Expected goals (xG) explained: what it measures and how to read it before a bet

One number for the quality of a chance

Expected goals, or xG, measures the quality of a chance by calculating how likely it is to be scored. It runs from 0 to 1: a 0.1 xG chance means a player would, on average, score one goal from every ten shots in that situation.

Add up the xG of every shot and you get a team's or a player's expected goals over a match or a season: what an average finisher would have scored from the same chances.

What the model looks at

Opta's model uses more than 20 variables. The most important:

• the distance and the angle to goal;
• the goalkeeper's position and how clear the goal mouth is;
• the pressure from defenders;
• the type of shot: foot, header, volley, one-on-one;
• the pattern of play: open play, fast break, free kick, corner;
• the action before the shot, such as the type of assist.

It was trained on nearly one million shots from 40 competitions, from 2018/19 to 2021/22, and women's football has its own model.

Penalties, npxG and the family of metrics

A penalty gets a fixed value equal to its historical conversion rate: 0.79 xG. Because penalties are so valuable and so rare, analysts often remove them: non-penalty xG (npxG) is the better measure of the chances a team creates from open play and set pieces.

• xGA (xG against): the xG of the chances a team concedes, a measure of how well it prevents them.
• xA (expected assists): how likely a pass is to become an assist, whether or not the receiver shoots.
• xGOT (xG on target): only shots on target, with where the ball ends up in the goal. A 0.3 xGOT shot has a 30% chance of going in.
• xG per 90: xG divided by minutes played, times 90, to compare players and teams with different playing time.

When the table and the chances disagree

After 11 matches of the 2025/26 Premier League, Opta found that Crystal Palace had scored 14 goals from 19.3 xG, the biggest underperformance in the league: on expected points, they would have been 3rd, not 10th. Wolves had 2 points against 11.2 expected points.

Shot location is what xG rewards. In 2022/23, Alexis Sánchez (Marseille) and Rúben Neves (Wolves) each took 63 non-penalty shots: Sánchez's chances were worth 10.2 xG and he scored 12; Neves's were worth 2.8 and he scored 3.

How to use xG before a bet

• Compare goals with xG, and goals conceded with xGA, over a large sample: large gaps driven by finishing or goalkeeping tend to shrink over time.
• Be wary of small samples: three or four matches of xG are very noisy.
• Use xG as one input for totals and both-teams-to-score lines, alongside team news and the fixture list, never as a guarantee.
• Prefer npxG when one team has had several penalties: they can inflate xG without saying much about open play.

💡 A team that has scored more than its xG has not "used up" its luck: past goals are banked and do not have to be paid back. Thinking otherwise is the gambler's fallacy.

What xG does not tell you

xG measures the quality of chances, not the expected result of a match. It does not account for the quality of the players involved, a top striker and an average one get the same value for the same shot, nor for the game state: a team leading 2-0 may happily concede low-value shots. Use it to ask better questions, not to replace them.

Bet responsibly

Betting is paid entertainment, for adults only (19 and over in Ontario, 18 and over in Alberta), never a way to make money. Set a budget before you start and stick to it, whatever happens.

💡 Free help, 24/7: ConnexOntario 1-866-531-2600 in Ontario, Addiction Helpline 1-866-332-2322 in Alberta.