Split the Teams Before the Argument Starts

Two minutes before kickoff someone types "6 vs 10?" into the chat. Two people have already pasted two different line-ups. A third one asks whether anyone remembered John. Nobody is arguing about football at this point. They're arguing about which version of the split is the real one, and who counted heads last. Meanwhile everyone is standing on the pitch you're paying for by the hour.

If you organize a regular pickup game, you know the shape of this. Making the split takes thirty seconds. Getting twelve people to agree on it takes forty minutes.

The four ways teams actually get picked

Almost every regular game already uses one of these, usually without anyone ever naming it out loud. All four are honest. All four break in a specific, predictable place.

Draw from a hat, or count off one-two-one-two

The oldest method there is: everyone numbers off, odds against evens. Fast, visibly neutral, and nobody can accuse you of stacking a side. The failure mode arrives around minute twenty, when it's 5-0 and the losing side is playing out the string. "It was just the draw" is completely true and completely useless as consolation. The other failure is quieter: a hat draw made before two people cancel and one guest arrives is no longer the draw everybody agreed to, and re-drawing feels like cheating even when it isn't.

Captains alternate picks

Two captains take turns: "I'll take Mike," "then Chris," back and forth to the end of the list. This is as close to a skill-aware split as an amateur game gets without spreadsheets, because the captains actually know who's in form this week. It's also the slowest method, it stalls completely when someone walks in mid-draft, and being picked last in front of a group chat is the kind of thing people remember for a whole season. Adults absolutely notice.

By sign-up order

First eight who signed up are one side, the rest are the other. Mechanical, instant, nothing to dispute: the list already exists, you just cut it in half. Its relationship to on-pitch balance is zero, so roughly every third week you get an accidentally stacked side. It also quietly rewards whoever checks their phone fastest, which over a season is its own kind of unfair.

One person decides and posts it

The fastest option in the moment, and exactly the scene from the top of this page. One person does the sorting from memory under time pressure while nobody checks the arithmetic. And when a second person helpfully does the same thing independently, you get two competing line-ups in the same thread ninety seconds apart, and now the group has to work out which paste is authoritative before anyone can work out which shirt to wear.

The argument is almost never about the algorithm

Look at what those four break on and it's the same three things every time.

The roster wasn't settled when you split. You divided fourteen people; twelve came. Two late "+1"s arrived after the teams were posted. Any method applied to a stale list produces a stale split, and that is what feels unfair, not the method. If sign-ups themselves are the leaky part for you, that's a separate problem worth fixing first: running sign-ups in a group chat when there are more than ten players.

Nobody owns the current version. The split existed in someone's head, then in one message, then in a competing message. Ten minutes later the honest answer to "which team am I on?" is "scroll up and hope."

One correction means redoing everything. Max shows up late. Under a hat draw or a snake draft there's no defined way to add one person, so people re-draw the whole thing, which invalidates the version everyone already accepted, which restarts the argument from zero.

Five rules that shorten it, with or without any software

1. Pick the method before the day, not on the day. Whatever you choose (hat, captains, sign-up order), announce it once in the chat and keep it. Arguments about which method are what actually eat the forty minutes; the method itself is rarely the issue.
2. Set a cut-off and split after it. "Teams get posted at 8pm the night before, whoever signs up after that gets slotted in." Late arrivals become an insertion, not a re-draw.
3. One person posts the split. Not two helpful people. One. Everything else is a suggestion sent to that person.
4. Insert, don't re-roll. Latecomer goes to the shorter side; someone drops, one name moves across. Two seconds, and the version everyone already accepted survives.
5. If you use captains, let them pick privately. Same teams, none of the "picked last in front of everyone" damage. Post the finished line-up, not the draft order.

None of that needs a bot. It's just what experienced organizers converged on, and it's worth stealing regardless of what you use to keep the list.

What this looks like when the chat holds the split

TeamUp doesn't invent a fifth method. It runs two of the ones above and then holds the result so it never has to be re-announced from memory.

Ask it to split the group and you get exactly two strategies. Random is the hat draw, done in one message: it shuffles the confirmed roster and deals people out round-robin, so the sides come out even; with an odd count they differ by one person, never by four. Say how many teams you want and it makes that many; two is the default. Manual is you or your captains dictating: name a team, list the players by their badge numbers, done. The bot keeps the assignment, you keep the judgment.

A demo: in a real chat the assistant's reply is written by the model in the chat's own language, so the wording will differ.

If the line-up already arrived as a finished paste grouped under team names ("Yellow: Alex, Mike, Emma… Purple: Dan, Chris, Max…"), hand it over as it is. It registers everyone and assigns the named teams in a single pass, including guests nobody has ever entered before. That's the competing-pastes situation from the intro, resolved by making one of the pastes the official record instead of arguing about timestamps.

After that, corrections are corrections. Moving Dan to the other side is its own small action by badge number, separate from the split, so a latecomer gets inserted instead of triggering a re-draw. Marking someone captain or goalkeeper is a role on an existing roster, not something baked into how teams were formed. And because a random re-split replaces existing teams outright rather than nudging them, asking for one when teams already exist gets you a check first: the bot shows the current sides and waits for a yes, instead of silently erasing the hand-fixes you already made.

Each player has a separate entry on the current roster. The number helps identify the person you mean, but can change after a cancellation or a complete roster replacement. Request the latest list and check the name before acting by number. get started

What it deliberately won't do

There is no skill-based split. Choose random distribution or manual teams. A player card contains registration history and personal tallies, but the splitting algorithm does not use it. If captains judge form and choose players, record their decision and adjust individual assignments afterwards. Registration history is neither verified attendance nor an ability rating.

Random doesn't secretly even things out. It's a hat, not a matchmaker. Equal head counts, nothing more. If last week's 5-0 still stings, the fix is switching to captains or manual for a week, not hoping the shuffle compensates.

A re-split is a replacement, not a tweak. That's precisely why it asks before overwriting, and why one-name fixes are their own action.

The forty minutes were never about fairness

They were about six people holding six versions of the truth at once. Any of the four methods above is fair enough for a Tuesday game. What decides whether you kick off on time is whether the split is settled before people arrive, whether one place holds it, and whether a latecomer costs you one move or a whole new draw. Fix those three and the method almost stops mattering. If you've spent a season refereeing line-up arguments, you know that's the actual win.