Technical guide
FairPlay false positives: inspect, trace, change one check
Use /fp alerts inspect and /fp latency diagnostics to verify a FairPlay flag before disabling checks, changing exemptions or recording an ML window.
Products and contents
A FairPlay alert is a signal, not a verdict. Before you disable a check or grant an exemption, find out which check fired, whether the alert is for the player you think it is, and what the latency trace shows. The FairPlay product page covers scope and the docs cover install.
Point your alert feed at one player
/fp alerts <enable|disable|toggle|inspect|clear> [player] controls alerts for the staff member running it.
enable,disableandtogglechange your own subscription and save it. Player-only.inspect <player>sets your alert target to one player. It doesn’t print history. It narrows the live feed until you clear it.clearremoves that target.
Two permission details matter. The dispatcher first requires fairplay.command. AlertsCommand then checks fairplay.command.alerts, which isn’t in the shipped plugin.yml. Ops bypass it, but a non-op staff rank needs the node granted. fairplay.notify is separate and controls receiving alerts. If an admin sees nothing, check both nodes before you assume the check is quiet.
Verify the flag before touching a check
- Reproduce with the same client version and the same action. A flag that shows up only during a lag spike is a different problem from one that shows up on every hit.
- Record the check name, VLS and reason from the alert. FairPlay’s built-in modules cover aimbot, autoclicker, autototem, fastbreak, hitbox, killaura, scaffold and timer. Which ones are active depends on your build.
- Watch the player for a session without punishment commands.
/fp checks <enable|disable|toggle> <check|category>changes check state./fp toggle <check|category>is the shorthand. - Disable one check or category at a time, then reproduce. Disabling a whole category hides unrelated flags and makes the next report harder to read.
Combat-only posture matters. FairPlay has no movement prediction checks. Movement packets are telemetry for combat, aim, hitbox, interaction, scaffold, inventory and timing checks. A “movement false positive” usually belongs to one of those, not to a fly or speed module.
Trace latency instead of guessing
/fp latency <player> <seconds 1..120> starts transaction-latency diagnostics for one player through the BacktrackCombatSpikes check. It streams records and ends with a summary: transaction samples, usable and unusable RTT samples, zero-RTT samples, attack attempts, accepted hits, spike matches, analyses, baseline rejections, resets and dropped records.
This is a diagnostic, not a latency compensation setting. Run it while the player reproduces the flag, then read the summary before you change mitigations.yml. The permission is fairplay.command.latency.
If the flag comes from reach compensation rather than detection, turn on reach-increase.debug in mitigations.yml and read the reason chain. reach-increase applied means the mitigation state was stored. range, aim and visibility_* show the first failed gate. Only reach-increase damage-applied proves a MONITOR-confirmed applied hit. Without that distinction, you might “fix” a compensation log that never applied damage.
Record an ML window when the alert won’t reproduce
For aim-model review, use the global window recorder:
/fp ml aim window record GLOBAL
/fp ml aim window status
/fp ml aim window stop
Recording needs fairplay.command.ml.global. Files land in plugins/FairPlay/windows/global/<player>/<UTC-date-and-session>.json after async finalization. They are per-player rotation captures and haven’t been reviewed. Treat them as evidence to inspect, not training data to merge blindly. The machine-learning-training config block is for training only and doesn’t affect runtime detection.
Check the optional hooks and config state
GrimAC and Vulcan are optional softdepend entries. FairPlay loads and runs without either, so a missing hook is expected, not a failure. PacketEvents is the hard dependency, so a packet-level failure usually starts there.
There’s no bundled YAML. Config is generated from annotated Java classes. A missing path logs [FairPlay] Missing config value for path '…'. That means the running build expects a key your file doesn’t have. It isn’t automatically corruption. show-cancel-messages defaults to false and cancel-message-cooldown-ms to 60000, so a silent cancellation doesn’t prove nothing happened.
Exemptions beat global disables
The descriptor ships three targeted nodes, all default false:
fairplay.exemptkeeps tracking active while removing the player from checks.fairplay.nosetbackstops setbacks but leaves detection.fairplay.nomodifypacketstops packet cancellation or modification.
Use these for a known legitimate player or a tournament exception. They’re auditable and reversible. Disabling a check for everyone is neither.
Frequently asked questions
Is a FairPlay alert a verdict?
No. It is a signal. Check which check fired, which player it names and what the latency trace shows.
Should I disable a check when it fires?
Not first. Verify the flag, then prefer an exemption over a global disable.
Which command controls alerts?
/fp alerts <enable|disable|toggle|inspect|clear> [player]. The command applies to the staff member who runs it.