Prediction
A match prediction is a numerical estimate of what may happen: the probabilities of a home win, a draw and an away win (1X2), the most likely score, the chances of both teams scoring or of more than 2.5 goals. Unlike a tip, it doesn't pick a side: it measures uncertainty.
A prediction rests on a model, meaning an explicit method that turns data (expected goals, form, league position, squads…) into probabilities. The typical output is a 1X2 distribution, for example 46% / 26% / 28%, complemented by derived markets: exact score, both teams to score (BTTS), over or under X goals.
The difference from a tip is fundamental. A tip picks a side; a prediction says how uncertain the outcome is. A team at 46% is still the favourite, yet it doesn't win one match in two: a blunt « Team A to win » would wipe out that nuance. A prediction also makes it easy to spot genuinely open matches, where no outcome goes above 40%.
A prediction isn't judged on a single match. A 28% outcome that happens doesn't prove the model was wrong: it was expected a little more than once in four. What matters is calibration over a large number of matches (outcomes rated at 30% should happen about 3 times in 10) and accuracy measures such as the Brier score.
Finally, a prediction is not a price: it contains no commercial margin. It helps you understand a match and its possible scenarios, without ever guaranteeing a result.
Example
With 1.5 expected goals for the home side and 1.1 for the visitors, a Poisson model gives 46% / 26% / 28% for the 1X2, a 52% chance of both teams scoring and 48% for over 2.5 goals. Yet the most likely score is 1-1 (about 12%): the most likely outcome (a home win) and the most likely score don't always match.
How Elofoot uses it
This is the core of Elofoot. Every figure is computed server-side by a Poisson model (actual xG, form, league position, division strength, home advantage); for big fixtures where a Polymarket market is liquid enough, the 1X2 blends 65% market and 35% model. In the AI analysis, the AI writes the text but produces no numbers.
Frequently asked questions
- How is a football match prediction calculated?
- A model estimates each team’s expected goals, then derives the probability of every score, often with the Poisson distribution. Adding up the relevant scores gives the 1X2, both teams to score or the total number of goals.
- Was a prediction wrong if the favourite loses?
- Not necessarily. A team rated at 46% loses or draws more than one time in two. A prediction is judged over hundreds of matches, by its calibration, not on a single result.
- What is the difference between a prediction and a tip?
- A tip names a single outcome; a prediction gives the probability of each one. It therefore also shows how uncertain the match is.
See also
Go further
Put these concepts into practice with our AI football match prediction.
Predict a match for free