Football happens before the shot

Why Expected Goals Only Tell Part of the Attacking Story

In Friday night's Premier League game, Liverpool defeated Ipswich 2-0. However, a closer examination of the statistics reveals a different story, suggesting that Ipswich actually had an advantage. Throughout the match, Ipswich outshot Liverpool, recording 14 attempts compared to Liverpool's 10, as well as more shots from inside the six-yard box (2-0) and more blocked shots (4-2). Additionally, Ipswich achieved a higher expected goals (xG) total of 0.76, while Liverpool's xG was 0.69. This discrepancy between the final score and the underlying statistics raises an intriguing question for fans and analysts:

How could Ipswich create more scoring opportunities yet still end up on the losing side?

This situation highlights a common misunderstanding about expected goals: while they serve as a useful metric for assessing performance, they do not encompass the entire attacking process. Understanding this complexity is key to interpreting match outcomes effectively. 

To gain a clearer picture of the game's dynamics, we need to incorporate expected goals on target (xGOT) and consider the overall threat generated over the course of the match. By doing so, we can better understand the various factors that contribute to a team's success or failure. Ultimately, this brings us to an essential aspect of xG that we need to grasp.

"How can you score one-and-a-half goals?"

You can't.

This is one of the most common criticisms whenever expected-goals numbers appear in the media.

xG is a probability, not a predicted score.

If a particular shot has an xG value of 0.20, the model is essentially saying that shots with similar characteristics have historically resulted in a goal around 20% of the time.

Take five identical 0.20 xG shots and you wouldn't expect each shot somehow to produce one-fifth of a goal.

You might score none.

You might score one.

You might score three.

Across thousands of those shots, however, roughly 20% should become goals.

Match xG simply adds those individual probabilities together (as shown in figure 1 below). So if a team accumulate 1.5 xG, it doesn't mean the model thinks the team should literally score 1.5 goals. It means the collection of chances they generated carried an expected value equivalent to 1.5 goals. That distinction matters.