How to Detect Cheating on Reaction Tests
What honest scores look like, and what manipulated ones almost always do.
TL;DR — Cheated reaction tests usually show one of four telltale patterns: sub-150 ms vision scores, sub-30 ms variance across attempts, runs that all end in the same final digit, or timing-source spoofing.
Genuine reaction-time data is messy. Variance, outliers and slight asymmetry are the signature of biology. The cleaner the numbers look, the more likely something is wrong.
Pattern 1 — Sub-floor times
Any visual reaction under 140 ms is biologically suspect. The nerve conduction lower bound is 150 ms even for elite athletes. Sub-100 ms scores are physically impossible without anticipation or input manipulation.
Pattern 2 — Unrealistically low variance
A real human varies 30–80 ms across 10 attempts. Bots and macros produce variance under 5 ms, often in distinctive cycles tied to refresh rate. A leaderboard run where every attempt lands within 8 ms of the average is almost certainly automated.
Pattern 3 — Round-number bias
Some cheats inject scores manually. Those scores cluster around round numbers (150, 175, 200) because humans pick round numbers. Real measurements distribute uniformly across all ending digits.
Pattern 4 — Timing-source inconsistency
Modern tests can compare the reported reaction time to the time between renderer events. A run where the input is faster than the render path itself never happened — the test was bypassed.
What honest data looks like
| Metric | Honest range | Suspicious range |
|---|---|---|
| Best attempt | 155–280 ms | Under 140 ms |
| Average | 180–320 ms | Under 160 ms |
| Variance | 30–80 ms | Under 10 ms |
| Final-digit spread | Uniform | Clustered at 0/5 |
Frequently Asked Questions
Can a person legitimately score under 150 ms?
On vision, no. On audio, occasionally (down to about 110 ms) under perfect conditions.
Does SERO check for cheating?
Yes. Server-side validation reviews timing patterns and rejects impossible sequences.
Why do some sites have leaderboards full of sub-100 ms scores?
Either no validation at all, or no randomised stimulus delay (so users anticipate). Both produce meaningless numbers.
Test Your Reaction Time Now
Free. 90 seconds. Global leaderboard. No download.
Start Test