CPS Limits vs Autoclicker Detection: Setting Thresholds

Why a flat CPS cap fails in both directions

A CPS limit is one number. Any click rate above it raises an alert. That is simple to set, but it measures the wrong thing. A player's peak clicks in a short burst tell you little about whether their clicking is automated across a whole fight.

A flat cap has two failure modes. It flags fast legit players during normal fights. It also misses automated clicking that stays below the cap. An autoclicker that limits its own rate can pass a cap set for human speed.

Which measurement methods trade off against each other?

Each method answers a different question. A flat cap is a cheap first filter. Averages over longer windows are steadier, but they can hide short bursts. Variance and pattern checks look at how the clicks are spaced, which is closer to what an automated clicker does wrong, but they need clean data to tune.

Human review of replays or recorded sessions gives the most context. It is also the slowest method, so use it for the cases the automated checks cannot settle.

How do I measure clean players before setting a limit?

Build the baseline before you change any number. Use sessions from staff and trusted players who agree to test in staging, and keep the input device and client version next to each sample.

One player's sessions are not a baseline. Include several people, because hand speed and input habits vary widely between players.

• Collect clean sessions from several players in staging, each with a recorded client version and input device.
• Record click timing for each session in the same format, so the samples can be compared directly.
• Include both calm and high-intensity fights, so the sample reflects real combat.
• Keep a separate set of known automated sessions if your team can produce them safely in staging.
• Write down the distribution of each measure for the clean group before you look at the automated group.

How do I tune a threshold without punishing legit players?

Tune in alert-only mode first. Alerts let you see how often the check fires on clean play before any punishment depends on it. Move the threshold in small steps and record the alert count after each step.

Compare each step with the clean baseline. If the alert rate rises on clean sessions, stop and find the cause before you go further. Use your appeals log as a second source. Repeated appeals on the same check are a sign the threshold is too tight for your players.

Watch the spread, not only the average. If a few clean players sit near the threshold, a small change in the limit moves them across it. Look at the top end of the clean distribution first. The threshold should sit clearly above that group, and the automated samples should still cross it. If the two groups overlap, the check needs a different measure, not a higher number.

• Set the check to alert-only and run it through a normal staging week.
• Adjust one threshold at a time, in small steps.
• After each step, count alerts from the clean sessions and from the automated sessions separately.
• Stop when clean alerts are rare and the automated sessions still trigger.
• Enable punishments only after the reviewers agree on the sample set and the result.

# Count alerts for one check in the current log (placeholder check name)
grep -c "<check-name>" logs/latest.log

What we have not verified

This article does not recommend a CPS number. A value from another server may not fit your players, input devices, or fight style. We have not tested any threshold against a specific anticheat, so do not copy one into your config without measuring it yourself.

The clean and automated samples on your server are the only data that decide the threshold.