A 1-0 win tells you who scored. It says far less about which side made the better chances, and that gap is where most football predictions are won or lost. Expected goals, or xG, was built to measure it. In just over a decade, it has moved from analysts’ spreadsheets to Saturday night television. For anyone following football predictions, it is one of the most useful numbers to look at before a match.
What xG measures
Expected goals measures how likely a shot is to be scored. Opta, the data company behind the most widely used model, rates each attempt on a scale from zero to one. Zero is a chance that cannot be scored and one is a chance that would be scored every time. A shot rated 0.2 was scored roughly one time in five across similar shots in the past.
Opta’s current model was trained on nearly one million shots and weighs more than 20 variables. They include the distance and angle from goal, the goalkeeper’s position, defensive pressure, the body part used and the type of pass before the shot. Penalties are the exception. They carry a fixed value of 0.79, which matches their historical conversion rate.
Add up a team’s shots for its match xG, then compare it with their opponents to see who created more, whatever the score.
Where it came from
Vic Barnett and Sarah Hilditch used the term expected goals in a 1993 paper on artificial pitches in English football. In 2004, Jake Ensum, Richard Pollard and Samuel Taylor studied 930 shots from the 2002 World Cup and found that distance, angle and the proximity of defenders all affected whether a shot went in.
The modern metric belongs to Sam Green, an analyst at Opta, who set out his model in April 2012 after studying more than 300,000 shots. It went mainstream in August 2017, when the BBC’s Match of the Day began showing xG figures at the start of the 2017-18 Premier League season.
Why it matters for predictions
Goals are rare, so luck plays a large part in any single result. A deflection, a goalkeeping error or a missed sitter can decide a match without saying much about how the teams played. The chances a team creates are a steadier guide. A side that keeps creating good chances tends to keep scoring, while a side that wins on a handful of shots tends to see that luck run out.
That is why analysts treat xG difference as a better guide to future results than goal difference. A team with poor results but a healthy xG difference is often a candidate to improve. A team high in the table on a negative xG difference is often due a fall.
Betting saw the value early. Matthew Benham, an Oxford physics graduate, founded the betting data company Smartodds in 2004. He later took control of Brentford and Midtjylland and ran both on analytics. Midtjylland won their first Danish title in 2015, and Brentford won promotion in 2021 to reach the top flight after 74 years away.
What xG does not tell you
One match is a small sample. A single xG total can swing on one or two big chances, so the figures carry far more weight across a run of games.
Models also differ. Providers such as Opta and StatsBomb use different data and variables, so the same shot can carry a different value. Stick to one provider rather than mixing them.
xG also rates the chance rather than the finisher. It will not tell you that one striker converts more often than another, and Opta itself notes that a higher xG total in a match does not necessarily mean a team should have won.
Reading xG before a match
Look at trends rather than single games. Check xG for and against over several matches, use non-penalty xG where the site offers it, and compare both with actual goals. Where the two have drifted far apart, the table may be flattering or underselling a side. That gap is where xG earns its place in a prediction.
