Paste a list and draw a winner fairly. Every draw uses your browser's cryptographic random generator with rejection sampling, not Math.random — which matters more than it sounds whenever the outcome has stakes. Math.random is not specified to be uniform, and the usual shortcut of taking a random number modulo the list length is measurably biased toward the entries at the top of the list. Neither is something you would want to defend if a giveaway result were questioned. Beyond a single winner you can draw several at once without repeats, remove each winner so the next draw excludes them, shuffle the whole list into a random order, or split everyone into balanced random teams. Duplicated names count as separate chances rather than being merged, which is what people expect when two entrants share a name. Nothing is uploaded and no list is saved.
| Method | Uniform? | Problem |
|---|---|---|
| Math.random() | Not guaranteed | Not specified for fairness; varies by engine |
| crypto value % length | No | Favours earlier entries when length is not a power of two |
| Rejection sampling (used here) | Yes | Redraws the biased tail; costs under two draws |
| Picking by value | — | Merges duplicate names into one chance |
| Picking by position (used here) | — | Two people named Sam get two chances |
Yes, and deliberately so. Indices come from crypto.getRandomValues with rejection sampling, which makes each entry exactly equally likely. The common approach — a random number modulo the list length — is biased toward the earlier entries whenever the length is not a power of two, and Math.random is not designed for outcomes that matter in the first place.
It is not specified to be uniformly distributed, its implementation varies between engines, and it is explicitly not intended for anything where fairness or unpredictability is required. For a rota it hardly matters; for a prize draw it is not something you could defend if challenged.
They are treated as separate entries, because picking is done by position rather than by value. Two people called Sam get two chances, which is the fair reading. Turn on dedupe if your list has accidental repeats you want collapsed.
Yes, two ways. Drawing several at once never repeats a position, and the remove-after-picking option takes each winner out of the list so subsequent draws exclude them.
No. The list lives in the page while the tab is open and nothing is uploaded, logged, or written to storage. Refreshing clears it — worth knowing before you close the tab on a long entry list.
Every draw comes from crypto.getRandomValues rather than Math.random, which is not specified to be uniform and is explicitly unsuitable where the outcome matters. Turning a random 32-bit value into an index needs care: taking it modulo the list length is biased whenever the length does not divide 2^32 evenly, which quietly favours the entries at the start of the list — with a list of three, the first entry would win fractionally more often than the other two, forever. Rejection sampling removes that: values falling in the incomplete final block are discarded and redrawn, which costs under two draws on average and makes the result exactly uniform. Multiple picks and group splits use a Fisher-Yates shuffle with each swap index drawn the same way, so every ordering is equally likely. Selection is by position rather than by value, so two entrants who happen to share a name are treated as two separate chances rather than being silently collapsed into one. Nothing is uploaded and no list is stored.