Minecraft anticheat guide
Backtrack Check Coverage Map: Reading Latency Alerts
Per FairPlay engineering notes, backtrack coverage asks separate latency questions. Learn how staff should read which question fired before acting.
Short answer: what the backtrack coverage map covers
Per FairPlay engineering notes, backtrack coverage is a set of separate checks. Each one asks a different question about latency that seems to lag behind where a target really was. The notes list seven backtrack checks. They read the same latency samples but look for different patterns: smoothing, repeated consistency, distance bias, client tick direction, movement resumption, combat versus non-combat shifts, and round-trip spikes. Read the question that fired, not a single latency number.
Why the checks are split by question
A single latency threshold would fire on ordinary lag. The notes reject a standalone ping or round-trip threshold, a one-sample spike, and a distance mismatch on its own. They also reject a second raw round-trip check. That check would reuse the same input and add no new causal boundary, which means more false positives under normal lag.
Each retained check adds a different condition: a stability control, a direction, a proximity range, a movement event, or a joint requirement.
| Question staff are asking | Coverage authority in plain terms | What must also be true |
|---|---|---|
| Does latency run higher when the target moves away than when it is closer? | Smoothed latency asymmetry | Enough samples, with noise staying low |
| Does that away-versus-closer gap repeat across batches? | Repeated consistency | Movement and distance controls agree |
| Does latency bias appear only at close range, where interpolation is plausible? | Proximity latency bias | Distance sits in the close or far bucket |
| Do client flying ticks lean the same way as latency? | Tick direction asymmetry | Same target and a bounded attack age |
| Does an attack land soon after movement resumes? | Release timing | A real movement gap, and connection leniency below the guard |
| Do combat hits show both a latency shift and a tick shift? | Combat latency shift | Non-combat latency stays stable |
| Do accepted hits line up with round-trip spikes against a baseline? | Combat spikes (experimental) | A stable baseline and a clean control window |
How to read a backtrack alert, step by step
Work through these steps before you act on an alert. Record your answer to each one in the case notes.
Fail-open behavior you should expect
Non-positive round-trip samples count as unusable. The notes say they stay out of the baseline, do not take part in spike matching, and do not expire pending hits as negatives. The pipeline fails open for that sample. Missing data should lower your confidence, not create a flag.
Ack-gap metrics are collected for diagnostics only and do not change detection math. The notes also say no claim is made that thresholds were tuned from a zero round-trip trace. Treat that as an open question, not a guarantee.
What the notes do not show
The backtrack notes describe code behavior. They do not record a live-server run, a false-positive rate, or a detection rate, and they make no claim of resistance to any particular client. Treat the coverage map as a list of questions the code asks, and test your own server before you rely on any single check.
A coverage map shows which code paths exist. It cannot tell you how often an alert will fire on your players, so measure that yourself.
Before you count alerts from the experimental combat spikes check, confirm its state in your configuration. Then test it in staging with legitimate high-latency connections you control.
FAQ
Does high ping on its own trigger a backtrack check?
Per FairPlay engineering notes, no. A standalone ping or round-trip threshold is rejected as a detection authority. Latency counts only alongside attack timing, direction, and target context. Still, run high-latency test accounts in staging before you trust your alert volume.
Can a backtrack alert fire without an attack?
The notes say the checks need an attack on a valid target within a bounded age window. A tick-only pattern with no attack context is rejected, so a backtrack alert should always point to a specific hit.
What happens to backtrack evidence after a teleport?
Per the notes, per-player state clears on teleport and on quit. Earlier samples do not carry over into the player's new position, so compare the alert time against the teleport time.
Why is the combat spikes check described as experimental?
The notes describe it as experimental and default-disabled. It builds its baseline from round-trip samples, and its diagnostics still run when the check is off. Confirm its state in your own configuration before you count its alerts.