Learn
Poisson Goal Models
The statistical workhorse that turns expected goals into match probabilities
What is a Poisson distribution?
From xG to match probabilities
The Dixon-Coles adjustment
Where Poisson still struggles
Frequently asked
Why use Poisson and not just match data?
Because Poisson lets you derive every market from one underlying surface. Fitting a separate model for 1X2, BTTS, Over/Under 2.5, etc. requires more data and introduces inconsistencies. A single Poisson scoreline matrix gives all markets consistently.
What's the lambda for an average Premier League team?
The league-wide average is around 1.35 goals per team per match. Stronger teams sit around 1.8–2.0; weaker teams around 0.9–1.1. These numbers shift when you apply the home advantage multiplier (~1.15–1.25 for home, ~0.80–0.90 for away).
Does BetsPlug use plain Poisson or Dixon-Coles?
We use a Dixon-Coles-style correction on top of the base Poisson. The correlation parameter is fit per league so we don't apply the same correction to high-draw and low-draw environments.
Can Poisson predict correct scores?
Yes - the scoreline matrix reads directly as correct-score probabilities. The most-likely score is usually 1-1 or 2-1 in balanced fixtures; the probabilities for exotic scorelines (4-3, 5-1) are small but measurable.
How often do you re-fit the Poisson parameters?
Team-level attack and defence strengths update after every match. The Dixon-Coles correlation parameter updates weekly from the previous 38-match rolling window per league, which is enough to capture seasonal trend shifts without over-reacting to single-match noise.