Expected Goals (xG): what it is and how it works

Category: News | Team Analysis

‘Expected goals’ (xG) is a metric that analyses the quality of the chances that take place in a match. Each shot has a value that shows the probability of becoming a goal. This number is the percentage (0.1 means that this shot ends up in goal 10% of the time) that is assigned according to […]


‘Expected goals’ (xG) is a metric that analyses the quality of the chances that take place in a match. Each shot has a value that shows the probability of becoming a goal. This number is the percentage (0.1 means that this shot ends up in goal 10% of the time) that is assigned according to the place of the shot, the distance to the goal and the angle, as well as the part of the body (head, foot or others) with which the ball is shot, adding to the variable the speed of the play, the pass that precedes – if any – the shot, etc. To facilitate the understanding of the tool we are going to show examples of the model of goals expected from Driblab.

How the expected goals began to be used (xG)

This model was started in 2013 by our CEO Salvador Carmona. This metric responds to the need to quantify the value of scoring chances. There were already models that answered these questions but at Driblab we work with different parameters and it differs from the rest of the companies in the amounts of xG it gives to the shots. The first ones who created a model of xG were the bookmakers, soon after, were people who worked in clubs or pretended to do it, as well as statisticians who made their models for pure hobby. Step by step the xG models have made a space in the analysis with pages like FiveThirtyEight, even having a place in international media like ESPN, The Guardian or FourFourTwo reaching the coaching staffs even being publicly mentioned as the example of Frank Lampard this season.

Our models have maps that allow the visualization of this metric. In our case our CTO Coré Ramiro developed maps where the Driblab model is shown in images like the one we see below corresponding to the second leg of the Champions League between Liverpool and Atletico Madrid:

Each circle represents a chance, the size is the xG value (the probability of becoming a goal), and the filled circles are the goals. The sum of xG in the score above is a sum of each team’s total xG. In this case we see how Atletico were very effective with three goals on low-probability shots while Liverpool had low scoring efficiency part of blame to Jan Oblak’s great game saving more than expected.

Clarifications of our xG model

When evaluating the expected goals (xG) we calculate the quality of the chances. So a key indicator, beyond the total volume of xG in a match, is how clear those chances were. For example, in the first leg Atletico 1-0 Liverpool (image below) we see a very similar sum of xG for the two teams that tells us about an even match but if we analyze the quality of those shots Liverpool conceded in the goal a shot with 0.59 xG (1-0 Saul), that is, a shot that is a goal 59% of the times; while the most dangerous shot of Liverpool was 0.21 xG. Looking at the quality of the shots is a clear indicator beyond the total volume of goals expected.

xG applied to player analysis

The value of the expected goals also allows for an analysis of a player’s effectiveness and ability to score. The use of this metric allows us to see the performance and evolution of the players at finishing and creating xG: the most relevant attacking players usually score more than expected. The example of Erling Haland is very enlightening as he stands out for his reliability in this regard: since he arrived at Borussia Dortmund he has scored more than twice as much as expected. In addition, another indicator that allows us to analyze the performance is the ‘xG per shot’ that shows us how efficient the selection of shots is: the higher xG per shot the less haste in those shots and the higher probability of scoring. In the case of Erling Haland we can see how this selection of shots with such a high xG per shot (0.22 xG per shot) is key to understanding why he is overperforming.

Similarly, the expected goals serve to compare the performance of defensive systems (how much they concede and the quality of the chances) and help to add context to goalkeeper performances. The example we will see below is very clear by analyzing the shots faced to goal (xG that a goalkeeper face) with the goals conceded. A clear example is Kepa Arrizabalaga, who this season is one of the worst goalkeepers in terms of expected goals. When evaluating this performance we must also consider the quality of the chances conceded (xG per shot) as well as the comparison between xG received and goals conceded. In this case we see how he is underperforming and that the opponents have found many shots in areas very close to the goal.

xG applied to team analysis

Another application that has the model of expected goals is its usefulness in measuring team performances. This analysis can differentiate the type of action with which these expected goals were created: open play, counter-attack, free kick, set piece or corner kick. For example, below we see the performance of Real Madrid in open play and the problems in finishing scoring fewer goals than expected this season. In this case, we are looking at a team that is scoring less than expected.

We are Driblab, a consultancy specialized in the statistical analysis of players and teams; our work is focused on advising and minimizing risk in professional football decision-making in areas related to talent detection and footballer evaluations. Our database has more than 130,000 players from more than 120 competitions, covering information from all over the world. Here you can learn more about how we work and what we offer.

Autor: Driblab
Para News | Team Analysiswe also recommend you:

What is going on with Chelsea?

Chelsea’s past transfer market suggested they were a title contender but the table puts them a long way from that fight. What are the problems Frank Lampard needs to fix?

Liverpool without Van Dijk: our collaboration in The Sunday Times

In the preview of one of the most eagerly awaited matches in the Premier League, we have collaborated with The Sunday Times in their analysis of Liverpool.

Passing Contribution: Introduction

Today we introduce passing contribution, a new way to measure the quality of passes. The algorithm takes into account both the decision-making and the output of the pass.

Driblab in Radio Marca: interview to Salvador Carmona

Our CEO, Salvador Carmona, was interviewed this week on Radio Marca to talk about football, big data and how both topics converge in Driblab. In the show called Marcador, which is broadcasted daily from 8pm to 11:30pm, Carmona explained how big data and our models can...

Letter from the CEO – End of the year 2020

Dear all, As this dark 2020 comes to an end it is time to reflect on what has happened in our industry over the past year.  Like in many industries we were taken by storm with the COVID-19 pandemic and football was not an exception: leagues stopped, games were...

Introducing Game Flow Charts

During the last World Football Summit we talked about the importance of data in the sports industry, now, following our will to keep creating insightful tools to help you understand football games from a data-driven perspective, today at Driblab we introduce our...

The promise of Brazil that Europe has in its sights

The next transfer window is just around the corner and one of the most talked about names in the old continent is the Brazilian player Gabriel Veron. The right winger, who made his debut in the first division of a Palmeiras - Fluminense team in November 2019, is...

Eduardo Coudet: analyzing the new coach of Celta de Vigo

The Argentinean Eduardo Coudet, new coach of Celta de Vigo, will make his debut this Sunday against Sevilla in a new fixture of La Liga. We used our models to understand the work that the coach did in his previous clubs. The ex-footballer, who previously coached Inter...

Driblab will be part of the World Football Summit

On November 24, our CEO, Salvador Carmona, will be present at the most important event in the world of football to discuss the importance of big data in the sports industry. The fourth edition of the World Football Summit will take place from November 23rd to 27th...

How significant are penalties in European football?

Taking data from the 2019/20 and 2020/21 seasons, we analyze the importance of this type of shooting situation, which means more than 10% of the total number of goals scored.


Corporate Information

We are a company based in Madrid founded in 2017 by Salvador Carmona and Cristian Coré Ramiro. Since our inception our work has focused on statistical analysis to help clubs in sports planning. We are a consulting firm that offers personalized services for each client and defends a mixed management model and constant communication to accompany the day to day of the institutions. Our strength is the widest coverage available in number of professional and junior tournaments. For more details please contact us.


Mentioned in: