Benchmark consistency check
One result can be lucky or noisy. Enter 3 to 20 repeats from the same test and setup to see how tightly the values cluster before you compare a change.
Repeat-run consistency analyzer
Enter 3 to 20 non-negative results from the same benchmark and setup. Separate values with commas, spaces or new lines.
Example values are ready. Replace them with your own repeats or analyze the sample.
Check repeatability before judging a change
Run the same benchmark several times under the same conditions, then paste the results here. BenchRush calculates descriptive statistics in this browser and does not add these values to PlayerScore or saved history.
Repeatability comes before a confident comparison
A tight cluster makes a before/after delta easier to interpret. Variation alone does not prove accuracy, speed or hardware health.
How to collect a useful sample
- Keep one setup. Use the same device, browser, power mode and test method.
- Run at least five repeats when practical. Three is the minimum for this quick check; more repeats make a noisy pattern easier to see.
- Avoid cherry-picking. Enter every valid run in sequence, including ordinary slow or low results.
- Compare like with like. Milliseconds, hertz, percentages and scores are different measurement boundaries.
What the variation label means
The label uses the coefficient of variation: the population standard deviation divided by the absolute mean. Under 2% is shown as very tight, under 5% as tight, under 10% as moderate and 10% or more as high variation. These are practical BenchRush labels, not laboratory pass/fail limits or population percentiles.
Use the median when one run jumps
The median is less sensitive than the mean to one unusually high or low value. The possible-outlier note uses a median-absolute-deviation screen only as a prompt to inspect the run; it never deletes a value or declares it invalid.
Then compare the change
Once both conditions are reasonably repeatable, use Setup Compare for the before/after delta and the methodology for the measurement boundary.
Questions people ask
Does this save or upload my results?
No. The values are parsed and analyzed on this page only. They are not added to PlayerScore, local history or a remote account.
Is a low variation percentage automatically good?
It means the entered values are tightly clustered relative to their mean. It does not prove the benchmark is accurate, the hardware is healthy or the result is fast.
Should I remove an outlier?
Only when you can identify a real invalid run, such as an interruption or a different setup. Do not remove a result merely because it is inconvenient.
Privacy by default
Your personal bests, XP and streak are stored locally in this browser. The current tool does not require an account or upload your individual run to a global leaderboard.