๐ŸคTwo by two, and nobody left standing.

โœ“ Tested & verified Updated: How the randomness works

Pair generator

  • 7 names ready โ€” press the button and they go two by two.

Shuffled with Fisherโ€“Yates, then taken in twos โ€” every possible pairing is equally likely

Game night, drawn

Weekly

One game, one way to decide who starts, and one rule to change โ€” drawn fresh each week. For the dice, cards and coin in the drawer.

Free, no account. One click unsubscribes and deletes your address.

Try also

How do you split a group into random pairs?

Shuffle the whole list once, then take it two at a time. That single pass is enough: if every ordering is equally likely, so is every possible set of pairs. The naive method โ€” pick two names, then two more from what is left โ€” arrives at the same place with more steps and more ways to go subtly wrong.

With an odd number somebody is left over. By default the last pair becomes a three, because a three works for almost every activity that works in twos โ€” and being the person left out is what everyone remembers about being paired up at school. Turn the option off and the leftover is shown explicitly as sitting out, not dropped off the end of the list.

Because the list is reshuffled each time, the same two people can be paired twice in a row: random is not a rota. If everyone must meet everyone exactly once, that is a schedule โ€” the round robin generator builds it.

How many different pairings there are

PeoplePairsPossible pairings
423
6315
84105
105945
12610,395
1682,027,025

The number climbs fast, which is why repeats feel surprising in a small group and never happen in a large one. With six people, one pairing in fifteen is the one you just had.

Where pairs beat teams

SituationWhy twos
Language practiceEveryone speaks half the time; in a four, one person carries it
Code reviewTwo people can hold one problem in their heads
Secret SantaOne giver, one receiver โ€” a chain, not a group
Interview practiceOne asks, one answers, then swap
Chess or cardsThe game itself needs exactly two

To split into more than two groups, the team generator deals a list into any number of sides. For gift-giving specifically, the secret santa generator draws a single closed circle so nobody gets themselves. To put a list in a random order without grouping it at all, use the list randomizer.

Frequently asked questions

What happens with an odd number of people?

One group of three by default, made from the last pair plus the leftover. Switch the option off and the extra person is listed on their own as sitting out. There is no third possibility โ€” with an odd list, someone is in a three or someone is out, and it is better to choose than to let the tool decide silently.

Can the same pair come up twice in a row?

Yes, and it will, roughly as often as chance says. Each press is an independent shuffle with no memory of the last one. With eight people there are 105 possible sets of pairs, so a repeat within a few presses is unremarkable rather than a sign of a broken tool.

How do I stop two particular people being paired?

You cannot, and adding that rule would quietly break the fairness โ€” every constraint changes the odds for everyone else, not just the pair you named. The practical answer is to press again, which costs a second, or to split the list into two groups and pair each separately.

Is it any good for large classes?

Up to 200 names, so a lecture hall is fine. Past about thirty pairs the on-screen list gets long to read out, and it is usually easier to project it than to call it. Numbered groups help โ€” each pair is numbered so you can say โ€œgroup seven, that tableโ€.

Does it remember previous rounds?

No. Nothing is stored, so every press starts fresh. If you want partners to rotate without repeats across a term, keep the list in a spreadsheet and use a round robin schedule instead โ€” that is exactly the problem it solves.

Can I share a set of pairs with someone?

The names live in the address bar, so the link reproduces your list โ€” but not the draw, which happens fresh in whoever’s browser opens it. To share the result rather than the setup, screenshot it. Nothing about a random draw survives being repeated.

How the randomness works

The list is read one name per line, shuffled once with Fisherโ€“Yates โ€” every swap index taken from crypto.getRandomValues with rejection sampling โ€” and then read off in twos. Because a uniformly random permutation is used, every possible pairing of the group is equally likely; for n people there are n! / (2^(n/2) ยท (n/2)!) distinct pairings and each has the same chance. With an odd count the final leftover is appended to the last pair, making one group of three, or listed separately if that option is switched off. Nothing is stored between draws.

Sources:

The problem with letting people choose

Letting people pick their own partners looks harmless and is the least fair method in the room. It sorts by confidence and existing friendship, produces the same pairs every week, and reliably leaves the same one or two people standing. Every teacher knows the shape of it; most of us remember being on one end of it.

A shuffle removes the social cost entirely, because there is nobody to be disappointed in. It also produces the mixing that pairing is usually for: new pairs mean new conversations, and the quiet student ends up next to somebody they would never have chosen, which is the entire pedagogical argument for doing it at all.