‘xG’ changing the way we analyse football?

5 min readMar 31, 2021

‘Expected goals’, more commonly known as ‘xG’ is a recently developed method for estimating the quality of chances created or conceded by a football team. The model uses a wide range of data and statistics collected from over hundreds of thousands of shots to put a numerical value on the quality of chances created or conceded. It has revolutionised the way football games can be analysed from a fan’s perspective, putting an end to the common subjective comments such as “He should’ve scored that” or “He’s not a good finisher”.

Personally, as a big football fan, when I first saw this method being introduced a few years ago my initial thoughts towards the concept were negative. I thought that football was a game based on opinions and that introducing data and statistics would only take away from the uniqueness of the sport.

However, a few years down the line, having seen the upsurge of data being used in football, and the necessity of statistical analysis in all sports. I can now safely say that I have seen the light.

So how is xG calculated?

Every chance or shot taken is compared with thousands of similar shots, to determine the probability that a shot will be scored. An xG with a probability of 0 is a certain miss, while an xG of 1 would be a certain goal. The model uses the characteristics of the shot taken and the events proceeding it to calculate this figure by comparing it with thousands of shots with similar characteristics. Some of these characteristics are:

· Angle of the shot

· Distance of the shot

· What part of the body was the shot taken with?

· Were there any defenders close by?

· Type of attack

· Type of pass proceeding the shot

Origin

In 2012, data analyst Sam Green from the sports statistics company Opta, first developed the method. He took inspiration from similar models being used in American sports to develop the model using data collected from the premier league.

However, it was not until 2017 until the model began to become of any relevance to football fans, when the model was used on BBC’s Match of the Day, which instantly made the model a relevant topic of conversation.

Since then, the model has gained lots of attention in the analytics industry, with lots of different analysts tweaking the model ever so slightly to try and perfect it. This means there is not strictly one model that xG follows, however the difference between the models is very miniscule.

How xG is used to analyse teams and individual players.

Team

· Comparing a team’s xG with the actual number of goals scored can indicate the quality of the team’s shooting ability. For example, if a team’s xG for a game was 3.72 goals, but they only actually scored 1 goal in the game, would suggest that their shooting was particularly bad in this game.

· Similarly, the defensive ability of a team can be assessed by comparing the oppositions xG with actual goals conceded. This can give an indication of how effective a team are at preventing the opposition from scoring.

This data can become especially helpful over the course of a season or sustained periods of time. As more data is collected, patterns begin to emerge. If a teams actual goals scored is lower than it’s xG for just one game, then this may be attributed to fluke or bad luck. However, if this is the case for the majority of games over the course of a season it is more than likely not just down to luck. Being able to identify problems like this through data would be of obvious huge benefit to the management or coaches, as they could make the necessary changes to fix this.

Players

· Comparing a player’s xG with actual goals scored can give an indication of a players shooting ability.

The best players usually score more than their xG. For example, Robert Lewandowski of Bayern Munich, who is known for being arguably one of the best strikers in the world, has scored 35 goals so far in the 2020/2021 season, yet he has an xG of just 26.27 goals.

Comparably, Roberto Firmino of Liverpool, who is having what some would say as a bad season in front of goal, has scored just 6 goals despite having an xG of 10.52

This model can be a great method for fans to analyse games and put an end to many old subjective arguments that cannot be backed up. However, there numerous limitations to the model which may be why the model is not yet traditionally used by football managers, coaches or pundits.

Limitations

As the model is based on averages, it is almost impossible to exactly calculate the probability of a shot being scored due to the rarity and uniqueness of each game of football that is played and subsequently the shots that are being taken.

The model has also faced a lot of criticism as many hardcore fans believe that the use of data to analyse the sport takes away from the beauty that football holds. This may be true and may be a strong reason why the xG is only typically used by certain fans currently. However, it is obvious that football’s data revolution has begun, and I would not be surprised at all to see the xG model gain even more popularity over the coming years.

~ Ted.

Bibliography

- Arastey, G. and Arastey, G., 2021. What are Expected Goals (xG)? | Sport Performance Analysis. [online] Sport Performance Analysis. Available at: <https://www.sportperformanceanalysis.com/article/what-are-expected-goals-xg> [Accessed 31 March 2021].

- Driblab.com. 2021. [online] Available at: <https://www.driblab.com/analysis-team/expected-goals-xg-what-it-is-and-how-it-works/> [Accessed 31 March 2021].

- FBref.com. 2021. xG Explained | FBref.com. [online] Available at: <https://fbref.com/en/expected-goals-model-explained/> [Accessed 31 March 2021].

‘xG’ changing the way we analyse football?

Written by Team Scotland