Is the wheel fair? 100,000 spins, tested

How the wheel picks a winner

When you spin, the wheel turns a random number of extra turns: between 5 and 10 on a computer, fewer on phones so the animation stays smooth. That number comes from your browser's cryptographically secure random number generator (crypto.getRandomValues), the same source browsers use for encryption keys, not the weaker Math.random. Wherever the wheel stops, the entry under the pointer wins. Because the random part always spans a whole number of turns, every stopping angle is equally likely, so every slice gets a chance equal to its size. With no weights set, all slices are the same size.

How this test works

The test does not animate anything. It runs the wheel's own code for choosing the number of turns and for reading the slice under the pointer, 100,000 times in a row, starting each spin where the last one stopped, exactly as the real wheel does. It then counts the wins for each entry. On a fair wheel with 10 entries, each should win about 10,000 times. They will never be exactly equal: with 100,000 random spins, a few hundred either way is normal.

What the chi-squared test tells you

The chi-squared test answers one question: are these counts further from equal than chance alone would make them? For each entry it takes the gap between its wins and its fair share, squares it, divides by the fair share, and adds the results up. That total is the chi-squared value. A fair wheel usually lands near the number of entries minus one (the "degrees of freedom").

The p-value turns that total into a plain probability: how often a perfectly fair wheel would give counts at least this uneven. A p-value above 0.05 means there is no evidence of bias. Below 0.05 is unusual but not proof of anything on its own: by definition a fair wheel falls below 0.05 in about 1 test out of 20. That's why the honest check is to run it again. A biased wheel fails every time.

What this test can't show

It checks that every entry has the same chance across many spins. It can't predict any single spin, and it doesn't test entries you give extra weight on purpose: those win in proportion to their weight. For running a draw people trust, see how to pick a winner fairly or use the raffle picker.