Expected points turn the quality of a team’s chances into the points it deserved, cutting through the luck that decides individual matches. This calculator takes both teams’ xG and returns expected points using the standard Poisson model from football analytics.
How the Poisson model works
Each team’s goals are treated as an independent Poisson random variable with a mean equal to its xG. The probability of any exact scoreline is the product of the two teams’ goal probabilities, and summing across the score matrix gives the three outcome probabilities:
P(home i goals) = e^-a * a^i / i! a = home xG
P(away j goals) = e^-b * b^j / j! b = away xG
P(home win) = sum of P(i)*P(j) for i > j
xPts(home) = 3 * P(home win) + 1 * P(draw)The away side’s expected points follow using its win probability and the shared draw probability. The model computes probabilities for scorelines up to 10 goals per side, which captures more than 99.99% of the probability mass for realistic xG inputs.
Worked example
Home xG = 1.8, away xG = 1.1:
- Home win probability ≈ 52%
- Draw probability ≈ 24%
- Away win probability ≈ 24%
- Home xPts = (3 × 0.52) + (1 × 0.24) = 1.56 + 0.24 = ~1.80 xPts
- Away xPts = (3 × 0.24) + (1 × 0.24) = 0.72 + 0.24 = ~0.96 xPts
The most likely single scoreline is often 1–1 or 1–0, because low-scoring results dominate the Poisson distribution at typical xG values. The 0–0 scoreline carries meaningful probability even when xG is above 1.
What xG ranges look like in practice
| Situation | Typical home xG | Typical away xG |
|---|---|---|
| Even match | 1.0–1.5 | 0.8–1.3 |
| Home side dominant | 1.8–2.5 | 0.4–0.8 |
| One-sided game | 2.5+ | 0.3–0.5 |
| Low-quality chances both sides | 0.5–0.8 | 0.4–0.7 |
Most competitive matches fall in the 0.8–2.0 xG range per team. Totals above 3 xG are unusual and typically indicate a very open match or a significant number of big chances.
Using xPts across a season
The real value of expected points is cumulative. In any single match, luck plays a large role — a team with 2.1 xG can still lose 0–1 to a wonder goal. Over 38 matches, luck averages out and a team’s actual points table increasingly converges toward its cumulative xPts.
Teams whose actual points significantly exceed their xPts are typically overperforming and often regress toward their xPts total in the second half of the season. Teams whose actual points significantly lag their xPts are often unlucky and may recover. This is why analysts use xPts as a season-level predictor and betting market signal, not just a match-level curiosity.
The model assumes the two teams’ goal-scoring are independent — the standard simplification in football analytics. Real matches include small correlations (a team chasing the game takes more risks, changing the scoring pattern), but the independent Poisson model is accurate enough for expected-points estimation and is the industry default.