Expected Goals (xG) - Explained

Hey, how has your Sunday been? Whether you've had a productive day, or a relaxing one, here's a little bit on expected goals inspired by a video I did on the same topic a few months ago.

Before I get into it, I will presume that you probably fall into 1 of 3 camps when it comes to xG.

1) You use it in your day-to-day work and pretty much know everything about it (in which case this week’s article may not be for you!)

2) Heard it and seen it banded around. Have an idea it is linked to probability of shots being scored but need a bit of knowledge top up (hopefully I can give you that!).

3) Have no clue what I am talking about. (You will after reading this!)

So, what is Expected Goals
In its simplest form, expected goals is effectively the probability that any given shot will result in a goal. This is then given as a decimal number, so a shot would have a score of between 0 and 1.

The closer the xG score is to 1, the more likelihood that particular shot has of ending up as a goal.

E.g., A player takes a shot that is a given an xG score of 0.32.

Therefore, this will (based on the model used - more about that later!) result in a goal 32% of the time. Make sense?

How to calculate Expected Goals
Okay, so you may be wondering how that number is actually calculated. Well, it's a great question. Going back to the example above of the 0.32 shot. What factors actually make this 0.32? Why isn't it 0.5, or 0.2?

It all depends on a number of factors based on previous shot data. Like anything in data analysis, the bigger the sample size, the (usually) more accurate things become. Therefore, to get these xG scores, models are created based on thousands of previous shots that we have collected data on.

Although you could create your own model, the more accurate models are normally produced by data companies to which professional clubs will subscribe to. The most basic of models will factor in the following.

• Distance to goal - As you can imagine, the closer to goal you are, the more likely you are to score.
• Angle to goal - Are you wide or central? I am sure you will agree that it is easier to score from the middle, than out wide.
• Body part used for the shot. Was it struck with the foot, or a header?
• Type of assist - As it says on the tin. How was the ball delivered to the shooter. Was it a through ball, cross, set piece, dribble etc.

Different data providers will have slightly different ways to calculate the end score of a shot. As mentioned, the more criteria you look at and consider, the more accurate you would expect the score to be.

Other factors to consider are the location of defenders (GK included), the location of other attackers, and the height of the ball.

How is xG used in professional football
Recruitment - This goes for both players and managers. Knowing the xG of a manager (or should I say his team), gives us an indication of the type of football he plays and the chances that his team typically creates. If this fits your club model, it could mean that manager could be a good fit for your team too.

As for players, it shows us the types of chances a player usually finds themselves in and how well they are converting chances. Could definitely be looked at and considered when looking at attacking players.

Predict performance - Although previous performance does not guarantee what will happen in the future, knowing xG of teams over a period of time and against specific opponents can help us try to predict results. You will find xG is used a lot in sports gambling.

Media - You've probably seen xG on the like of Match of the Day and Sky Sports as it becomes more mainstream.

What have we learnt so far?
A few things to note that we have learnt so far from historical shot data include:

• Central shots are better (quite obvious really!)
• Shots with feet convert better than headers.
• Crosses are hard to convert!
• Shot quality is a better indicator of a player’s skill than their finishing ability (more on this below)

Over a large enough sample, most strikers will have a very similar score when it comes to xG. They may have seasons where they run hot or others when they have a bad spell, but over time, things will always even out, and we find that (unless you are a player like Messi) you won’t deviate too far from the average over a long period of time.

With this in mind, it is the type of chances that a player takes that is a better indicator or performance. For example, a better player may have better movement and be able to navigate themselves into better scoring positions. Over time, if you are constantly finding yourself in better shooting positions, you will score more goals.

Still need a bit more information? No problem! Check out this video I did on this exact topic...

Building on Expected Goals (or xG), you also have things like npxG (non-penalty expected goals, and xA (expected assists). Lots of fun to be had!

📧 If you enjoyed this post, why not consider subscribing to my free weekly newsletter where I share more tips and industry insights direct to your inbox. Just enter your email address in the field below.

📺 I also have a load of useful videos over on my YouTube channel including a podcast where I interview industry professionals to share more behind the scenes knowledge, advice and ideas.

📝 If you are wanting to improve your CV for a job in sport, you can grab my free template HERE.

💻 If you are a coach or analyst on the look-out for video analysis software (that is actually affordable!) then I got you… check THIS out.