5/8/2026 · GoalVio Analysis
The Poisson Distribution: The Maths Behind Football Predictions
Serious football analysts use the Poisson distribution to predict scorelines. The intuition is simpler than the name suggests — and it works better than you'd expect.
# The Poisson Distribution: The Maths Behind Football Predictions
When GoalVio Intelligence generates a predicted scoreline, it uses a mathematical model called the Poisson distribution. The name sounds intimidating. The idea isn't.
## Why Football Suits This Model
Football is a low-scoring sport. Most matches end 1-0, 1-1, 2-1, or 2-0. Goals are rare events that happen at roughly predictable rates.
That's exactly the kind of process the Poisson distribution was designed to model: counting events that happen at a known average rate, independently of each other. Classic examples include calls arriving at a call centre per hour, accidents at an intersection per month, and goals a football team scores per match.
## The Formula
**P(k) = (λᵏ × e⁻λ) / k!**
Where:
- **λ (lambda)** is the expected number of events — in football, expected goals
- **k** is the number of events you're calculating the probability for
- **e** is Euler's number (≈ 2.718)
## A Worked Example
Arsenal's expected goals for a match: 1.6 (based on their recent form and the opponent's defensive record).
- P(0 goals) = **20.2%**
- P(1 goal) = **32.3%**
- P(2 goals) = **25.8%**
- P(3 goals) = **13.8%**
- P(4+ goals) = **7.9%**
Now Chelsea, with an xG of 1.1:
- P(0 goals) = **33.3%**
- P(1 goal) = **36.6%**
- P(2 goals) = **20.1%**
- P(3 goals) = **7.4%**
## Building the Scoreline Matrix
Multiply the two distributions together to get the probability of every possible scoreline.
P(Arsenal 2–1 Chelsea) = P(Arsenal scores 2) × P(Chelsea scores 1) = 25.8% × 36.6% = **9.4%**
Do this for every combination up to 5–5 and you have a complete scoreline probability matrix. The most likely scoreline has the highest probability. Sum the right cells and you get win/draw/loss probabilities.
## Why It Works
The Poisson model makes two assumptions: goals happen at a constant average rate, and goals are independent of each other. Neither is perfectly true. But both are close enough that the model outperforms simpler approaches — especially for predicting exact scorelines and over/under markets.
Studies consistently show Poisson-based models beat naive win/draw/loss probability models when tested against real results.
## The Limitations
The basic model treats both teams' scoring as independent. In reality there's a negative correlation — when one team scores, the other often changes their approach. Some advanced models use a "bivariate Poisson" to account for this.
The model also can't know about late team news, weather, or psychological factors. But over a large sample, those things average out. The Poisson model remains the industry standard because it's principled, transparent, and validated against thousands of real matches.
## How GoalVio Implements It
We calculate λ for each team by blending their recent form (70%) with historical win percentages (30%), then adjusting for relative attack and defence strength. We run the full Poisson matrix to generate scoreline probabilities, from which we extract the predicted score, win probabilities, and over/under figures.
The maths is the same maths professional analysts use. We just make it visible.
---
*See the Poisson model in action on any match's Intelligence tab at GoalVio.*