xG’s: Why it is flawed and pointless

Over the last few days I have had a few debates with people as to why the “expected goals” (xG) statistic is pointless. It seems there are plenty who defend it.

To work out a team’s xG for a match, every shot must be analysed and given an “Expected goal value” (EGV).

EGV is the probability that any given shot will end up as a goal.

As Patrick Lucey, director of data science at STATS, explains, EGV is based on a number of factors, such as where the shot was taken from, the proximity of defenders, the nature of the attack (i.e a direct free-kick or a penalty). The EGV of a shot assumes it is being taken by someone of average ability in the league, so it expects for instance that a shot from 10 yards out plum in front of goal with no defenders nearby has a high chance of ending up as a goal.

People then use this statistic to highlight sides who are over performing or under performing.

One journalist from the Mirror recently wrote an article stating that Arsenal should only actually have scored 11 goals this season, not the 19 we have actually scored. And should be mid table.

Some outlets are actually printing a weekly xG table in an attempt to highlight where sides will be based on if shots that are expected to go in did, and vice versa. Arsenal are currently 11th in the league based on expected points.

The glaring issue with xG it is that it assumes every player is born equally.

The xG value of shot assumes it is being taken by someone of average ability in the league.

It puts the same value on a chance going in regardless of if it is Alexandre Lacazette taking a shot or Yaya Sanogo.

And it assumes the same with a goal keeper.

It does not account for David de Gea being a better goal keeper than Manuel Almunia . It assumes that a shot is just as likely to beat both.

So Sanogo against de Gea is given the exact same value as Lacazette against Almunia.

Football is all about exceptional talents. It is what separates a £50m striker or a £80m goal keeper from one that costs £5m. It is why Manchester City top the league and Newcastle are bottom. Why Sergio Aguero outscores Joselu.

Lionel Messi has “outperformed” his xG prediction throughout his career. That is because he isn’t average. He is the greatest player of his generation. He should not be compared to the average within the league.

As soon as you start disregarding individual talent and assume that everyone is the same, your analysis becomes flawed.

So Arsenal are performing above what their xG believes they should have scored.

Perhaps the truth is that in Alexandre Lacazette and Pierre-Emerick Aubameyang; we just have above average strikers?

Ultimately, xG is pointless. It means nothing. It is a spreadsheet which, after a game, “predicts” what should have happened based on the law of averages. Not actually what happened.

And it is what happened that is ultimately important.

A world class goal keeping performance, a keeper letting the ball through his legs, a striker missing a tap in or scoring a worldie. A beach ball entering the field of play. These are all actual events. Events an algorithm can not predict.

If you wish to give weight to xG’s, that is your prerogative. But I am going to live in a world of actual goals. Actual results. A world that rewards the exceptional talents over the average.

xG is pointless.


5 thoughts on “xG’s: Why it is flawed and pointless

  1. Pony Eye

    Your logic is correct but you’ve done away with the fact that a lot of the joy of football comes from speculations using all sorts of parameters. xG happens to be one of the most accurate parameters. It’s another instrument for self entertainment.

    However your arguement is a basis for creating the xG for each player against each keeper for each situation assuming that is possible. It would make the xG even more accurate but never the reality. The reality is that Lacazette or de Gea this weakened can be a far cry from Lacazette or de Gea next week end. If football becomes simple mathematics it would lose its appeal.

    Bottom line is most of us like to take trips away from reality and back to it.


    1. keenosafc Post author

      Completely agree with your middle paragraph. Likewise a chance in the 8th minute is not the same as a chance in 118th minute. No two situations are ever the same, no two players are the same. So creating a algorithm which assuming things are the same is flawed.


  2. Peter

    xG is not pointless. The problem is most people don’t understand what it’s telling them. It’s not really about goals at all, but about a player’s/team’s ability to get into good scoring positions. Or their lack of ability to do so. It has some minor predictive use: for example, if we see what we assume is a poor shooting team, say Norwich, outscoring their xG, we can guess they are on a hot streak that will cool soon enough. But, of course, that requires an assumption, and also kinda tells us what we already knew: that they are playing over their heads.



Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.