Randomness Audit
Check the wheel's fairness for yourself. Spin it up to 100,000 times with the same picking method the real wheel uses, and compare how often each entry wins with the odds it should have.
Type in your own entries, make some slices bigger than others, and see the results come in. When the run is done, copy a plain-language report to keep with a giveaway's records or to show a class. Ready to spin for real? Head to the wheel spinner or pick one of the templates.
How the Wheel Picks a Winner
Pick First
The moment you spin, the winner is chosen, before the wheel moves at all.
Use a Secure Source
The choice comes from the same kind of secure randomness used to make passwords, not a pattern that could be predicted.
Weigh It Fairly
Bigger slices win more often, in proportion to their size. When every slice is the same size, every entry has exactly the same chance.
Then Animate
The wheel turns until the chosen slice stops under the pointer. The animation shows the result; it never changes it.
How the Audit Works
The audit picks a winner the same way the real wheel does, as many times as you choose, without the animation, and counts how often each entry wins. It then works out how often each one should have won, based on the size of its slice.
No fair wheel hits those numbers exactly, so the audit measures how far the counts landed from them with a standard statistics test, the chi-squared test. Its p-value says how often a perfectly fair wheel would stray at least that far. Above 0.01, the run looks like fair chance; between 0.01 and 0.001, it is unusual but expected about once in a hundred runs; below 0.001, a fair wheel does it about once in a thousand runs, so run the audit again.
The audit also checks the random source itself: that it's available, that the page is secure, and that nothing on the page has tampered with it. That last check is a quick sign of trouble, not a full security review of your device.
As the U.S. National Institute of Standards and Technology points out in its guide to testing random number generators, statistical tests are a useful first step, but no set of them can certify a generator on its own. The audit is that first step, done in the open, where anyone can repeat it.
What the Audit Can't Tell You
An audit is a sample, not a proof. A good result shows the wheel's picks follow your odds across many spins, the same as a truly random source would, but it can't say anything about one particular spin beyond its odds. It also checks this browser on this device: the same code runs everywhere, but each browser supplies its own random source. And it is a transparent check you can run yourself, not a legal certification, gambling compliance or an independent cryptographic review.
References
- MDN Web Docs: Crypto.getRandomValues(), the browser function the wheel draws its random numbers from.
- MDN Web Docs: Web Crypto API, the browser's built-in cryptography, of which it is part.
- NIST SP 800-90A Rev. 1, the standard for the kind of random bit generators that cryptographic sources like it are built on.
- NIST SP 800-22 Rev. 1, NIST's statistical test suite for random number generators, on what tests like this one can and can't show.
Frequently Asked Questions
Quick answers about how the wheel works, what it saves and how fair it is. Want proof? Try the randomness audit.
Does a good audit prove every spin is fair?
No test can prove that, and an honest audit doesn't claim to. What it shows is that over many spins the wheel's picks match the odds you set as closely as a truly random source would. Every real spin uses the same code and the same random source, so it has the same odds.
What makes the wheel's randomness secure?
The wheel uses the same kind of randomness your device uses to create passwords. Its picks can't be predicted from earlier ones, so no one can work out what the next spin will land on.
How do bigger slices change the audit?
Each entry is expected to win in proportion to the size of its slice. If Pizza's slice is three times bigger than three other entries, Pizza should win half the spins and the others a sixth each. The audit compares every entry with its own expected share.
Why might a fair wheel show a warning?
Because randomness is lumpy. By the definition of the p-value, a perfectly fair wheel lands in the borderline range about one run in a hundred, and in the unusual range about one in a thousand. Run the audit again: a fair source won't keep doing it, and a biased one would.
Does the spinning animation affect who wins?
No. The winner is chosen the instant you spin, and the wheel is then turned to stop on it. Spin time, speed and where on its slice the pointer stops are all just for show.
Does the audit send my entries anywhere?
No. It runs entirely in your browser, like the wheel itself. Nothing you type is uploaded.
Ready to Spin?
Your wheel is waiting at the top of the page, just as you left it.