Stats

The numbers behind Hustle.

Every chart here is drawn so that a bigger bar means better. Measured in our own test lab during development, mostly with bot players. We'll re-run every test on the release build. The caveats explain exactly what each number means.

5,000bot players at a full 20 TPS on one 6-core machine
30× fasterthan Leaf (tuned) in the same 700-player fight
98%+of 2,435 tracked vanilla features matched
99.69%of world-gen blocks identical to vanilla
7/ 8popular Paper plugins tested and working
148chunks/sworld generation, up from 84, output identical to vanilla
TPSGame speed

20 TPS is full speed. Lower means the whole game slows down. Higher is better.

MSPTWork per update

Time each game update takes. Over 50 ms, the server lags. Lower is better.

SpareSpare capacity

How much of each 50 ms update is left over: room for more players. Higher is better.

Player scaling

What happens as the crowd grows.

A dense fight of bot players, with more added each run. Hover a point for every number.

HustleLeaf (tuned)Paper (tuned)

The 700–5,000 runs used earlier builds, before some vanilla systems were added. On the current build with every feature on: 500 players ≈ 5.2 ms and 1,000 ≈ 13.8 ms per tick. At 5,000 players the network card's packet limit was reached before the server's.

Multi-core

More cores, more speed.

The same 700-player fight, run with a different number of worker threads. Speed is compared to running on one thread.

1 thread9.3 ms/tick
1.0×
2 threads5.4 ms/tick
1.7×
4 threads3.4 ms/tick
2.7×
6 threads2.8 ms/tick
3.3×
12 threads2.6 ms/tick
3.6×

Vanilla parity

Faster, and still real Minecraft.

A real vanilla 26.3 server runs next to Hustle. Scripted players fight, mine, build, farm, trade and run redstone on both, and we compare the packets, blocks, health, drops and sounds.

58/ 58end-to-end checks passing against a live vanilla server, plus 1,300+ per-feature scenario checks
99.69%of terrain identical over 224 chunks. Two vanilla runs of the same seed only agree 99.68% with each other.
16/ 16Nether and End chunks identical to vanilla, byte for byte
47/ 47online-mode login tests identical to vanilla

Engineering

How much faster it got.

Chunk savingSame machine, same world
8.850.50 ms/chunk
17.7× faster
512 villagers at workBeds, workstations, a bell
51.914.8 ms/tick
3.5× faster
~1,900 hoppers, chests and furnacesA base full of block entities
10.53.95 ms/tick
2.7× faster
WorldEdit: 442k-block //setWorst tick, work spread across ticks
49~6 ms
8× faster
Creating a Multiverse worldServer stall while the spawn area generates
3.1 s5–41 ms
75×+ faster
Live stats coming soon

Across Victory Lap

Live from our servers.

Once we launch, this section will show real-time numbers from every Victory Lap server. The values below are placeholders.

Players online now—
Servers running—
Average TPS—
Biggest event this month—
Players online, last 24 hoursSample shape, not real data

Methodology

How to read these numbers.

  1. Bots, not people. Player counts were tested with scripted Java bot clients. Real players read packets at different speeds and move in more varied ways.
  2. Development builds. Measured in September 2026 on builds still in development. Some scale runs predate features added since. Everything will be re-run on the release build.
  3. Our tests of other servers. Paper and Leaf ran their newest supported builds (1.21.11) with our tuning on our hardware: an Intel i7-8700 (6 cores / 12 threads), 16 GB RAM, bots on a separate PC.
  4. About Folia. Folia splits the world into regions. Our tests were one big fight in one region, which isn't where Folia shines, so we left it off the chart.
  5. Noise. Some results came from a shared development PC with about ±1.5 ms of noise. Treat relative gains as solid and exact milliseconds as approximate.
  6. "98%+ parity" counts features on our tracker. The remaining gaps are tracked one by one.

Put Hustle to work.

Start your server →