Expected Goals in the Premier League: How to Read xG

Expected goals (xG) measures the quality of the shots a player or team takes—not how many goals they were guaranteed to score. Each shot receives an estimated scoring probability between 0 and 1. A chance worth 0.20 xG would be scored about 20 times in 100 comparable attempts, according to the model. Add the shot values together to get a player’s or team’s total. Opta’s explanation sets out this distinction.

That makes xG useful alongside a Premier League scoreline: it separates creating chances from converting them. But it is not an alternative result or a verdict on who deserved to win.

What goes into the number?

An xG model compares a shot with historical attempts, considering features such as distance, angle, body part and the preceding action. Opta’s published methodology also includes goalkeeper position, defensive pressure and how clearly the shooter can see the goalmouth. A central tap-in and a header from a tight angle therefore receive different values, even if both count as one shot. Source: Opta methodology.

Total xG is a sum, not a rate: it is expressed in expected goals, with no shots or minutes denominator. Two related rates answer different questions:

• xG per shot = total xG ÷ shots: how good was the average chance?
• Player xG per 90 = total xG ÷ minutes played × 90: how much scoring opportunity did the player get relative to playing time?

Use the same set of attempts for the numerator and denominator: non-penalty xG per shot, for example, needs non-penalty shots. Neither rate replaces the total. A substitute can have an impressive per-90 figure from very few minutes; a team can produce high-quality shots but too few of them.

A real example: Liverpool 4–2 Newcastle

Liverpool’s home win on 1 January 2024 shows why goals, shots and xG tell different parts of the story. They took 34 shots and generated 7.27 xG in Opta’s model, yet scored four goals. Sources: Sky Sports match details and match report.

Liverpool statistic	Value	What it tells you
Goals	4	The actual scoring output
Shots	34	The volume of attempts
Opta xG	7.27	The sum of the estimated scoring probabilities for those attempts
xG per shot	About 0.21	The average chance quality: 7.27 ÷ 34

The average is calculated from the reported figures; it does not mean every shot had a 21% scoring probability. Unlike an individual shot’s probability, a team’s xG total can exceed 1.

The match contained both quantity and quality. Darwin Núñez squared for Mohamed Salah’s opener, and Diogo Jota supplied Curtis Jones for another close-range finish. Liverpool also had two penalties, one saved and one scored. Those spot-kicks matter when interpreting the total: it was not all chance creation from open play. Source: Sky Sports report.

Liverpool finished 3.27 goals below their xG, but that subtraction does not prove they should have scored exactly seven. Newcastle goalkeeper Martin Dúbravka made 10 saves. Neither that save count nor the goals-minus-xG gap alone separates Liverpool’s shot execution from his shot stopping.

How to read a Premier League xG table

For public team and player data, Understat’s EPL page provides xG tables and filters. Select the intended season and date range before comparing clubs.

Useful measures include:

• xG: expected goals from shots taken; higher means more combined scoring opportunity.
• xGA: expected goals against, from opponents’ shots; lower means less opportunity conceded.
• xG difference: xG minus xGA, balancing attacking output against chances allowed. Understat’s NPxGD column is the non-penalty version, excluding penalties on both sides.
• Non-penalty xG (NPxG): removes penalty attempts, helping separate spot-kick opportunities from other shots—not just open-play chances.

Check matches played before comparing totals. For players, check minutes and penalty involvement too. A goals-minus-xG figure describes conversion above or below the model’s estimate; it does not, by itself, establish lasting finishing ability. Hudl Statsbomb’s explanation stresses the importance of large samples when identifying finishing ability.

Where xG can mislead

Keep comparisons within one provider’s model. Different inputs can produce different values for the same shot, as Hudl Statsbomb explains. Do not splice Understat figures into an Opta season total.

Also remember what is missing: a dangerous attack that ends without a shot contributes nothing to standard shot-based xG. Read the shot sequence alongside the match situation—especially the score, red cards and timing of chances—rather than treating the final total as a complete measure of control.

Finally, ordinary xG measures the chance before the shot’s outcome. To assess goalkeeper shot stopping, a post-shot measure also considers the execution of on-target attempts, such as where they were placed in the goal. Opta explains that distinction; our guide to goals saved above expected explains how to read the goalkeeper statistic. Use xG to ask whether a team is creating and conceding good chances, not to rewrite the scoreline.