What is xG (expected goals)?

xG is the probability that a shot becomes a goal, on a scale from 0 to 1.

Elliot Hayward

Editor

Updated , 3 min read

Illustration of half a football pitch from above with dots of different sizes, like a shot map.
On this page
  1. How is an xG value worked out?
  2. Why do apps show different xG for the same match?
  3. Is xG a good predictor?
  4. What are the limits of xG?
  5. How can you use xG as a fan?

xG stands for expected goals. A shot with an xG of 0.2 would generally be scored about twice in every 10 attempts, Hudl says. Read it as chance quality over several matches, not as the expected outcome of one game: The Analyst says xG measures the quality of a chance and not the expected outcome of the game.

How is an xG value worked out?

A model compares the circumstances of a shot with a large set of past shots. The Analyst's explainer says its model "evaluates how over 20 variables affect the likelihood of a goal being scored", among them distance, angle, goalkeeper position, pressure and shot type, and that it is powered by nearly one million shots from 40 competitions between 2018-19 and 2021-22.

Hudl says standard models use distance, angle, body part and the type of assist or previous action. Hudl Statsbomb adds goalkeeper position and the position of all attackers and defenders. Its example: a standard model might give a chance 0.30, where Hudl Statsbomb gives 0.65 "if the goalkeeper was out of position".

Why do apps show different xG for the same match?

Providers build their own models from different inputs, so the same shot can get different values. Hudl's example shows it: 0.30 from a standard model, 0.65 from Hudl Statsbomb when the goalkeeper was out of position.

We haven't compared any two apps' xG for the same match, and the pages we read don't publish the model behind an app's number.

Is xG a good predictor?

In the studies we found, xG predicted future results better than the traditional statistics it was compared with. In Brechot and Flepp's data it still leaves most of the variation unexplained.

Brechot and Flepp used 170,688 shots from 7,304 matches in five top European leagues, 2013/14 to 2016/17. The xG difference over a team's previous ten matches explained 32.0% of the variation in points in the next ten, against 25.3% for points won. The authors call their xG model "very basic", and the 2020 journal version of the paper may differ from this 2018 working paper.

Mead, O'Hare and McMenemy (2023) used Wyscout data from the 2017-18 season of the same five leagues. Their abstract says xG "was found to be a superior predictor of a football team's future success when compared to traditional statistics".

What are the limits of xG?

xG describes the quality of chances. It doesn't tell you who will win a match or how good a finisher is.

  • Variance. A goal is binary and xG is a probability, so Hudl says "some variance between goals and xG is expected over shorter periods".
  • One moment. Hudl says its approach values the goalkeeper's position only at the instant of the shot.
  • Finishing. Davis and Robberechts, in a 2024 arXiv preprint, say sustained overperformance of cumulative xG "requires both high shot volumes and exceptional finishing". So xG alone doesn't settle whether a player finishes well or badly.

How can you use xG as a fan?

Use it over several matches, next to the result, and treat it as one input. Open a match page and check that your app shows xG for your league before you rely on it.

Frequently asked questions

What does an xG of 0.3 mean?

The model gave that shot a probability of 0.3 of becoming a goal. Flashscore's FAQ reads an xG of 0.5 as a 50 % chance.

Is xG the same on every app?

There is no reason to expect it, because providers use different models and inputs. The Analyst's model uses 0.79 for a penalty and Hudl 0.78; we haven't compared two apps' xG for the same match.

Does a team with higher xG always win?

No. The Analyst says xG measures chance quality, not the expected outcome of the game, and Hudl says some variance between goals and xG is expected over shorter periods.

What is the xG of a penalty?

It depends on the provider. The Analyst's model gives penalties a constant 0.79, reflecting their historical conversion rate, while Hudl says most models use 0.76 and that its 2022 update changed its value to 0.78.

Why did my team lose with 2.1 xG?

A shot is either a goal or not, while xG is a probability, so results and xG can diverge. FotMob's FAQ says a team that loses 1-0 with an xG of 2.1 was "unlucky to lose".

Where can I see xG on Flashscore, FotMob or Sofascore?

All three list xG: Flashscore's Google Play listing names "xG data", FotMob's App Store listing names xG and shot maps, and Sofascore's names xG, shot maps and heatmaps. 365Scores, GOAL and Soccerway list it too. None of the listings we read says how fast xG appears or which leagues have it.