any position.
Order last reviewed 2026-08-26 09:44 UTC. Player counts on the ranking page are pinged every three minutes.
The three inputs, re-run monthly
None of this is a formula. It is three sources of evidence that we go back over every month, so a position has to be re-earned rather than inherited:
- Hands-on testing. Dozens of testers play the servers on our shortlist, past the first hour, and report back on the things below.
- A player vote. We poll players on which survival server they would actually recommend, and re-run that poll monthly rather than citing an old one.
- Our own measurements. Every listed server is pinged every three minutes for its live count, its 24 hour range and whether it answers on the Bedrock port.
What comes out is an editorial judgement, and the page on our number one sets out the reasoning for the top spot specifically.
What the testers weigh
Roughly in the order it matters:
- How the survival actually plays. Whether the progression has anywhere to go past the first hour, whether the systems are built for this gamemode or bolted on, and how long a new player can play before hitting a wall.
- How much of it is behind the store. Whether ranks, crates and rewards are obtainable in game, and whether the expensive end of the store sells cosmetics, acceleration, or things you cannot otherwise get.
- Live population, measured by us. We ping every listed address every three minutes and keep the history. A server that cannot hold players does not stay on the list.
- Whether it is actually maintained. Seasons that land, updates that ship, staff who answer.
- Public standing. Where the server sits on TopG, Minecraft-MP and MinecraftServers.org, as a sanity check on our own view rather than as the deciding input.
We do not publish a weighted formula, because we do not use one. A list of ten servers ordered by an arithmetic score implies a precision that nobody ranking Minecraft servers actually has, and every site that publishes one is still choosing the weights until the output matches its opinion. This is our opinion, formed by playing them, and it is labelled as one.
What we measure rather than judge
Everything numeric on the ranking page is measured, not estimated:
- Player counts come from our own status ping to the server's own address, every three minutes. Nothing is copied from another list and nothing is typed in by hand.
- The 24 hour range beside each count is the real low and high we recorded that day.
- Bedrock support is a reply on port 19132, not a claim on a website. It takes two replies out of three attempts before we mark an entry as supporting Bedrock, because a single probe throws false positives.
- Version ranges are what the server itself advertises in its status reply, or what its own server message states.
- Network counts are labelled. Where a server reports one figure across every gamemode it runs rather than for its survival world, the entry says so, because otherwise the number answers a different question to the one you are asking.
When counts go stale
Three rules keep the numbers honest about their own age.
- If a server's last successful ping is more than 30 minutes old, its entry switches from "online now" to its seven day average and says so in those words.
- If our pinger has not had a single successful reply anywhere for two hours, every "live" and "now" on the site is replaced with the timestamp of the last verified reading, including in the JSON feed and llms.txt.
- Every entry prints how long ago it was last verified online, so you never have to trust that a number is current.
Two consecutive failed ping cycles also raise an alert on our side, so an outage on our end gets fixed rather than quietly served as fact.
One measurement quirk worth naming: a couple of large proxied networks answer status pings from more than one backend, reporting wildly different populations seconds apart. Where we have measured that, the entry is pinned to the backend that serves the main world and replies from anywhere else are discarded rather than averaged in.
What gets a server listed, and what gets it removed
- Every listed server has answered a real status ping from us. Nothing is listed on a claim alone.
- Descriptions state only what the server's own materials and our own measurements support. If we cannot verify a feature, it is not on the page.
- Placement is never sold. No server on this list has paid to be here, in any position.
- We publish no star ratings or review scores as structured data for any server we rank, because a rating you issue yourself about somebody else, on a page you control, tells a reader nothing.
- A server that stops answering pings for a sustained period comes off, and the changes page records the move.
- Some servers are excluded from these lists entirely at our discretion. They are not ranked last, they are simply not listed.
Corrections
The order comes from joining and playing every server on this list rather than from anybody's marketing copy. Placement has never been for sale and no server has ever paid to appear here.
If something here about your server is wrong, out of date, or reads as an unfair characterisation, we would rather fix it than argue about it. Factual errors get corrected the same day.