RandomPick

Split into teams

A number of groups, or groups of a size. Shuffled properly, and dealt so the remainder spreads out instead of landing on one group.

Saved in the address bar — bookmark or share this page and the list comes with it.

How the picking works. Every choice comes from crypto.getRandomValues, the browser's cryptographic random source, mapped to your list length using rejection sampling so no entry is even slightly favoured by the arithmetic. It is cryptographic-quality pseudorandomness — not "true" randomness harvested from physical noise, which is a different and stronger claim we are not making. Nothing about your list is sent anywhere, and nothing is weighted unless you say so.

Shuffled, then dealt

The split works the way you would deal cards: the list is shuffled with a Fisher–Yates shuffle using the cryptographic random source, then dealt one at a time around the groups. The consequence is that when the numbers do not divide evenly, the remainder is spread — nineteen people into four groups gives 5, 5, 5, 4, not 5, 5, 5 and a lonely 4 built last.

That is a small thing that matters in a classroom, where a visibly short group reads as a punishment.

The two modes are not interchangeable

Choosing "four groups" fixes the number of groups and lets the sizes fall out. Choosing "groups of four" fixes the size and lets the number of groups fall out — including a possible undersized final group. If you have four tables, use the first. If you have an activity that needs exactly four people, use the second and deal with the remainder deliberately.

Running the session as well as splitting it? The countdown and interval timers have a full-screen mode built for exactly this — visible from the back of a room, and the alarm sounds even if the tab is buried.

Questions

Groups of a size, or a number of groups?
Both, and the distinction matters more than people expect. 'Four groups' from eighteen people gives you groups of 5, 5, 4, 4. 'Groups of four' from eighteen gives you four groups of four and one of two — which may be exactly what you want, or may leave a pair stranded. Pick the mode that matches the constraint you actually have.
Is the split even?
As even as the numbers allow. The list is shuffled first, then dealt round-robin into the groups, which distributes any remainder one person at a time rather than dumping it all on the last group. That is why group sizes differ by at most one.
Can I stop two people ending up together?
No, and that is a deliberate limit rather than an oversight. Constraint-based grouping — keep these apart, keep those together — is a genuinely different tool with a real interface, and a half-built version that silently ignores your constraints would be worse than not having it. Split, and swap two people by hand if you need to.
Does it reuse the list from the wheel?
Yes. Anything you pasted on the wheel page is already here, and edits here show up there. One list, several tools.