How we calculate the probabilities
No "sure tips". Here we explain how we estimate probabilities, what the model can't do and how we check how often it gets things right.
Team strength is measured on the quality of chances created and conceded (expected goals) and, where these are missing, on goals.
For each team we estimate attack and defence over all its matches and add the advantage of playing at home, measured on the league.
From expected goals we derive 1X2, Over/Under, BTTS and the correct-score map.
Absences and line-ups adjust the expected goals before the final calculation.
In short
We start from expected goals, which measure the quality of chances and not just the goals scored. For each match we estimate expected goals by combining one team's attack, the other team's defence and the league's home advantage. With few matches played, the estimates move towards the league average. Then we adjust them with the news.
From the expected goals a Poisson model derives 1X2, Over/Under, BTTS and the correct-score grid. They are estimates: a 70% probability is wrong three times out of ten, and we say so.
1. Expected goals
For each match we estimate each team's expected goals by combining how one attacks and how the other defends, relative to the league average, plus the advantage of the home side.
× attack of the home team
× defence of the visitors
Attack and defence are calculated over all of the season's matches (xG where available, otherwise goals); with few matches they move towards the league average. Then the news Radar adjustments are applied.
2. From goals to probabilities
The Poisson distribution tells you how likely a team is to score 0, 1, 2, 3… goals. Combining the two teams gives the probability of every correct score, and adding those up gives 1X2, Over/Under and BTTS.
Calculator
It's the same calculation as the dashboard. The grid shows the probability of every correct score (rows: home goals, columns: away goals).
Correct scores · %
How to read it
If the home team has 1.5 expected goals and the visitors 1.0, the single most likely result is often 1-0 or 1-1, but none goes above 15%: that's why we look at aggregate probabilities rather than a single result.
- Green: home win
- Grey: draw
- Blue: away win
The model's limits
Every model simplifies. These are the cases in which the estimates should be read with more caution.
With few matches played the data is unstable: a single game weighs heavily on the averages.
A team already safe or already qualified may play differently from what the numbers say.
The Radar updates the estimates when a story comes out, not before. Official line-ups arrive about an hour before kick-off.
Red cards, penalties and stoppage-time goals are hard to predict by definition.
Reliability
Every day we compare the model's probabilities with matches already played. No number is made up.
The numbers are calculated from the saved results and appear here at the next data update.
The probabilities of all of today's matches, calculated this way, are in the Scout Specialist Pro dashboard.
See pricing