How wheel randomness works: odds, repeats, and duplicate entries

The result is chosen before the animation

Wheel Tools chooses an active entry when you press Spin. The animation then moves the wheel to that entry. A slow finish or a stop close to a divider is part of the presentation; it does not give the neighboring entry an extra opportunity to win. Pressing the button harder or waiting for a particular moment is not a way to steer the result.

In browsers with the Web Crypto API, the wheel draws an unsigned 32-bit random integer. It rejects values above the largest whole multiple of the entry count, then uses the remainder to select an entry. This rejection step avoids the small bias that would otherwise come from dividing a fixed-size random range unevenly. If Web Crypto is unavailable, the implementation falls back to Math.random; that fallback is not cryptographic randomness.

Equal odds apply to entries, not unique names

With four active rows, each row has a one-in-four chance on an ordinary spin. If Alex occupies two rows and Bea and Casey occupy one each, Alex has two chances out of four. Duplicate rows therefore act as weights. The wheel does not know whether two matching names are the same person or two different people.

For equal chances per person, use one row per person and distinguish people who share a name, for example Alex A and Alex B. Review the active entry count after pasting. The wheel uses at most 200 active entries, so do not paste a larger roster and assume everyone is included.

Alex
Alex
Bea
Casey

Alex: 2/4 = 50%
Bea: 1/4 = 25%
Casey: 1/4 = 25%

Choose the repeat rule before the first spin

Ordinary spins sample from the same active list each time. A previous result does not become less likely just because it already won. With four distinct entries, a specified entry has a 25% chance on the next spin even if it appeared on the previous two spins. A short streak is compatible with random selection.

No repeats until all are picked uses a shuffled bag of entry positions. Each position appears once in a round, then a new round begins. The implementation also avoids selecting the previous round's final position as the next round's first position. This mode intentionally creates dependent draws; it is not the same probability experiment as independent spins.

Remove winner after each spin changes the active list. The remaining entries become more likely as the list shrinks. Use this for an elimination draw; use no-repeat mode for repeating rounds where the roster should remain available. Avoid combining rules unless you have explained the resulting behavior to the group. Duplicate labels can still produce the same visible name from different entry positions.

A small experiment you can reproduce

Start with four distinct entries and ordinary spins. Record 20 results without editing the list or discarding outcomes. Each entry has an expected count of five, but that is an average over repeated experiments, not a promise that this run will divide evenly. A record of seven, five, five, and three is not by itself evidence of an unfair wheel.

Now enable no-repeat mode and record a complete four-spin round. Each entry should appear once. That check tests the rule you selected, rather than expecting independent randomness to behave like a rotation schedule. Keep the list unchanged during the experiment: editing entries changes the pool and can reset the round.

What this tool does not prove

The wheel runs in the visitor's browser. It does not publish a signed audit log, commit to a public seed, or certify that a host has not changed the list or repeated a draw. A share link carries the entries and theme, not a verifiable history of how a winner was chosen. For an informal classroom or game-night pick, showing the roster and agreeing on rules can be enough. Do not describe a screen recording or a copied result as independent certification.