How to Convert Expected Goals Into Football Betting Probabilities

Woman taking notes beside a laptop showing a live football match and betting data by the pool. — Expected goals can be converted into score probabilities with a relatively compact mathematical model.

Expected goals can be converted into score probabilities with a relatively compact mathematical model. Take a hypothetical match in which the home side is assigned 1.60 xG and the away side 1.10 xG. A Poisson distribution can turn those two estimates into probabilities for each goal total. Those figures can then be combined into exact scores and match-result percentages. In mixed sports-and-games catalogues, Thimblescan appear beside football markets while the statistical calculation remains focused on the match itself. The useful part is not finding one supposedly certain score. It is seeing how assumptions about scoring rates translate into a complete probability distribution.

Step 1: Expected-goals inputs set the starting rates

A Poisson model needs one expected scoring rate for each team. In this example, the home rate is 1.60 and the away rate is 1.10.

Those numbers can be estimated from recent attacking xG and the quality of chances conceded by the opponent. Home-away performance can refine the picture. Opponent strength provides another adjustment.

Expected lineup changes also matter because the absence of a regular attacker can alter the likely chance quality. The same applies when a defensive change materially affects the opposition's attacking outlook.

Consistency in the data source is important. xG is a model-generated estimate rather than a universal measurement. Two providers can assign different values to the same shot because their models use different inputs or weighting methods. Mixing one team's figure from one provider with its opponent's figure from another can therefore create an inconsistent starting point.

Historical xG is also descriptive. It measures the estimated quality of previous chances rather than guaranteeing that the next match will produce chances of the same quality.

Step 2: The Poisson formula converts xG into goal probabilities

For a basic model, each team's xG estimate becomes the Poisson rate, represented by λ.

The formula is:

P(X = k) = (e^-λ × λ^k) / k!

Here, k represents the number of goals being tested. The value λ is the expected scoring rate for the team. The constant e is approximately 2.71828. The factorial k! multiplies all positive integers from k down to 1.

For the home team, λ equals 1.60. The probability of scoring exactly once is therefore:

P(X = 1) = e^-1.60 × 1.60¹ / 1!

That produces approximately 32.30%.

For zero goals, λ remains 1.60 but k becomes zero. The result is about 20.19%.

The same process is repeated for the away team with λ set to 1.10. Once enough goal totals have been calculated, the model provides the building blocks for the scoreline matrix.

Step 3: Individual goal probabilities form two scoring distributions

Using 1.60 xG for the home side produces the following distribution. The same calculation with 1.10 gives the away probabilities.

Goals	Home probability	Away probability
0	20.19%	33.29%
1	32.30%	36.62%
2	25.84%	20.14%
3	13.78%	7.38%
4	5.51%	2.03%
5	1.76%	0.45%

The home distribution peaks at one goal. The away distribution does the same, although its probability of scoring zero is considerably higher.

That difference begins to explain why the home side later receives the larger match-win probability. It still does not identify one match result as certain.

The table stops at five goals for readability. A full calculation can continue beyond that point because Poisson probabilities never become exactly zero. Those small tail probabilities are included when calculating the final match-result totals.

Step 4: The scoreline grid combines both teams

An individual score probability comes from multiplying the relevant home probability by the corresponding away probability.

For 0–0:

20.19% × 33.29% ≈ 6.72%

For 1–0:

32.30% × 33.29% ≈ 10.75%

For 1–1:

32.30% × 36.62% ≈ 11.83%

For 2–1:

25.84% × 36.62% ≈ 9.46%

A larger section of the matrix looks like this:

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%

The largest individual cell is 1–1 at about 11.83%. Calling it the most likely exact score is mathematically accurate within this model. Calling it likely to happen in an everyday sense would be misleading because its probability remains below 12%.

Within a mixed catalogue, Spin & Wincan sit beside sports markets while this type of score calculation remains a separate football model based on expected scoring rates.

Step 5: Match-result probabilities convert into fair decimal odds

The score matrix can also be grouped by result.

Every cell where the home score exceeds the away score contributes to the home-win probability. Equal scores form the draw probability. The remaining cells contribute to the away win.

Using the full Poisson distributions rather than stopping at five goals gives:

Match outcome	Model probability	Fair decimal odds
Home win	48.96%	2.04
Draw	24.89%	4.02
Away win	26.15%	3.82

Fair decimal odds come from the reciprocal of the probability:

Fair odds = 1 / probability

The home probability of 48.96%, expressed as 0.4896, therefore gives:

1 ÷ 0.4896 ≈ 2.04

The same method produces approximately 4.02 for the draw and 3.82 for the away result.

These are model prices without a bookmaker margin. A quoted market price can be compared with them, but a small numerical difference by itself is weak evidence because the original xG assumptions and the statistical model both contain uncertainty.

A price above the calculated fair level can indicate a gap between the market and the model. It does not establish that the selection will succeed or produce a positive return.

Step 6: Model limitations define how the fair prices are read

The basic Poisson approach makes several simplifying assumptions. It treats the scoring processes of the two teams as independent. It also assumes that each team's expected scoring rate remains constant across the match.

Real football is less tidy. A goal can alter how both sides play. Tactical changes can affect chance creation after the initial xG estimate has been set.

Unavailable players can change the quality of the inputs as well. Match circumstances may also produce a pattern that historical averages did not capture.

Low-scoring outcomes deserve particular attention. Standard independent Poisson models can misrepresent score combinations around 0–0 or 1–1. The Dixon–Coles approach is one established extension designed to adjust some of these low-score effects.

Sample size creates another limitation. A short run of matches can produce unstable attacking or defensive xG estimates, particularly when the opposition has varied substantially in strength.

The worked example therefore shows a reproducible chain rather than a prediction formula with guaranteed results: 1.60 and 1.10 become goal distributions, the distributions become scorelines, and the full matrix produces fair match-result prices. The arithmetic is exact once the inputs are fixed. The quality of the final estimate still depends on how realistic those inputs are and how well the model represents the match being assessed.