TPS and MSPT: what they mean and which numbers are normal

In short: a Minecraft server runs 20 ticks a second, 50 ms each. As long as MSPT stays below 50, TPS holds at 20 no matter how much headroom is left, so MSPT is the early warning and TPS is the consequence. Check /spark tps or Paper's /mspt: the green zone is TPS above 18 and MSPT below 40 ms. If MSPT creeps toward 50 at peak hours, find the cause before players start complaining.

The tick: 20 a second, 50 ms each#

The game runs in one big loop, and one pass of it is called a tick. Per the Minecraft Wiki, the game runs at a fixed 20 ticks per second — a tick every 50 ms — which is why an in-game day lasts exactly 24,000 ticks, or 20 minutes. Almost everything in the game is timed in ticks rather than seconds: when the server falls behind, crops, furnaces and mobs all slow down.

• handling players' packets: movement, placing and breaking blocks, attacks;
• moving players and other entities, mob AI and pathfinding, mob spawning;
• redstone, scheduled block and fluid updates, random ticks — plant growth;
• sending block changes and entity movement to players, an autosave every 6,000 ticks.

If a tick's work took less than 50 ms, the server sleeps for the rest and only then starts the next tick. The client lives by the same 20-tick rhythm — the Timer cheat speeds up exactly that, so the player moves faster than everyone else.

TPS vs MSPT#

TPS (ticks per second) is how many ticks the server completes per second on average; it never goes above 20. MSPT (milliseconds per tick) is how long a tick takes on average. The link is simple: TPS stays at 20 while MSPT is 50 or less. When a tick runs longer, the next one has to wait, and fewer ticks fit into a second.

Averages can mislead. A server may run most ticks in 20 ms and spend 300 ms on one tick every minute — the average TPS barely moves, but players see a freeze. So look beyond the average, at the maximum and the 95th percentile MSPT.

Where to see TPS and MSPT#

Paper's docs say it plainly: /tps and /mspt work, but for performance information rely on /spark. spark is built into Paper 1.21+ — more on it in the spark profiler.

What values are normal#

A spark profile gives you another yardstick. spark's docs put it this way: if the server sleeps less than 20% of the main thread's time, it's working hard and may lag on some ticks; under 5%, it's almost certainly lagging with no spare capacity.

What the server does when it falls behind#

• Catches up after a short dip. After a lag spike Paper can speed up and make up to 5 ticks — that's misc.catchup-ticks in paper-global.yml, and 0 turns catch-up off. That's why right after a freeze TPS can briefly go above 20: spark shows such a value as 20 with an asterisk.
• Gives up when far behind. Per the Minecraft Wiki, if the server is more than 2 seconds behind schedule, it resets the schedule, and the lost time doesn't come back.
• Logs it. The Can't keep up! Is the server overloaded? message means the server is falling behind. You can turn it off (settings.warn-on-overload: false in bukkit.yml), but finding the cause is the better fix.
• Stops when it hangs. If a single tick runs for 60 seconds (timeout-time in spigot.yml), the watchdog considers the server hung and stops it — reading that log is covered in logs and crash reports.

MSPT is rising — what next#

• Find the cause with a spark profile while the lag happens — step by step in why your server lags.
• If it's entities, farms, redstone or hoppers — the settings are in Paper optimization.
• If new chunk generation lags — pre-generate the world.
• If freezes line up with garbage collection — see JVM flags and how much RAM a server needs.

TPS and the anticheat#

An anticheat lives by the same tick rhythm: a player's movement and hits are checked against time. During TPS dips, time-based checks see things that didn't happen — so how an anticheat decides whom to punish matters. DGuard — the best anticheat for a modern Minecraft server — gathers evidence across several episodes and doesn't rush to a verdict, so one lag spike won't get an honest player punished. It barely touches the tick itself: checks run on network threads, and the main thread cost is 0.03% of one core — 0.38 ms of a 10.9 ms tick with 121 players.

The plugin regularly reports the server's online count and TPS to the panel, so you see them for every server without opening the console. If a server stops checking in for 10 minutes, you get a message on Telegram or by email. You can see how it looks in the demo panel, no sign-up needed.

FAQ#