How to Pick a Name at Random — Without Anyone Doubting the Result

What fair actually means here

A draw is fair when every entry has the same probability of being selected. That sounds obvious, but it is broken constantly in small ways: a list where a name appears twice doubles that person's odds, and a physical hat favours whatever was dropped in last and sits on top. The list itself, not the randomiser, is where most unfairness enters.

Fairness also has a second half that people forget: the outcome has to be accepted. A mathematically perfect draw that nobody witnessed and that cannot be checked afterwards will still be disputed. Treat the draw as a process with an audience, not just a computation.

With or without replacement, and why it matters

Drawing with replacement puts each name back after it is picked, so the same person can win twice. Drawing without replacement removes them, which is what you want for multiple prizes or for assigning people to slots. For three winners out of forty entries, without replacement each person has a 3 in 40 chance, or 7.5 percent; with replacement the arithmetic changes and some names may repeat.

Decide this before the draw, not when a repeat appears. The same applies to reserves: if you may need substitutes, draw the full ordered list in one go and keep positions four, five and six as the reserve order, rather than running a fresh draw later with a smaller pool.

When a browser random number is not enough

For a classroom draw, a team split or a giveaway among friends, a standard generator is entirely adequate — the distribution is uniform enough that no one is disadvantaged. What it is not is unpredictable to someone who can inspect the page: the usual browser function is explicitly documented as not cryptographically secure, and its output can in principle be reconstructed.

When money, places or legal obligations are involved, that matters. Public randomness beacons such as the one operated by NIST publish a signed random value at fixed intervals, so a draw can commit to a future beacon pulse in advance and anyone can verify afterwards that the result followed from a value nobody could have known when the rules were set.

Frequently asked questions

Is drawing from a hat fairer than using a website?

Not statistically. Slips of paper are rarely mixed enough, and the last ones added tend to sit on top, so a hat is usually less uniform than a software draw. Its real advantage is that everyone can watch it — which you can match by recording a digital draw.

Can I redraw if the winner does not reply?

Only if you said so beforehand. Publish a deadline — for example seven days to claim — as part of the rules, and draw the reserve order at the same time as the winner. Improvising a second draw after the fact is what makes participants suspect the first one.

How do I weight a draw so some entries count more?

Repeat a name as many times as it has entries, then draw once from the expanded list. Someone with three tickets appears three times among a hundred lines and has a 3 percent chance. Say clearly that the draw is weighted, since a weighted draw is not what most people assume by default.