What are Expected Goals (xG)?

By Footovision  |  April 6, 2021

Sum of Expected Goals of each team during a match, linked to the live score

The Expected Goals metric measures the quality of a shot based on several variables. The nature of the assist, shot angle and distance from the goal, and whether it was a header, volley, or first-time finish all determine whether a shot is deemed a 'big chance'. By adding up a player or team's expected goals, it gives us an indication of how many goals a player or team should score on average, compared to how many goals they actually scored. In football, many more shots and chances are created than goals scored. Therefore, a goal is a rare event (an average of 2.5 goals are scored per game), the number of goals scored in the game does not necessarily provide us with enough data to properly assess who deserved to win.

This brings the xG model into the picture. This model emerged in the USA in the 1990s through sports like baseball, basketball and American football. The characteristics which determine an xG, for instance, are the angle, distance, or power behind the shot. Whilst xG is a calculated average chance of a shot being scored, a team or player may outperform or underperform their xG value! Players can score from a chance that on average a player would be expected to miss, or vice-versa.

Representation of a half of the field divided into different zones. Within each zone, the percentage of successful shots and the number of expected goals generated are indicated.

How does this model work?

By exporting all shot data that we have in our database (mostly from top European leagues), we can frame each shot in the context of additional information like the position of the shooter, goalkeeper, and which body part was used to score the goal. At Footovision, we then build the xG model translating all this data into Python. Python is a prominent level programming language which, when combined with an algorithm which uses machine learning (the 'Neural Network') creates an accurate and complete account of the number of expected goals. The xG model is often used to analyze various scenarios and is crucial for:

Ranking of players according to the number of expected goals generated by eachduring a match. Also shown: offensive dangerousness (expected goals/shots), number of shots, number of shots on target, number of shots outside and number of goals for each player.

What once was a technical term used by analysts has now become a global metric, integral to many broadcasters' live feeds. It is increasingly common to see 'xG' below 'possession', 'shots on target' etc. in the list of statistics at half time. The template is growing too: don't be surprised if pundits and commentators start talking about xA (expected assists) or xT (expected threat) in the near future!

Related Readings