How BeTeam Crowd numbers are calculated
Every percentage BeTeam publishes comes from predictions real people submitted before a match kicked off. This page defines exactly what is counted, what is excluded, and where the numbers are weak — in enough detail that you could reproduce them if you had the data.
What counts as a prediction
An exact-score forecast submitted by a BeTeam player for a specific match, before that match kicked off. BeTeam’s prediction endpoint refuses writes from kickoff onward, so every published forecast was genuinely made in advance. There is no way to add or edit a prediction once a match has started.
One person, one prediction
A BeTeam player can be in several private leagues that all cover the same match. We count that person once. Where a player predicted the same fixture differently in different leagues, we take their most recently updated forecast. This is why our sample sizes are smaller than our raw prediction counts, and it is deliberate: “1,000 predictions” should mean 1,000 people.
What is excluded
- Auto-filled predictions. When a player misses a deadline, BeTeam writes a placeholder so their league still scores. Those are not forecasts and never appear in these numbers.
- Internal accounts. BeTeam’s own test accounts are excluded from every public figure.
- Suspended accounts. Excluded.
- Manually created matches. Players can add fixtures that aren’t in our data provider’s feed, for their own league. Those stay private and are never published.
Minimum sample
A match is only published once at least 25 people have predicted it. Below that, a percentage says more about who happened to be online than about what fans think. The sample size is printed next to every figure so you can weigh it yourself. Raising this threshold later affects which new matches get published; it never empties a page that already cleared it.
Frozen at kickoff
The moment a match kicks off, we freeze that match’s crowd split and keep it. The page you read a year later shows what the crowd expected before the match, not a figure recalculated afterwards.
Known limitations
- These are shares, not probabilities. “62% expect a home win” means 62% of the people in that sample predicted one. It is not a modelled likelihood, and we do not present it as one.
- The crowd is self-selected. BeTeam players skew towards the competitions their friend groups follow. This is not a representative sample of all football fans.
- The crowd under-predicts draws. Our study of 194,812 World Cup predictions found fans pick draws at roughly half the rate they actually occur, while calling the correct outcome 66% of the time overall.
- Rounded percentages may not total 100. Three shares rounded independently can sum to 99.9 or 100.1. We do not adjust a number to tidy up the arithmetic.
The AI prediction
Some match pages carry a separate, clearly-marked forecast from a language model, generated shortly before kickoff. It is there to be compared against the crowd, not confused with it: it is not counted in any percentage on the page, it does not affect the sample size, and the model that produced it is named so the claim is checkable. It is the only text on these pages a model wrote.
Languages
A match page is published in English and, where the competition’s own audience reads another language, in that language too. Both versions are built from the same numbers by the same code — a translation never changes a figure.
Privacy
Published data is aggregate only. No individual prediction, player, or private league is ever exposed, and these pages contain nothing that could identify a BeTeam user.
Not betting advice
BeTeam is a free social prediction game with no real money in it: no deposits, no odds, no stakes, no cash prizes. These pages report what our players predicted. They are not tips, and BeTeam is not a sportsbook.