How Scoring Works

Score Composition

The total score is built from three main categories, each testing a different aspect of server performance:

Single-core and gameplay together make up 80% of the score because Minecraft is fundamentally a single-threaded game. The fastest disk in the world won't help if the CPU can't keep up with tick processing.

Single-Core Performance (40%)

This category directly measures the server's ability to maintain a stable 20 TPS (ticks per second) under CPU-intensive workloads.

Tests

• Redstone stress — Rapid piston activation and signal propagation
• Block physics — Falling blocks, water flow, and physics calculations

What Matters Most

• MSPT (Milliseconds Per Tick) is the primary metric. Unlike TPS which caps at 20, MSPT shows the true margin: a server at 20 TPS with 10ms MSPT has 40ms of headroom, while 20 TPS with 48ms MSPT is on the edge.
• Worst-case performance is weighted heavily. We look at percentiles (P1, P5, P95, P99) not just averages. A server that runs smooth 99% of the time but freezes for 200ms every minute will score lower than one with consistent 30ms ticks.
• Stability is rewarded. Low variance in TPS and MSPT across the test duration earns bonus points.

Gameplay Score (40%)

This is where real Minecraft gameplay is simulated. The plugin creates actual in-game scenarios and measures how the server handles them.

Sub-categories

Exploration

• Chunk loading speed and throughput
• Chunk ticking performance under load
• Lighting engine update speed

Entities

• Mob AI pathfinding
• Entity spawning throughput
• Performance with hundreds of active entities

Mechanics — the most important gameplay sub-category

• Hopper item transfer throughput
• Explosion calculations (TNT chain reactions)
• Tile entity ticking (furnaces, hoppers, etc.)
• World save performance under load
• Crop and tree growth (bone meal processing)

Why Mechanics Are Weighted Highest

Mechanical interactions are what players actually do on survival and technical servers. A server that handles chunk loading well but chokes on hopper chains isn't useful for a typical SMP or technical server.

Test-Specific Metrics

Each gameplay test measures what actually matters for that operation:

• Chunk loading measures chunks per second, not just TPS
• Hopper test measures items transferred per second
• Explosion test measures blocks destroyed per second and TNT count
• World save measures write throughput and TPS impact

TPS and MSPT serve as secondary indicators to ensure the server wasn't struggling during the test.

Hardware Score (20%)

Raw system capabilities independent of Minecraft load:

Memory — The most impactful hardware metric for Minecraft

• Sequential and random read/write throughput
• Memory latency
• Random access is weighted highest because Minecraft's world data access patterns are largely random

Disk I/O

• Sequential and random I/O throughput
• Latency at the 99th percentile (worst-case)
• IOPS (I/O operations per second)
• Matters for world saves, chunk loading from disk, and crash recovery

Multi-Core

• Parallel workload throughput
• Core scaling efficiency
• Bonus for higher core counts (useful for Paper's async features, GC, plugins)

Why Hardware Is Only 20%

Because Minecraft is single-threaded. A server with a blazing fast NVMe and 128 GB of RAM but a weak single-core CPU will still have bad TPS. Hardware matters, but CPU single-thread performance dominates.

Multipliers

After the base score is calculated, several multipliers adjust it to reflect real-world conditions:

Duration

How long the benchmark took to complete. Faster servers finish quicker, which is itself a performance indicator. Very slow completions (>20 minutes) receive a penalty.

Garbage Collection

GC pauses freeze the server's main thread. The scoring system penalizes:

• High GC overhead (pause time as % of total time)
• Individual long pauses (>100ms)
• High pause frequency (many small pauses)

This is why JVM tuning matters — a well-configured GC can be the difference between a B and an A rank.

Stability

Consistency across the entire benchmark. Measured by:

• TPS standard deviation across all tests
• MSPT standard deviation
• Cross-test consistency (is TPS similar across redstone, chunk loading, and mob AI tests?)

Inconsistent results suggest external interference (other processes, CPU throttling, shared hosting noise).

Synergy

A bonus for balanced performance. A server that scores well in all three categories gets a bonus, while one that excels in hardware but fails in gameplay gets a penalty. This encourages well-rounded server configurations.

Failed Tests

Each test that fails or produces invalid data applies a cumulative penalty. This prevents servers from getting a high score by only passing easy tests.

Rank Badges

The final score maps to a letter rank:

Key Takeaways

1. MSPT matters more than TPS — TPS caps at 20, MSPT shows the real picture
2. Worst-case matters more than average — Lag spikes kill player experience
3. Stability is rewarded — Consistent 18 TPS beats erratic 20-then-12 TPS
4. Real gameplay > synthetic benchmarks — Tests use actual Minecraft mechanics
5. GC tuning is free performance — Proper JVM flags can significantly improve your score
6. Hardware helps but doesn't dominate — 80% of the score depends on CPU single-thread and gameplay handling
7. Balance is valued — A well-rounded server outscores one with a single strength