How models predict football
The Poisson goal model
5 min
How do you turn "this team is good" into a probability of 2–1? The workhorse of football prediction is the Poisson model, and it's the heart of how FinalSkore predicts football.
The core idea
Goals arrive at a roughly steady but random rate, and the Poisson distribution is the standard maths for "how many rare, independent events happen in a fixed window". Feed it an expected goals rate for each team and it returns the probability of 0, 1, 2, 3… goals for that side.
Where the rates come from
FinalSkore estimates each team's scoring and conceding rate from a rolling last-10-matches window, adjusted by head-to-head history and home/away context. Combine the home team's attack with the away team's defence (and vice versa) and you get an expected goals figure for each side.
From rates to every market
Once you have a goal-probability grid for both teams, almost every market falls out of it:
- 1X2 — sum the grid cells where home > away, home = away, away > home.
- Over/Under 2.5 — sum every cell where total goals clear (or miss) the line.
- BTTS — the probability both teams' goal count is at least 1.
That's the power of a goal model: one estimate of expected goals quietly prices a whole board of markets at once.
Poisson assumes goals are independent and the rate is stable — neither is perfectly true. It's a strong approximation, not reality.