Random Group Generator

Paste a list of names and split them into random groups — by how many groups you want, or by how many people each should hold. It uses a Fisher-Yates shuffle with cryptographic randomness, which matters more than it sounds: the one-line trick most tools use, sorting with a random comparator, is not a shuffle. It breaks the contract sorting relies on, and in practice leaves people near where they started, so the same names keep landing together. Groups are dealt round-robin rather than sliced, so eleven people into four groups gives 3, 3, 3, 2 rather than 4, 4, 3 and an empty group.

Dealing against slicing

11 people, 4 groupsResult
Dealt round-robin3, 3, 3, 2
Sliced into chunks4, 4, 3, 0 — an empty group
Fisher-YatesEvery arrangement equally likely
Random comparator sortUneven; people stay near where they were

When to use it

  1. Split a class into project teams.
  2. Draw teams for a game.
  3. Assign a secret santa or a buddy system.
  4. Divide a workshop into breakout groups.
  5. Randomly allocate shifts or tasks.

How to Use

  1. Paste your names, one per line.
  2. Choose a number of groups, or a group size.
  3. Generate — and again for a different split.

Features

  1. Fisher-Yates shuffle, not the sort trick
  2. Cryptographic randomness
  3. Split by group count or group size
  4. Round-robin dealing, so no empty groups
  5. Blank lines ignored
  6. Copy or download the result
  7. Nothing transmitted

FAQs

Is it actually random?

Yes. It uses Fisher-Yates with indices drawn from the browser's cryptographic generator by rejection sampling. The common shortcut — sorting with a comparator that returns a random sign — is not a shuffle: it gives an inconsistent answer to the same comparison, which no sorting algorithm can cope with, and the resulting distribution is badly uneven in every engine.

Why do the same people keep getting grouped together in other tools?

That is the symptom of the sort trick. Because sorting algorithms do limited work when things already look ordered, elements tend to stay near where they started — so a list in the same order produces similar groups. A real shuffle has no such tendency.

How are uneven numbers handled?

By dealing round-robin, like cards. Eleven people into four groups gives 3, 3, 3, 2. Slicing the list into equal chunks instead would give 4, 4, 3 and an empty fourth group, which is a bug rather than a rounding choice.

Can I split by size instead of by count?

Yes. Ask for groups of four and the number of groups is worked out — eleven people needs three groups, not two and a leftover.

Can I re-roll?

Generate again for a completely fresh shuffle. Nothing is remembered between runs, so nobody is nudged toward or away from anyone.

Are the names sent anywhere?

No. They stay in the page, which matters when the list is a class register or a staff rota.

Technical Details

The shuffle is Fisher-Yates, and the reason that matters is that the one-liner almost everyone reaches for is not a shuffle at all. `array.sort(() => Math.random() - 0.5)` hands the sort a comparator that answers inconsistently, which breaks the contract every sorting algorithm relies on — so the result depends entirely on which algorithm the engine happens to use, and is badly non-uniform in all of them. In V8 some orderings come up orders of magnitude more often than others and elements tend to stay near where they started, which for a group generator means the same people keep ending up together. It looks random enough to pass a glance, which is exactly why it survives in so much code. Fisher-Yates walks backwards swapping each position with a randomly chosen one at or below it, and the inclusivity there is the detail that makes every permutation equally likely: drawing strictly below instead produces a Sattolo shuffle, which never leaves any element in its original position and is therefore not uniform either. The indices come from the same rejection-sampled cryptographic source the rest of these tools use rather than a modulo, which would reintroduce a small bias at every swap. Allocation deals round-robin rather than slicing the shuffled list into chunks. Slicing looks equivalent and is not: eleven people into four groups gives 4, 4, 3 and an empty fourth group, because the first three slices consume everyone. Dealing gives 3, 3, 3, 2, which is the even split people expect and cannot produce an empty group. Splitting by group size rather than group count rounds the resulting count up for the same reason — eleven people in groups of four needs three groups, not two and a leftover. Nothing is transmitted; the names stay in the page.

Random & Fun Tools

Random & Fun Tools

  • Random Picker
  • Random Picker Wheel
  • Pictionary Word Generator
  • 2048 Game
  • Dice Roller
  • Coin Flip