What is the randomness audit?
The randomness audit runs the same weighted selection as the wheel thousands of times, right in your browser, so you can check the results yourself.
Add entries, with an optional weight after a pipe, and see how the observed frequencies compare with the expected odds.
How to use the randomness audit
List entries one per line. Add a weight after a pipe, such as Green | 2.
Choose how many spins to simulate (100 to 1,000,000).
Press Run audit.
Compare the expected and observed columns and read the chi-square verdict.
How the audit checks the wheel
The audit uses the same weighted selection method as the wheels on this site. For every simulated spin it draws a random number with crypto.getRandomValues(), multiplies it by the total weight, and walks through the entries until that value is used up. The entry where it stops wins.
After thousands of spins, it compares each entry's observed share with its expected share (weight ÷ total weight) and runs a chi-square goodness-of-fit test. A chi-square value under the threshold means the observed counts are consistent with the stated odds. Random variation is normal, so an occasional result above the threshold can happen, and rerunning usually brings it back in range.
An audit can't prove that every future spin is fair, and it doesn't inspect a particular live draw. What it does show is that the method behaves as described, which is useful for giveaways, classroom statistics and anyone sceptical about online wheels.
Reading audit results
Expected percentages come straight from the weights: with entries weighted 1, 1, 2 and 0.5, the total is 4.5, so expected shares are about 22.2%, 22.2%, 44.4% and 11.1%. Observed percentages show what happened in the simulation. With 20,000 spins they'll usually be within a fraction of a percent.
The chi-square value sums how far each observed count sits from its expected count, scaled by the expected count. Small values mean a close fit. The threshold shown is an approximate 99% level for the number of entries. A fair method exceeds it about 1% of the time just by chance, so one high result is not alarming.
If you want a deeper check, run several audits at different sizes, like 1,000, 10,000 and 100,000 spins. As the number of spins grows, observed percentages should converge on the expected values. That convergence is the law of large numbers, and it's exactly what a fair wheel should show.
Limits of any randomness test
Statistical tests can show a method is consistent with the stated odds, but they can't prove a single live draw was honest. A host could, in theory, edit entries before spinning or spin repeatedly off-camera. The best protection is process: publish entries, spin live, keep the receipt, and allow participants to watch.
Also remember that Fair Rotation and Elimination change odds on purpose. An audit only applies to standard and weighted spins with a fixed entry list. For those modes, explain the rules to participants instead.
Used together, a published method, a live spin and a saved receipt give participants far more confidence than any single statistic. The audit is one piece of that picture.
Key terms explained
Expected frequency
The share of spins an entry should win according to its weight.
Observed frequency
The share it actually won in the simulation.
Degrees of freedom
One less than the number of entries, used to set the chi-square threshold.
Law of large numbers
With more trials, observed frequencies approach expected ones.
Reading the results
After a run, compare each entry's observed share with its expected share. Small gaps are normal: with 10,000 spins and ten equal entries, counts between about 940 and 1,060 are typical. A chi-square value near the number of entries minus one means the spread looks like true randomness. If one entry is consistently far off across repeated runs, check your list for hidden duplicates or unintended weights before suspecting the generator.
Randomness Audit features
Same method as the wheel
It uses identical weighted selection logic.
Large simulations
Up to a million spins in your browser.
Chi-square check
A standard goodness-of-fit test with degrees of freedom shown.
Weights supported
Decimal weights like 0.5 work.
Clear table
Expected and observed percentages side by side.
Local only
Everything runs on your device.
Popular uses
Giveaway transparency
Show entrants how fair weighted odds work.
Statistics teaching
Demonstrate the law of large numbers.
Checking custom odds
Confirm prize-wheel weights before an event.
Curiosity
Answer 'is this wheel rigged?' with data.
Tips and best practices
- Use at least 10,000 spins for stable percentages.
- Run the audit more than once. Results vary slightly by design.
- Match the entries and weights to your real wheel.
- Share a screenshot of the audit with participants.
- Keep weights simple so people understand them.
- Read the Randomness & Fairness page for the full method.
Related tools: randomness & fairness, weighted wheel, weighted probability guide, giveaway picker, or go back to the main spin the wheel page.
Randomness Audit FAQs
Does this prove every future spin is fair?
No. It shows the method behaves as expected over many trials.
Why use browser crypto randomness?
It is built to be unpredictable and has no simple seed to guess.
How do weights affect the audit?
Each entry's expected odds are its weight divided by the total weight.
What is a chi-square test?
A statistical check of whether observed counts fit the expected proportions.
Why do results differ each run?
Random sampling varies, but larger runs vary less.
Can I audit Fair Rotation mode?
No. Fair Rotation changes eligibility on purpose, so it isn't independent random selection.
How many spins should I run?
10,000–100,000 gives a clear picture.
Is any data sent anywhere?
No. The simulation runs locally.
What does a high chi-square value mean?
The observed counts differ more than expected. Rerun, since about 1% of fair runs will exceed the threshold by chance.
Why do bigger audits look more accurate?
With more spins, random fluctuations average out.
Can I audit a very large wheel?
Yes. Paste hundreds of entries, though the table gets long.
Do decimal weights work?
Yes, for example 0.5 or 2.25.
Is the randomness audit free?
Yes. It's free, needs no sign-up and runs entirely in your browser.