Poisson Score Predictor
Enter home and away expected goals (xG). The Poisson distribution computes the probability of every score from 0-0 to 8-8, the home win / draw / away win probabilities, and the most likely scoreline.
Pre-match xG is the expected number of goals a team is forecast to score, derived from historical scoring rates, opponent defence, and venue. World Cup group-stage averages: favourites ~1.6, mid-table ~1.1.
| 1 - 1 | 12.05% |
| 1 - 0 | 10.95% |
| 2 - 1 | 9.34% |
| 2 - 0 | 8.49% |
| 0 - 1 | 7.77% |
| 0 - 0 | 7.07% |
What is the Poisson Predictor?
The Poisson distribution models goal-scoring as a random process driven by each team\'s expected goals (xG). Plug in the home and away xG for any match and this tool returns 1X2 probabilities, fair decimal odds, over/under totals, both-teams-to-score probability, and the most likely correct scores. It\'s the foundational model behind almost every football prediction engine.
How to use the Poisson Predictor
- Enter expected goals (xG) for the home team. Look up our xG API or use a back-of-envelope estimate (~1.4-1.8 for World Cup favourites).
- Enter expected goals for the away team.
- Read off the 1X2 percentages, fair odds, totals markets, and correct-score distribution.
- Compare fair odds to bookmaker prices to find value bets.
The formula
P(team scores k goals) = (λ^k × e^−λ) / k!
where λ = team's expected goals (xG)
Match scoreline probability assumes independence between teams:
P(Home=h, Away=a) = P_home(h) × P_away(a)
Then sum over the 9×9 score matrix to get 1X2, totals, BTTS, etc.
Fair decimal odds = 1 / probability.Worked example
England (xG 1.65) vs Senegal (xG 0.95) in a World Cup R16 match.
- P(England wins) ≈ 56.0% → Fair odds 1.79
- P(Draw) ≈ 22.5% → Fair odds 4.44
- P(Senegal wins) ≈ 21.5% → Fair odds 4.65
- Most likely scorelines: 1-1 (12.8%), 1-0 (10.4%), 2-1 (10.5%)
- Over 2.5: ~52%, BTTS: ~57%
If your bookmaker prices England at 1.85, that\'s implied 54.0% vs the model\'s 56.0% → +3.7% edge.
Get team-level xG for every match via API
7-day free trial. All endpoints on every plan.
Frequently Asked Questions
How accurate is the Poisson model?
Surprisingly good for a single-parameter model. It systematically under-predicts draws (because real-world goals aren't fully independent) and high-scoring outliers. Pro models like Dixon-Coles correct for this, but raw Poisson is a strong baseline used by most public predictors.
Where do I get xG numbers?
Use our [player & team stats API](/blog/world-cup-player-stats-api) which exposes match-level xG. Alternatively, Understat, FBref, and StatsBomb publish historical xG. For pre-match projections, average a team's last 5-10 matches' xG and adjust for opponent quality.
Why does the model under-predict draws?
Real football has scoring dependence (defensive shells when ahead, push-for-equaliser when behind) that Poisson assumes away. The Dixon-Coles correction boosts low-scoring draws (0-0, 1-1) and reduces 1-0/0-1 to fit historical data better.
Can I use this for live in-play predictions?
Yes, but adjust xG to reflect remaining match time (e.g. λ_remaining = λ × minutes_left/90). The model is most useful pre-match where the underlying scoring rate is stable.
How do I incorporate home advantage?
Add ~0.2-0.3 xG to the home team and subtract ~0.1-0.15 from the away team. For neutral-venue World Cup matches, no adjustment is needed.
Why are my fair odds different from bookmaker odds?
Bookmakers add 5-8% margin and have access to better data (injuries, motivation, weather). The difference between your fair odds and the bookmaker's isn't pure edge - it's a starting point for finding mispriced markets.