GuidePlayer guides
How to find a good Minecraft server, and spot a dead one
Check 30-day uptime, players at your hours and the trend before you join. Why votes and star ratings mislead, with numbers from servers we ping.
You can tell a good Minecraft server from a dying one before you join by checking three measured things: whether it stayed up over the last 30 days, how many people are online at the hours you actually play, and whether that number is growing or shrinking. After that, read the rules and the store yourself, and give vote counts and star ratings the least weight of anything on the page.
Measured numbers are hard to fake. Votes and stars are easy to inflate. ServerList pings every listed server every minute and keeps the history, so you can check the measured side of every listing here.
Start with the mode, then your hours
The mode decides what you do all day, so choose it first. A world that a group builds up over months is an SMP. PvP where every kill changes your health bar is lifesteal. Classic progression with extras such as land claims or an economy is survival. The server types guide explains every other label.
Then think about when you play. A server packed at 20:00 in New York can be nearly empty when you log on after school in Berlin. Here is a day of player counts across every server we track, in your time zone:
Busiest at with about 17,845 players, quietest at . The busiest hour holds 2.6 times the players of the quietest.
Measured 19 Oct 2026, 22:55 UTCHow we measure
Single servers swing harder than the network. Among servers averaging 5 or more players, the busiest hour typically has 2.7 times as many players as the quietest. In the four weeks to 16 October 2026 that was 2.8 times, and more than 4 times for over a quarter of them. The breakdown by region and mode is in when Minecraft servers are busiest.
Signals you can measure
Uptime over 30 days
Uptime is the share of our pings that found the server playable, which is stricter than "it answered". Free hosts often answer with a sleeping placeholder when nobody is online, and others answer with "Maintenance" or zero slots. You cannot join a placeholder, so it does not count. The methodology page lists exactly which answers count as not playable.
What the numbers mean over a 30-day month:
| 30-day uptime | Downtime in 30 days | What it usually means |
|---|---|---|
| 99.9% or more | under 45 minutes | Restarts only |
| 99% to 99.9% | 45 minutes to about 7 hours | A few restarts plus a short outage |
| 90% to 99% | about 7 to 72 hours | Regular outages, a host move or a rough patch |
| Under 90% | more than 3 days | A free host that sleeps, a server in trouble, or one closing down |
- Never answered65
- Placeholder only18
- Under 50%25
- 50 to 90%21
- 90 to 99%19
- 99 to 99.9%31
- 99.9% or more33
83 of 212 measured listings were never playable. 33 were playable at 99.9% of our pings or more.
Measured 19 Oct 2026, 22:55 UTCHow we measure
In the 30 days to 16 October 2026, 78 of the 216 listings on serverlist.net were not playable even once: 64 never answered a ping, and 14 more answered only with placeholders. Among the 138 that were playable at least once, the median uptime was 98.9%, roughly 8 hours of downtime a month, and 67 managed 99% or better. Right now: 83 of 216 never playable, 64 at 99% or better, and a median of 98.9% among the rest. The uptime report breaks this down by region and host type.
A daily restart is healthy. A server that vanishes for a day every week is telling you something about its hosting or its budget.
Players at the hours you play
Look at the typical player count, not the peak. A peak of 200 can come from one event night. Our server pages show the typical count as a median of daily averages, which one busy weekend cannot drag up.
Most servers are small, and that is fine if you know it going in:
- 20
- 43Under 1
- 151 to 5
- 205 to 20
- 1620 to 100
- 17100 to 500
- 5500 or more
- Median server
- 4.8
- Five biggest
- 56%
- Ten biggest
- 71%
Each column counts servers by their average players. Half of the 118 servers average 4.8 players or fewer.
Measured 19 Oct 2026, 22:55 UTCHow we measure
In the two weeks to 16 October 2026, 119 listings were playable at least once. The median one averaged 4.4 players, 45 averaged fewer than 1, and the five busiest held 60% of all the players we counted. Right now: a median of 4.8, with 56% of players on the top five. More on this in how many players a Minecraft server really has.
Ignore the advertised slots. The number after the slash ("12/500") is whatever the owner typed into the config. Over the same two weeks the median server advertised 500 slots and at its busiest moment filled about 12% of them. Only 9 of the 119 ever reached half.
A small SMP with eight regulars at your hours can be the best server you ever play on. A server averaging under one player is a single-player world with extra lag.
Growing or fading
Compare the last two weeks with the two before. The stats page draws 30-day and 90-day curves, so this takes a glance. Of the 58 servers that averaged 5 or more players from 19 September to 2 October 2026, 6 grew by more than 20% over the next two weeks, 15 shrank by more than 20%, and 37 stayed within 20% either way. Right now: 7 growing and 12 fading out of 58.
A fade is not always bad news. Seasonal servers empty out toward the end of a season and fill again at the reset, so check the season schedule first. A steady fade with no season in sight, plus a listing nobody updates, is a server winding down.
Version support
The newest Java release is 26.3. When we wrote this it was 26.3, released on 15 September 2026. Servers that accept many versions usually run ViaVersion, which lets newer clients join, and ViaBackwards, which lets older ones in. On an older client, blocks and items it does not know show up as the closest older ones.
One catch: text such as "Velocity 1.7.2-26.3" comes from the proxy, the gateway software in front of a network. It lists what the proxy can speak, not what the game servers behind it accept, so the listing's own version list is the better guide. If your version is refused, the game says "Incompatible client! Please use" followed by the version the server wants (older game versions say "Outdated client" or "Outdated server"). At one check on 18 October 2026, 37 of the 108 servers playable at that moment named 26.3 in their version text. Right now: 35 of 100.
Region and ping
Light in fiber covers about 1,000 km in 5 milliseconds, so every 1,000 km between you and the server adds at least 10 ms to the round trip, and real routes add more. Building feels fine at 100 ms. In PvP you notice anything above about 80 ms. Go by the declared country, then add the server to your multiplayer list and watch the bars. Our pinger measures from one location, so the latency on our pages is ours, not yours.
Reading a real stats page
Here is what we measured for one listing in the four weeks to 16 October 2026, the same signals its stats page shows. We picked SurvivorSMP because it shows every signal at once, not as a recommendation, and the live page has moved on since.
| Signal | Measured | How to read it |
|---|---|---|
| Playable uptime, 28 days | 99.6% | Under 3 hours down in four weeks |
| Average players, last 14 days | 82.5 | Up from 48.6 in the 14 days before, a server on the rise |
| Busiest hour, 28 days | 16:00 UTC, about 83 players | 18:00 in Amsterdam, so evenings in Europe are lively |
| Quietest hour, 28 days | 00:00 UTC, about 49 players | Even the quiet hour has people around |
| Busiest days, 28 days | Friday and Sunday | Tuesday is the quietest day |
| Version text | Velocity 1.7.2-26.3 | The proxy's own range. The listing says it runs the latest release, so join on 26.3 |
| Country | Netherlands | Low ping from most of Europe |
Two minutes on that page tell you more than the description will. The description is what the owner hopes. The stats page is what happened. And when a chart shows one spike far above everything around it, trust the typical line. A lone spike is usually one event, or a glitch in what the server reported.
Signals you have to read
Rules and staff
Open the rules before you join. Good rules are specific ("no client modifications except those on this list", "griefing claimed land is a ban, unclaimed land is fair game") and tell you how appeals work. Vague rules ("be respectful, staff decide") usually mean decisions depend on who is online.
Look for a staff list and check whether questions in the server's Discord get answers within a day. Owners on serverlist.net can fill in a "Rules and moderation" section on their listing. A filled-in one at least shows the owner has thought about moderation.
The store and pay to win
Most servers have a store, and that is allowed. Mojang's Minecraft Usage Guidelines let servers charge for access, sell cosmetics and even sell things that affect gameplay, as long as those do not ruin other players' experience or give a competitive advantage. They also require servers to show their content and prices before you join, so read the store before you invest time.
Cosmetic ranks, chat colors, pets and particles are fine. Kits with enchanted gear, crate keys that drop top-tier items or "skip the grind" bundles mean paying players will always be ahead. Crate keys are paid random rewards, so treat them like loot boxes in any other game. The owner side is in making money within Mojang's rules.
Reviews that say something
A useful review names something specific: "the economy reset after three weeks and nobody announced it", "staff answered my ticket in an hour". "Best server ever!!" tells you how the writer felt the day they voted. Read the three-star reviews first, then the owner's replies, because how an owner answers criticism is how they will treat you when something goes wrong.
Why vote counts and star ratings can mislead
Votes measure how often a server asks for votes and what it pays out for them, not how many people play. In the 30 days to 16 October 2026:
- 15,648 votes were cast on serverlist.net, and only 54 servers received any. The three most-voted servers took 61% of them.
- Of the 10 servers with the most players on average, only 1 was among the 10 most voted, and 6 of the 10 received no votes at all.
- 1,324 votes, 8.5% of the total, were cast during an hour in which that server was not playable at any of our pings.
Right now: 15,237 votes in 30 days, 62% to the top three, 11.8% cast while the server was down, and 2 of the 10 most played servers among the 10 most voted.
Votes also run on their own clock. Many servers remind you to vote when you join, which is probably why votes bunch up a few hours before the busiest hour for players:
Share of votesShare of players
Votes peak at , players at , 3 hours apart. Based on 15,246 votes.
Measured 19 Oct 2026, 22:52 UTCHow we measure
Star ratings have a different problem. On serverlist.net you can only review a server right after voting for it, so most reviewers already like it. On 18 October 2026 there were 2,949 visible reviews with an average of 4.78 stars, and 86% gave five. Right now: 2,955 reviews averaging 4.8, 86% with five stars. When nearly everything has five stars, the average stops telling servers apart.
This is why our default order puts servers online in the last 24 hours first and only then looks at votes, and why players and uptime sit next to the vote count on every list. The methodology page explains the ranking in full, and how server voting works covers the vote side.
Try three servers for a week
- Shortlist three servers in your mode that pass the measured checks: 99% or better uptime, a typical player count you like at your hours, a flat or rising curve.
- Read each one's rules and store page. Drop any that sell gear or power you would need to keep up.
- Join each at the time you normally play, not on a Saturday afternoon if you play on weekday evenings.
- In the first session, ask something in chat and see whether anyone answers. Run
/helpand see whether the server explains itself. - Play a few sessions on each. Keep the one you think about when you are not playing, and vote for it, which is how servers you enjoy stay findable.
If you want a starting point, these SMPs were playable at the last ping with 99% or better uptime:
Ping 1 millisecond. ExtremeCraft EXTREMECRAFT Spears! Survival: New Eco TreasureRelics Bosses and more! 430 of 5,000 players online Up 99.99% in 30 days2,707 votes in 30 daysjoin.extremecraft.net
Ping 83 milliseconds. BendersMC FOUR ELEMENTS MINECRAFT SERVER ʜᴛᴛᴘꜱ://ᴅɪꜱᴄᴏʀᴅ.ɢɢ/ʙᴇɴᴅᴇʀꜱᴍᴄ 288 of 1,200 players online Up 100% in 30 days1,102 votes in 30 daysplay.bendersMC.co
Ping 87 milliseconds. AugustaMC AugustaMC [1.21.10] 👑 Towny | ★ ᴀᴅᴠᴇɴᴛᴜʀᴇ ᴀᴡᴀɪᴛѕ ★ | augustamc.com 26 of 500 players online Up 100% in 30 days952 votes in 30 dayssl.augustamc.com
Measured 19 Oct 2026, 22:55 UTCHow we measure
The site-wide numbers behind all of this are on the stats page.
FAQ
How many players should a good server have?
Enough to have people around at your hours, and that depends on the mode. A whitelisted SMP can be great with 10 regulars, while a minigame network needs hundreds so queues fill. Skip servers whose typical count at your hours is under 1.
What uptime is good?
99% or more over 30 days. Below 90%, you will often find it offline when you want to play.
Are paid ranks a red flag?
Not on their own. Cosmetic ranks and a few convenience commands are normal and pay for hosting. Ranks or crate keys that sell gear, money or protection other players cannot match are the red flag, and Mojang's guidelines do not allow paid extras that give a competitive advantage.
Read next
Keep going.
Why Minecraft servers ask you to vote, and how it works
What a vote does, how it travels from the list to your reward, the 24-hour rule, Bedrock names, and what to check when a vote reward never arrives.
Minecraft server types and slang, explained A to Z
Every common Minecraft server type, from anarchy to whitelist, in plain words, with live counts of servers you can join and the slang you will hear.