Poisson Distribution for Football Betting: Model and Tool

Use the calculator with expected-goal inputs that were estimated before the match. It does not infer team strength from names, odds or a live score⁠.

POISSON SCORE CALCULATOR

Turn two expected-goal inputs into a score matrix

Enter independent pre-match expected goals for each team. The calculator applies a basic independent Poisson model. It does not estimate the inputs or correct for changing team strength, game state or score dependence⁠.

Home expected goals (lambda)Away expected goals (lambda)

Home win

48.96%

Draw

24.89%

Away win

26.15%

Over 2.5

50.64%

Rank 1: 1-1

Model probability

11.83%

Fair odds

8.45

Rank 2: 1-0

Model probability

10.75%

Fair odds

9.30

Rank 3: 2-1

Model probability

9.46%

Fair odds

10.57

Rank 4: 2-0

Model probability

8.60%

Fair odds

11.62

Rank 5: 0-1

Model probability

7.39%

Fair odds

13.53

Rank	Score	Model probability	Fair odds before margin
1	1-1	11.83%	8.45
2	1-0	10.75%	9.30
3	2-1	9.46%	10.57
4	2-0	8.60%	11.62
5	0-1	7.39%	13.53

Accessible 0-5 score matrix. Every cell contains the model probability; colour is only a visual aid.

Home 0 goals

Away 0

6.72%

Away 1

7.39%

Away 2

4.07%

Away 3

1.49%

Away 4

0.41%

Away 5

0.09%

Home 1 goals

Away 0

10.75%

Away 1

11.83%

Away 2

6.51%

Away 3

2.39%

Away 4

0.66%

Away 5

0.14%

Home 2 goals

Away 0

8.60%

Away 1

9.46%

Away 2

5.20%

Away 3

1.91%

Away 4

0.52%

Away 5

0.12%

Home 3 goals

Away 0

4.59%

Away 1

5.05%

Away 2

2.78%

Away 3

1.02%

Away 4

0.28%

Away 5

0.06%

Home 4 goals

Away 0

1.84%

Away 1

2.02%

Away 2

1.11%

Away 3

0.41%

Away 4

0.11%

Away 5

0.02%

Home 5 goals

Away 0

0.59%

Away 1

0.65%

Away 2

0.36%

Away 3

0.13%

Away 4

0.04%

Away 5

0.01%

Poisson probability for each home and away score from zero to five

Home / away	0	1	2	3	4	5
0	6.72%	7.39%	4.07%	1.49%	0.41%	0.09%
1	10.75%	11.83%	6.51%	2.39%	0.66%	0.14%
2	8.60%	9.46%	5.20%	1.91%	0.52%	0.12%
3	4.59%	5.05%	2.78%	1.02%	0.28%	0.06%
4	1.84%	2.02%	1.11%	0.41%	0.11%	0.02%
5	0.59%	0.65%	0.36%	0.13%	0.04%	0.01%

Total line 0.5

Under probability

6.72%

Over probability

93.28%

Over fair odds

1.07

Total line 1.5

Under probability

24.87%

Over probability

75.13%

Over fair odds

1.33

Total line 2.5

Under probability

49.36%

Over probability

50.64%

Over fair odds

1.97

Total line 3.5

Under probability

71.41%

Over probability

28.59%

Over fair odds

3.50

Total line 4.5

Under probability

86.29%

Over probability

13.71%

Over fair odds

7.29

Total line	Under probability	Over probability	Over fair odds
0.5	6.72%	93.28%	1.07
1.5	24.87%	75.13%	1.33
2.5	49.36%	50.64%	1.97
3.5	71.41%	28.59%	3.50
4.5	86.29%	13.71%	7.29

Under 2.5: 49.36%. Both teams to score: 53.24%. The matrix covers 100.00% through dynamically selected tails of 17 home and 15 away goals; per team; displayed fair odds are the inverse model probability before margin⁠.

For each team and goal count k, the calculator uses P(X=k) = e^-lambda x lambda^k / k!, then multiplies the two team probabilities for each scoreline.

§ ON THIS PAGE · 9 sections

The Poisson football formula

For one team⁠:

P(X = k) = e^-lambda x lambda^k / k!

Where⁠:

• X is the number of goals⁠;
• k is the exact goal count being calculated⁠;
• lambda is the team's expected goals for the match⁠;
• e is the mathematical constant⁠;
• k! is the factorial of the goal count⁠.

If the home team's lambda is 1.60, the probability of exactly one home goal is⁠:

e^-1.60 x 1.60^1 / 1! = 32.30%

The probability of exactly two home goals is⁠:

e^-1.60 x 1.60^2 / 2! = 25.84%

Repeat for enough goal counts that the remaining unmodelled tail is negligible. The calculator expands the score grid dynamically for the selected lambdas, retains the remaining probability tails within its tolerance and reports how much total probability the displayed grid covers⁠.

Worked Poisson match: 1.60 versus 1.10 expected goals

Assume the home team has lambda 1.60 and the away team 1.10. The basic independent model produces approximately⁠:

For a 1‑1 score⁠:

P(home scores 1) x P(away scores 1)

32.30% x 36.62% = 11.83%

The five largest score cells are 1-1, 1-0, 2-1, 2-0 and 0-1. "Most likely score" does not mean likely in ordinary language: even the largest cell is only about 11.83 percent in this example⁠.

Turn the score matrix into market probabilities

Each cell represents one final score under the stated model⁠:

• sum cells where home goals exceed away goals for the home‑win probability⁠;
• sum the diagonal cells for the draw⁠;
• sum cells where away goals exceed home goals for the away win⁠;
• sum cells with three or more total goals for over 2.5⁠;
• sum cells where both goal counts exceed zero for both teams to score⁠.

Do not select only the single largest cell and ignore the rest. A home win, for example, can occur through many different scores. The full market probability is the sum of all qualifying cells⁠.

The draw betting guide explains the same aggregation principle for tied scorelines. The home versus away guide covers venue inputs that may change lambda⁠.

How should expected goals, lambda, be estimated?

Poisson does not create the expected-goal input. A basic reproducible method can start with⁠:

1. separate home and away scoring rates⁠;
2. league-level home and away baselines⁠;
3. each team's attack strength relative to the matching baseline⁠;
4. each opponent's defensive strength⁠;
5. a frozen date window and minimum sample rule⁠;
6. declared adjustments for team news or structural changes⁠.

The Smarkets source gives an example that combines attack strength, defence strength and league averages. The same source warns that an overly long window can become irrelevant while an overly short one is vulnerable to outliers⁠.

Never tune lambda after seeing the final score. Save the inputs, data cutoff and model version before kickoff⁠.

Where a basic Poisson model can fail

The simple two‑team calculator assumes⁠:

• each team's scoring follows a Poisson distribution⁠;
• its goal rate is constant through the match⁠;
• the two goal counts are independent⁠;
• the pre-match lambdas represent current strength⁠;
• the event and settlement period match the data⁠.

Football violates these assumptions in useful ways. A red card, early goal or tactical change alters the game state. Low scores can be dependent. Team strength changes over time. The original Dixon-Coles research uses a more involved model, and current research continues to test alternative dispersion structures⁠.

Treat basic Poisson as a transparent baseline, not a claim that every goal is independent or that one formula is universally best⁠.

A leakage-resistant backtest contract

Score the probability forecast, not only whether the top score happened. Useful diagnostics include calibration by probability band, log loss or Brier score, plus profit and loss at prices that were genuinely available⁠.

Compare a baseline with its proposed adjustment on the same frozen out-of-sample matches. If a more complex version improves the training data but not the held-out data, it has not demonstrated a better betting model⁠.

Compare model probability with the accepted price

Convert a model probability to margin‑free decimal odds⁠:

model fair odds = 1 / model probability

If the model says 48.96 percent for a home win, its fair price is about 2.04. That does not establish value. You must still account for model error, market margin, price movement and whether the offered market uses the same period⁠.

Do not compare a 90-minute Poisson output with a market that includes extra time. Do not compare a model's 1X2 draw probability with draw-no-bet. Use the betting value, ROI and risk guide to keep forecast probability and settled performance separate⁠.

Evidence manifest3 primary sources mapped to this guideView sources

Each source below is retained with the claims it supports. Operator sources describe published terms, not independent first‑hand performance⁠.

1. Dixon and Coles football score model ↗ (opens in a new tab)
   • Association football scores have been modelled with Poisson regression
   • The model requires adjustments for the data structure and changing team performance
2. Smarkets Poisson football calculation guide ↗ (opens in a new tab)
   • Poisson converts an expected goal rate into probabilities for individual goal counts
   • The matrix multiplies each team's goal probabilities and groups combinations into a home win, draw or away win
   • The model uses past results and does not account for squad changes, injuries, weather or correlations
3. Bayesian Conway-Maxwell-Poisson football score research ↗ (opens in a new tab)
   • The model quantifies departures from Poisson equidispersion and is applied to English Premier League scores
   • Team-specific dispersion can affect model fit and predictive performance

Update record, 21 August 2026: aligned the explanation with the calculator's dynamic probability-tail coverage instead of claiming a fixed 0–12 grid⁠.

Frequently asked questions

What does lambda mean in a football Poisson model?

Lambda is the expected number of goals for the team in the modelled match period. It is an input estimate, not the most likely exact goal count⁠.

Can Poisson predict the correct score?

It assigns probabilities to scorelines. It cannot guarantee the final score, and the most probable single score can still have a low absolute probability⁠.

How do I calculate a draw probability?

Multiply the home and away probabilities for each tied score, such as 0-0, 1-1 and 2-2, then sum those diagonal cells⁠.

Is Poisson suitable for live betting?

Not without updating the model for the current score, remaining time, red cards and game state. A frozen pre-match lambda is not automatically a live lambda⁠.

Is a Poisson model profitable?

The formula alone cannot establish that. Profitability requires calibrated inputs, prices above the model's true break-even threshold and controlled execution over a complete record⁠.