SCULK BENCHMARKS
These are the validation tracks behind Sculk's combat, movement, packet, and player checks. Precision, recall, and F1 are deliberately held until a versioned replay corpus and review protocol can be published with each result.
PRECISION
How often a raised signal is actually the intended behaviour.
RECALL
How much of the labelled behaviour a track identifies.
F1 SCORE
The balance that stops one impressive number hiding another problem.
TRACKS
A blank metric is intentional: it means the evidence is not ready to publish.
COMBAT
A replay track for rotation, target switching, reach boundaries, and click cadence that no player should have to guess at.
—
PRECISION
Keeps legitimate players out of the ban path.
—
RECALL
Measures whether the intended replay fixture is caught.
—
F1 SCORE
Balances both sides before a release can graduate.
CHECK FAMILIES IN THIS TRACK
MOVEMENT
A controlled set of terrain, effects, latency, and clock conditions for Fly, Speed, Timer, and Phase checks.
—
PRECISION
Keeps legitimate players out of the ban path.
—
RECALL
Measures whether the intended replay fixture is caught.
—
F1 SCORE
Balances both sides before a release can graduate.
CHECK FAMILIES IN THIS TRACK
PACKET
Malformed order, withheld movement, and anticheat-disabling packet patterns are reviewed against raw trace fixtures.
—
PRECISION
Keeps legitimate players out of the ban path.
—
RECALL
Measures whether the intended replay fixture is caught.
—
F1 SCORE
Balances both sides before a release can graduate.
CHECK FAMILIES IN THIS TRACK
PLAYER
AutoTotem, FastUse, and inventory movement need a larger labelled corpus before public detection figures mean anything.
—
PRECISION
Keeps legitimate players out of the ban path.
—
RECALL
Measures whether the intended replay fixture is caught.
—
F1 SCORE
Balances both sides before a release can graduate.
CHECK FAMILIES IN THIS TRACK
PUBLICATION POLICY
When Sculk publishes a figure, it should include the server build, replay conditions, labelled cases, and the false-positive review that produced it. Until then, these tracks show the scope without pretending uncertainty is a score.