Randomize a list and pick winners fairly

It favours nothing: every one of the possible orders is equally likely. The list is shuffled with Fisher-Yates, which walks from the last position to the first and swaps each one with a slot picked from the entries not yet fixed, and the picking uses crypto.getRandomValues rather than the ordinary random function. There is one more subtlety most implementations skip: taking a raw 32-bit number modulo the list length quietly favours the lower positions, because 4,294,967,296 does not divide evenly by 7 or by 1,000. Draws that land in the ragged tail above the last whole multiple are thrown away and taken again, which removes that lean completely and costs less than one wasted draw in 40,000 on a list of 100,000 names.

Not from one draw — a shuffle moves each entry to exactly one position, so nobody can occupy two. If a name is sitting in your list several times it holds several tickets and can win more than one prize, which is usually what you want for a raffle where people earned extra entries and never what you want for a class register. The fold-repeats checkbox collapses identical lines into one ticket and ignores capitals while doing it, so Ana Marino and ana marino are treated as the same person.

Repeat their line once for every extra chance and leave the fold-repeats box unticked. Somebody who earned three entries appears on three lines and holds three of the tickets in the hat; with 120 lines in total their odds are 3 in 120. This is the honest way to weight a draw, because the arithmetic stays visible in the list itself — anyone can count the lines and check. It also means the pool grows quickly, so keep the paste under a few hundred thousand lines and the draw still lands instantly.

Because sorting with a coin-flip comparison does not produce equal odds, and the result depends on which sorting algorithm your browser happens to use. The comparison is asked inconsistent questions — it may say A beats B, B beats C and C beats A — and a sort given contradictory answers takes shortcuts. On a three-item list the popular one-liner leaves some orders turning up twice as often as others, and the skew grows with the list. It is the single most common way an online shuffler gets this wrong, and it is invisible until somebody counts thousands of runs.

The randomness is: it comes from the same generator a browser uses to make encryption keys. What a promotion usually needs on top of that is evidence, which is why the draw record exists — it names the winners and reserves, the size of the pool, when the draw happened and how the order was produced. Screen-record the draw if the prize is worth arguing about, and keep the entrant list as it stood at closing time. Prize-promotion rules in the UK, the EU and several US states also require a stated closing date and a published method, and no shuffler can supply those for you.

No — reloading clears both the list and the draw, and nothing is remembered between visits. That is deliberate: a draw you can silently re-run until you like the winner is not a draw. Copy the result or save the record before you close the tab. Changing the entries after a draw does not secretly re-run it either; the panel simply warns you that the result on screen belongs to the older list.

Anything that fits on a line: quiz questions before a round, the order teams present in, chore assignments, flashcards, a playlist you want out of album order, the sequence of interview prompts. Teachers use it to pull a name for the next question without falling into the habit of asking the same three people. The tool never looks at what a line contains, so a line can be a sentence, a URL or an emoji.