Reaction Time Distribution
This is the actual distribution of reaction-time scores submitted by people who took the ReflexArcade reaction time test. Each score is the average of five rounds, per our methodology. It is not an estimate and not sourced from anywhere else — it is exactly what was measured, updated on recomputed from the live dataset at most a minute apart.
How to read this page
- N is the number of submitted scores in that segment. It is always the true submitted count, whether or not any values were excluded as outliers for the percentile calculation.
- Percentiles. p50 is the median: half of scores fall below it. Because reaction time is measured in milliseconds, lower is better here — p90 is the slowest decile, not the fastest. At sample sizes near 10, p90 rests on the slowest one or two observations and should be read loosely.
- Rounding. Reaction time is rounded to the nearest whole millisecond.
- Outliers. In segments with at least 10 submissions, we exclude values outside 1.5× the interquartile range (Tukey's rule) before computing percentiles, and state how many were excluded — including when that number is zero. The rule is applied to each segment on its own, so an overall table and a device table can exclude different values. Below 10 submissions we do not attempt this — there is not enough data for the rule to mean anything.
- Segments below 10 submissions are labelled "insufficient sample" with the sample size shown, and no percentile is computed. We would rather show nothing than a percentile built on a handful of points.
Measured distribution
Reaction time test (5-round average)
Reaction time test (5-round average): insufficient sample (N=7) — not enough data has accumulated for a reliable percentile, so we omit one rather than show a shaky number.
By device — Reaction time test (5-round average)
desktop: insufficient sample (N=4) — not enough data has accumulated for a reliable percentile, so we omit one rather than show a shaky number.
mobile: insufficient sample (N=3) — not enough data has accumulated for a reliable percentile, so we omit one rather than show a shaky number.
Why this differs from "average reaction time" articles
Published reaction-time averages (including our own age-group breakdown, which cites external aggregated research rather than this site's own users) rarely state a sample size or a collection period. This page states both, every time, and says plainly when a breakdown is too thin to trust.
Next
Take the reaction time test yourself and add to this dataset, read the methodology behind the timing and the outlier rule, check the privacy policy for exactly what is stored, or go back to the data hub.