👥Split a list into even teams, without the playground negotiation.

✓ Tested & verified Updated: Aug 6, 2026 How the randomness works

Random team generator

12 names ready — press the button and the sides are dealt fairly.

Shuffled with Fisher–Yates, dealt round-robin — team sizes differ by one at most

Names — one per line

Number of teams

Try also

How do you split a group into random teams fairly?

Shuffle the whole list first, then deal it out one name at a time like cards. That is precisely what happens here: a Fisher–Yates shuffle randomises the order, and the shuffled names are dealt round-robin into however many teams you asked for. Because the deal goes one per team per pass, the teams can never differ in size by more than a single person — twelve into two is 6 and 6, eleven into three is 4, 4 and 3.

The shuffle is the part that matters. Fisher–Yates, published in 1938 and given its modern in-place form by Durstenfeld in 1964, makes every possible ordering equally likely — a stronger guarantee than “looks mixed up”. Every name has the same chance of landing in every team, whatever order you pasted them in.

What it deliberately does not do is balance ability, seniority or who has a car. It is a random split, not a fair-fight algorithm — the difference is the subject of the note further down.

Round-robin dealing means the split is decided by arithmetic, not by luck. These are the sizes you get for the awkward headcounts.

People	2 teams	3 teams	4 teams
10	5, 5	4, 3, 3	3, 3, 2, 2
11	6, 5	4, 4, 3	3, 3, 3, 2
12	6, 6	4, 4, 4	3, 3, 3, 3
13	7, 6	5, 4, 4	4, 3, 3, 3
15	8, 7	5, 5, 5	4, 4, 4, 3

Which team gets the extra person is itself random, because the deal order follows the shuffle. That matters more than it sounds in a PE lesson, where being on the seven rather than the eight is the sort of thing that gets noticed.

Need a queue instead of a split? Use the list randomizer. Need one person rather than a group — a captain, a volunteer, a scapegoat? The random name picker does that in a single spin.

Frequently asked questions

How does it keep the teams even?

By dealing rather than dividing. After the shuffle, names go out one per team in rotation, so with 13 people and 4 teams you get 4, 3, 3 and 3 — never 6, 3, 3 and 1. The maximum difference between the largest and smallest team is always one person, whatever the numbers, and the order of the deal is itself randomised by the shuffle.

Can I make teams of a fixed size instead?

Set the number of teams to your headcount divided by the size you want — 20 people in teams of 5 means 4 teams. The generator works from the team count because that is the number people actually decide on first, and because a fixed size only divides cleanly when the numbers cooperate, which they rarely do.

Does it balance skill levels?

No, and no random generator can. It has no information about anyone beyond their name. A genuine random split will sometimes put the three strongest players together, which is exactly what randomness means. If the sides need to be competitive rather than merely impartial, generate a split and then let both teams agree on one swap.

Can I run it again if the split looks bad?

You can, but decide the rule first. Rerolling until the teams look right is not a random split, it is a selection process with extra steps — and everyone in the room can see it happening. The version that keeps the peace is one generation, accepted, with an agreed number of mutually approved swaps afterwards.

How many people can it handle?

Any number you can paste. It stays readable well into the hundreds, which covers a school year group or a conference breakout. The practical limit is your ability to read the output on screen — for very large groups, generate the split and screenshot it rather than trying to read names off a phone in a sports hall.

Can I share the teams with everyone?

Copy the page link. Your entries are carried in the address bar, so the person you send it to gets the same list — though pressing generate on their machine produces a fresh shuffle, since the split is not stored anywhere. To share the actual result, screenshot it or paste the teams into your group chat.

How the randomness works

The list of names is shuffled with the Fisher–Yates algorithm — walking backwards through the array and swapping each element with one chosen at or before its position — using indices drawn from crypto.getRandomValues with rejection sampling, so every one of the possible orderings is equally likely. The shuffled list is then dealt round-robin into the requested number of teams, which guarantees sizes differ by at most one and gives every name an equal chance of landing in any team. Names are read from the text box, mirrored into the page URL for sharing, and never sent to a server.

Sources:

Random is not the same as balanced

A random split solves a social problem, not a sporting one. Captains picking sides is a fair procedure that produces a genuinely unpleasant experience for whoever is chosen last, and every adult who was ever picked last remembers it. A generator removes the ranking entirely — nobody is chosen after anyone else, because nobody is chosen at all.

Where it needs help is competitiveness. Random splits are unbiased, which is not the same as balanced: with twelve players of varying ability, a genuinely random draw will regularly hand one side the better half. Two conventions handle it well. Either run it once and accept the game you get, or run it once and allow each team a single swap they both agree to. Both are defensible; rerolling until it looks right is not, and everyone knows it.

For a running order rather than a split, the list randomizer shuffles the same names into a queue.