Random Team Generator

Random team generator: split a roster into evenly sized groups

Paste a list of names, one per line, say how many teams you need, and press Generate. The whole roster is shuffled and dealt out, so the teams come back the same size or within one person of each other, and nobody was picked first or last in front of everyone else. That last part is the actual reason this exists: captains picking sides in front of the group is a small public ranking, and a shuffle removes it entirely.

It runs in your browser. The roster is never uploaded, never stored on a server and never leaves the page — which matters more than it sounds when the list is a class register or a list of employees. This tool partitions a whole list into groups; if you only need to draw one person out of a list, the random name picker is the one for that.

Choosing team count versus team size

The input here is number of teams, from 2 up to 50. There is no field for team size, and that is worth knowing before you start, because most people arrive with a size in mind: "groups of four." Converting is one division — divide the roster by the size you want and round up.

Rounding up rather than down keeps groups at or below your target size. Rounding down keeps them at or above it. Which you want depends on the activity: a workshop exercise with four worksheets wants a hard ceiling of four, while a sports drill that needs at least three players wants a floor. That last row shows why the choice matters — for pairings, an odd roster is better handled as one group of three than as one person on their own.

When the roster does not divide evenly

This is the case that breaks naive splitters, and the tool handles it by dealing rather than slicing. The roster is shuffled once, then names are dealt around the teams in rotation, the way you would deal a deck of cards: first name to team 1, second to team 2, and around again. The consequence is that team sizes always differ by at most one, and the remainder is spread out instead of dumped into a final undersized team.

With 23 people across 5 teams you get three teams of 5 and two of 4 — never four teams of 5 and one lonely team of 3. Which teams get the extra person is determined by the shuffle, so it is not always the first ones on screen. Each team card shows its own headcount, so you can see the split at a glance rather than counting names.

If an uneven split is genuinely unacceptable — a tournament bracket, a drill that requires exact numbers — you have two options: trim the roster to a multiple of the team count and give the extras a role off the teams (referee, scorer, timekeeper, floater), or add the placeholder names yourself so the split comes out even. Typing "Sub" or "Bye" as a line in the roster is a legitimate trick; the generator treats it as any other entry.

What "balanced" does and does not mean here

Be clear about this before you use the output for anything competitive. The teams are balanced in size. They are not balanced by skill, seniority, department, gender, ability, or anything else, because the tool has no information about your people beyond their names, and it does not ask for any. Every name is equally likely to land on every team, which is exactly what fairness means for a lottery and is not at all what it means for a competitive match.

So a purely random split can absolutely hand one team all three of your strongest players. If the activity is competitive, the usual fix is to stratify by hand: split the roster into tiers yourself, then run the generator once per tier and merge the results. For 16 players into 4 teams, take the top 4, generate 4 teams from them so each team gets one, then repeat with the next 4, and so on. That takes four passes and gives you a genuinely level set of teams, with the randomness deciding who goes where inside each tier — which is still enough to defuse any argument about favouritism.

The same trick handles constraints. There is no keep-these-two-apart or keep-these-two-together option, so if two people must not be on the same team, place them manually first and generate for everyone else, or generate and re-roll until the split works. On small rosters a re-roll takes seconds.

Keeping it fair across repeated rounds

The honest answer first: each generation is independent. The tool keeps no memory of previous splits, so nothing stops the same two people ending up together three sessions running. It is not a rotation scheduler and it will not tell you who has already worked with whom.

It helps to know how likely a repeat actually is. For a roster of N people split into teams of size k, the chance that any one specific other person lands on your team is (k−1)/(N−1). Twelve people in three teams of four gives 3/11 — a bit over one in four, per pair, per round. Across a term of weekly sessions, some pair repeating is not a coincidence, it is arithmetic. People notice repeats and do not notice the far larger number of non-repeats, which is why random splits get accused of being rigged.

Practical ways to spread the mixing when it matters:

• Use more teams. Smaller groups mean a smaller k, and repeat pairing falls straight out of the formula. Six pairs mix far more than three groups of four.
• Anchor and rotate. Fix one person per team by hand each round — a rotating leader — and randomise only the remainder. This guarantees the anchors never repeat with each other.
• Round-robin the deliberate part. For a team that must genuinely see everyone, a fixed rotation schedule beats randomness outright. Randomness gives you unpredictability, not coverage.
• Re-roll in front of the group. If a split looks lopsided, press Generate again with everyone watching. A visible re-roll costs nothing and answers the fairness objection better than any explanation.
• Keep your own log. The tool does not save results, so if repeat-avoidance matters, screenshot or copy each round's teams. Nothing is stored between generations or between visits.

Where people use it

• PE and school classes. The original problem case. Sides get picked without anyone being chosen last, and the split is visibly not the teacher's opinion of anyone. For mixed-ability activities, the tier-by-tier method above keeps games from becoming a rout.
• Workshops and training. Breakout groups for an exercise, and a fast way to break up the cliques that form when a room self-organises. Paste the attendee list, pick the number of breakout rooms, project the result.
• Standups and engineering teams. Pairing rotations, code-review buddies, on-call shadowing, and splitting a team for a spike. For pairs, set the team count to half the headcount — an odd number gives one trio, which is usually the right outcome anyway.
• Tournaments and game nights. Seeding groups for a bracket, splitting a pub quiz, dividing a games night. Remember the size-only caveat: for a serious bracket, seed by hand.
• Icebreakers and offsites. Cross-department mixing, where the whole point is that the groups are not the ones people would have chosen.
• Family and household use. Chore teams, holiday gift groups, deciding who is on washing-up.

Getting the roster in cleanly

Entries are one per line. Blank lines are ignored, so a list pasted out of a spreadsheet with gaps in it works without cleaning. Leading and trailing spaces are stripped. The live count under the box tells you how many people the tool actually sees, which is the fastest way to catch a paste that went wrong — if you expected 28 and it says 27, a name is on the same line as another one.

Duplicate names are kept as separate entries, so two people called Sam are two people, not one. That means it is on you to disambiguate them — "Sam K" and "Sam T" — or you will not know which team each is on. If you paste a column that includes a header row, delete the header, or you will end up with a team member called Name.

Randomness, privacy and the limits

The shuffle is a Fisher–Yates shuffle, which is the correct algorithm for the job: every possible ordering of your roster is equally likely, and it does not suffer from the subtle bias of the sort-by-random-key approach a lot of quick scripts use. It draws from the browser's built-in random number generator. That is uniform and entirely appropriate for picking teams, but it is not a cryptographically secure source, so do not use this for a prize draw with real money on it, a legal ballot, or anything where somebody has a motive to predict the outcome.

On privacy: the names stay in the page. There is no upload, no account, no analytics on the content of your list, and nothing written to a server. Closing or reloading the tab clears everything, including the teams you just generated — there is no saved history and no export button, so capture the result on screen before you navigate away.

Two other things the page does that are easy to miss. There is a mode switch at the top: Make Teams splits the roster, and Pick One (Wheel) spins the same roster as a wheel to draw a single name — handy when you have already made teams and now need to choose who presents. And the team count is clamped to the 2–50 range, so typing a larger number quietly caps it rather than producing fifty-one one-person teams.