Expected goals explained simply
What xG measures, why it tells the story of a match better than the result, and how to read it on a team page.
Read →From expected goals to 1X2, Over/Under, BTTS and correct-score probabilities, step by step.
The Poisson model is the most common way of turning a match's expected goals into probabilities. It isn't the only one and it has limits, but it's simple, transparent and works surprisingly well. Let's look at it with an example.
Suppose that, combining both teams' attack and defence and home advantage, the model estimates 1.6 expected goals for the home team and 1.1 for the visitors.
The Poisson distribution tells you how likely a team is to score 0, 1, 2, 3… goals, given its expected goals. With 1.6 expected goals the home team has about a 20% chance of not scoring, 32% of scoring 1, 26% of scoring 2 and 14% of scoring 3. The visitors, with 1.1, have a 33% chance of drawing a blank, 37% of scoring 1 and 20% of scoring 2.
Multiplying the two teams' probabilities gives the probability of every correct score. In our example the most likely are 1-1 (about 11.8%), 1-0 (10.8%), 2-1 (9.5%) and 2-0 (8.6%). Note that even the most likely result doesn't reach 12%.
Adding up the right results gives all the other probabilities:
The model treats the two teams' goals as independent and knows nothing about motivation, fatigue or last-minute news. That's why expected goals need adjusting with the news, which is what the Radar does. Read more on the Method page, where you can also try the calculator.