Splitting even teams: how to pick fair sides
One-sided games are the quietest reason weekly groups fall apart. Nobody quits over a single 9-2, but after a month of them the losing half simply stops turning up. Here is what does not work, and what does.
The four common methods, and why they fail
| Method | What actually happens |
|---|---|
| Two captains pick | Reasonably accurate, but slow, and it humiliates whoever goes last every single week. It is the most common reason a new player does not come back. |
| Random | Feels fair, is not. Splitting ten players at random produces a close game about a third of the time. |
| Self-declared skill level | Almost everybody rates themselves as average. The data is worthless within a fortnight. |
| "Same teams as last week" | Works until somebody does not show, at which point you are back to square one with everyone already on the pitch. |
What they have in common is that none of them use the information you already have: how people actually played the previous times.
What "even" actually means
Even does not mean each side gets an identical amount of talent. It means each side arrives at a similar chance of winning. Two things routinely get forgotten:
- A goalkeeper is worth more than any outfield player. A split that ignores that will produce lopsided games even when the averages match perfectly.
- Spread matters as much as the average. A team with two excellent players and three weak ones is not equal to five average ones, even when the mean is identical.
Why the ratings are blind
The only way to know how good somebody is, is to ask the people who played with them. The problem is that visible ratings turn into politics immediately: people rate their friends highly, and anyone marked down knows exactly who did it. So in KROO the rating is blind.
After the game, not before it
You rate the people you played with and against while it is fresh, rather than guessing in advance.
Nobody sees who rated them
The rater is never shown on any screen. Not to them, and not to you.
One bad week does not stick
The score is a weighted average across games rather than the latest number. The more games there are, the more it settles, and the less any single odd night moves it.
Why it is a Bayesian average
A new player with two ratings does not have enough data behind them to trust that average. Rather than treating those two ratings as the truth, KROO pulls the score toward the group average and releases that pull as games accumulate. The effect is that a newcomer does not get an extreme rating off the back of one unusually good or bad evening.
How Kroo Balance actually splits the sides
Close the list
The split runs on who confirmed, not on who might come.
Sort by score and deal in a snake order
First player to side A, second and third to side B, fourth and fifth to side A, and so on. A snake draft balances both the average and the spread, which is exactly what a random split misses.
Place the goalkeepers first
So that one side does not end up with both of them.
You can still drag
If you know something the data does not, an injury, somebody back after a month out, drag a player across. The algorithm is a proposal, not a verdict.
Common questions
Is there an app that splits football teams evenly?
Yes, that is what Kroo Balance does inside KROO. Once the confirmed list closes, it splits the players into two sides using the blind ratings their teammates gave them, and places goalkeepers separately. If you disagree with the result you can drag a player across.
Why is a random split not good enough?
Because it throws away everything you already know. Splitting ten players at random gives a close game about a third of the time, and the rest are decided inside ten minutes. It feels fair because nobody chose, but the outcome usually is not balanced.
Who can see the rating I gave a player?
Nobody. The rater is never shown on any screen in the app, including to you. That is the only reason the ratings are worth anything: the moment people know who rated them, ratings become politics rather than information.
What happens to a new player with no ratings yet?
They start near the group average and their score moves gradually as games accumulate. That prevents one unusually good or bad evening from deciding how they are treated permanently.
