Expected goals (xG) explained
By ProbaPredict Data Desk · Last reviewed
Expected goals (xG) is the number of goals a team would be expected to score on average, given its chances. Shot-based xG adds up the probability that each shot becomes a goal. Model-based expected goals, like ours, forecast a team's goals before a match from its attack, the opponent's defence and home advantage.
What is shot-based xG?
Data companies assign every shot a probability of becoming a goal, based on things like distance, angle, body part and type of assist. A penalty is worth about 0.76 xG, a tap-in from two yards perhaps 0.6, a 30-yard effort 0.02. Add up a team's shots and you get its xG for the match. If a team has 2.3 xG and scores once, it created chances that usually produce more than one goal.
What is model-based expected goals?
Prediction models also produce an "expected goals" number, but before kick-off. Ours estimates each team's attack strength and defence strength from past results, adds a home-advantage factor and combines them: expected home goals = home attack × away defence × home advantage. It is a forecast of the scoreline, not a summary of shots. Each match page shows both teams' expected goals.
How do expected goals turn into predictions?
Goals in football are rare, roughly independent events, so a Poisson distribution describes them well. With expected goals of 1.5, a team's chances of scoring 0, 1, 2 and 3 are 22.3%, 33.5%, 25.1% and 12.6%. Combining both teams' distributions gives every scoreline, and from there the 1X2, Over/Under and most-likely-score figures. The steps are on our methodology page.
How many goals do teams score per game?
League scoring levels are the backdrop for every expected-goals figure. Over the last five seasons, matches in our nine leagues averaged 2.73 goals: 1.52 for the home side and 1.21 for the away side. The Bundesliga was the highest-scoring league and the Brasileirão Série A the lowest.
| League | Period | Matches | Over 1.5 | Over 2.5 | Over 3.5 | Goals per game |
|---|---|---|---|---|---|---|
| Premier League | This season (2026-27) | 50 | 72.0% | 52.0% | 34.0% | 2.82 |
| Last 5 seasons | 1900 | 79.8% | 56.6% | 34.2% | 2.93 | |
| Championship | This season (2026-27) | 95 | 80.0% | 64.2% | 36.8% | 2.92 |
| Last 5 seasons | 2765 | 72.7% | 47.5% | 24.2% | 2.53 | |
| La Liga | This season (2026-27) | 69 | 76.8% | 56.5% | 39.1% | 3.04 |
| Last 5 seasons | 1900 | 72.4% | 47.4% | 25.0% | 2.59 | |
| Bundesliga | This season (2026-27) | 36 | 86.1% | 77.8% | 61.1% | 3.81 |
| Last 5 seasons | 1530 | 83.5% | 61.0% | 40.8% | 3.18 | |
| Serie A | This season (2026-27) | 50 | 82.0% | 54.0% | 28.0% | 2.90 |
| Last 5 seasons | 1900 | 74.4% | 48.9% | 26.3% | 2.61 | |
| Ligue 1 | This season (2026-27) | 45 | 77.8% | 60.0% | 33.3% | 2.84 |
| Last 5 seasons | 1677 | 77.2% | 53.5% | 31.5% | 2.82 | |
| Eredivisie | This season (2026-27) | 63 | 90.5% | 76.2% | 63.5% | 3.81 |
| Last 5 seasons | 1530 | 81.6% | 59.1% | 37.3% | 3.06 | |
| Primeira Liga | This season (2026-27) | 62 | 75.8% | 54.8% | 37.1% | 2.74 |
| Last 5 seasons | 1530 | 73.3% | 51.0% | 28.4% | 2.65 | |
| Brasileirão Série A | This season (2026) | 277 | 78.3% | 51.6% | 25.6% | 2.67 |
| Last 5 seasons | 1900 | 69.8% | 44.3% | 21.1% | 2.41 | |
| All nine leagues | Last five seasons | 16632 | 75.7% | 51.6% | 29.2% | 2.73 |
Source: ProbaPredict analysis of 17,379 matches, updated Fri, 2 Oct 2026. "Last five seasons" covers 2021-22 to 2025-26 (calendar years for the Brasileirão). Scores are after 90 minutes.
Worked example
Suppose our model has the home side's attack at 1.30 (30% better than league average), the away side's defence at 1.10 (concedes 10% more than average), a league baseline of 1.25 goals per team and a home factor of 1.12. Expected home goals ≈ 1.25 × 1.30 × 1.10 × 1.12 = 2.00. If the away side's expected goals come out at 0.95, the total is 2.95, and the model would give Over 2.5 goals roughly a 57% chance.
Why do people use xG?
Because goals are noisy. A team can win 1-0 with one shot or lose 0-1 after dominating. Shot-based xG strips out some of that luck, which helps judge performance. Model expected goals do something similar for forecasting: they average over many matches rather than reacting to the last result.
Limits of expected goals
Shot-based xG ignores chances that never became shots, and different providers' models disagree. Our model-based figure only knows results, not line-ups or injuries. Neither tells you what will happen in a single match; both describe averages.
UK and US terminology
"xG" is used on both sides of the Atlantic. US coverage sometimes says "expected goals" in full on soccer broadcasts; the meaning is identical.
Common mistakes with xG
- Treating xG as a prediction of the exact score. 1.5 xG often means one or two goals, sometimes none.
- Mixing up shot-based and model xG. One measures chances; the other forecasts goals.
- Over-reading one match. xG is most useful over many games.
- Assuming every provider agrees. Different xG models give different numbers.
How do expected goals relate to Over/Under and BTTS?
The total of both teams' expected goals is the single biggest driver of Over/Under probabilities. A combined total of 2.2 typically gives Over 2.5 around 38–40%; 2.8 gives about 53%; 3.4 gives about 66%. BTTS depends more on the lower of the two figures, because both teams need to score: a 2.5 vs 0.6 split has the same total as 1.6 vs 1.5 but a much lower BTTS chance.
That's why our match pages show both teams' expected goals rather than just the total. Together with the most likely scorelines, they let you see the shape of the forecast, not just its headline numbers.
Frequently asked questions
- What does xG of 1.5 mean?
- That the team would score 1.5 goals on average if the same situation were repeated many times. In a single match it might score 0, 1, 2 or more.
- Is your expected goals figure shot-based xG?
- No. Our expected goals come from a pre-match model based on results, not from shot data. It is a forecast of goals, not a measure of chances created.
- How is xG used to predict scores?
- Each team's expected goals feed a Poisson distribution, which gives the chance of scoring 0, 1, 2 and so on. Combining both teams gives every scoreline's probability.
- Can a team beat its xG for long?
- Over a short run, yes, because of finishing variance and goalkeeping. Over a long run, goals and xG tend to converge, though elite finishers can sit slightly above.
Related pages
- Our methodology
- How prediction models work
- Scoreline probabilities
- Today's predictions with expected goals
- How Do Football Prediction Models Work?
- What Does a 60% Chance Really Mean?
- How Accurate Are Football Predictions?
- How Big Is Home Advantage in Football?
- All football betting guides
18+. Gambling can be addictive; only bet what you can afford to lose. Free help: BeGambleAware.org · GamStop.