GGOALVIO

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.*