IS THE WHEEL REALLY RANDOM?
How the winner is picked, why the same name sometimes comes up twice in a row, and what happened when we spun, drew and rolled millions of times.
Whenever someone gets picked twice in a row, someone else asks whether the wheel is rigged. It's a fair question, and it can be answered with data. We spun this site's wheel 1,200,000 times and also tested the name picker, dice, coin, numbers, cards and rock paper scissors. Here are the method and the results.
How the winner is picked
In every tool the result is decided before the animation, using the browser's cryptographic random generator, the same kind used to create passwords. With n options, the winner is Math.floor(Math.random() × n), and on this site Math.random draws its numbers from that cryptographic generator: every option has exactly a 1/n chance. The wheel then spins to show it. The order of the list, the speed, the time, who presses the button and previous results make no difference, and there are no hidden weights or favorites. The code is in the page itself, for anyone to see.
The test: 1,200,000 spins with 6 options
If the wheel is fair, each of 6 options should come up 16.67% of the time. This is what happened:
| Option | Expected | Measured |
|---|---|---|
| Option 1 | 16.67% | 16.74% |
| Option 2 | 16.67% | 16.59% |
| Option 3 | 16.67% | 16.66% |
| Option 4 | 16.67% | 16.67% |
| Option 5 | 16.67% | 16.71% |
| Option 6 | 16.67% | 16.63% |
None is more than 0.5% away from the expected value. The chi-square test gives a p-value of 0.052: the differences are the kind randomness produces, not a trick. To be suspicious you'd need a very small p-value, below 0.01, repeated in another test.
Why does the same name come up twice in a row?
Because the wheel has no memory: each spin starts from scratch. With 6 options, the chance of repeating the last winner is 1 in 6, and in our spins it repeated 16.66% of the time, almost exactly. With 10 names, a repeat happens 1 spin in 10. What would be strange is the opposite: never repeating. Thinking “it must be someone else's turn” is the gambler's fallacy. If you want nobody repeated until everyone has been picked, use the random name picker, which draws each name only once.
The other tools
We did the same with the rest, each with its own way of choosing a result:
| Tool and trials | Expected | Measured | p-value |
|---|---|---|---|
| Wheel (6 options) 1,200,000 · each option | 16.67% | 16.59%– | 0.052 |
| Draw of 10 names 1,000,000 · each name in each place | 10.00% | 9.77%– | 0.240 |
| 6-sided die 1,000,000 · each face | 16.67% | 16.62%– | 0.634 |
| Coin flip 1,000,000 · heads | 50.00% | 50.07% | 0.137 |
| Numbers 1 to 100 1,000,000 · each number | 1.00% | 0.973%– | 0.434 |
| Card from a 52-card deck 1,040,000 · an ace | 7.69% | 7.72% | 0.803 |
| Casino roulette (0–36) 1,110,000 · zero | 2.70% | 2.70% | 0.869 |
| Rock paper scissors 999,999 · rock / paper / scissors | 33.3% | 33.4% / | 0.419 |
In the draw of 10 names, each name should land in each place (first, second… tenth) 10% of the time: it did so between 9.77% and 10.25%. Every p-value is above 0.05: no tool strays from what fair randomness gives.
What a p-value means
The chi-square test compares what came up with what should have come up, and works out the chance of seeing differences this big (or bigger) if everything were fair. A high p-value means “nothing unusual”; a very low one, such as 0.001, would raise suspicion. Note that even when everything is fair, a low p-value turns up now and then by pure chance (1 test in 20 falls below 0.05), which is why you look at the whole picture.
What we don't do
- No weights or favorites: every option always has the same chance.
- We don't store who wins: the draw happens on your device and the names never leave it.
- Nobody can choose the result: not you, not the organizer, not us. If others need to see it, draw live or record it.
The data
The tests were run in the browser with the same code the tools use. The full counts are in this data file, free to cite. In the name draw, since each draw places all ten names at once, the hundred counts aren't independent, and the test accounts for that (81 degrees of freedom, not 99). The dice figures, with their tables, are in dice probability.
Frequently asked questions
Is the random wheel rigged?
No. The winner is chosen with the browser's cryptographic generator and every option has the same chance. In 1,200,000 spins with 6 options, each came up between 16.59% and 16.74% of the time.
Why does the same name come up twice in a row?
Because every spin is independent. With 6 options, repeating the last winner happens 1 time in 6; in our test, 16.66%. If you don't want repeats, use the random name picker, which draws each name only once.
Do the order of the list or the spin speed matter?
No. The result is decided before the animation and does not depend on the order, the speed or previous results.
Can I make a specific name win?
No. There are no weights or hidden options for anyone. If you want to give an option more chance, add it to the list several times.
Which random number generator do you use?
The browser's cryptographic generator (crypto.getRandomValues), the same kind used to create passwords. It works the same on phones and computers.
Keep reading
- Dice probability — 2d6 and 3d6 charts, rolling a 6, doubles, the 7 and the d20 with advantage, with an experiment of a million rolls.
- How to do a fair random draw — Five steps that leave nobody room to complain: clean list, rules set in advance, a truly random method, recorded proof and backup winners.
- Random numbers: the guide — What “random” really means, how computers generate it, practical uses in draws and raffles, and the gambler's fallacy explained.