Why xG can’t always tell us how good a team really is

If you look at the expected goals (xG) for the Premier League results last weekend, you would probably be thinking the statistic is a load of rubbish.

Of the 10 matches, the xG score for each team matched the result in only three fixtures. For instance, Manchester United (2.27) lost 0-1 at home to Everton (0.16).

Yet xG is not going to take into account that Everton had a man sent off or that they played with their backs to the wall for most of the game.

It can tell us that United did not create many big chances from their 25 shots, Mason Mount (0.44) and Luke Shaw (0.38) being the best of them.

If the xG is so wildly out, is it time to stop giving it so much prominence? Not so fast, perhaps.

In week 11, the xG matched seven results, and in week 10 it had eight correct.

Since 2021-22, xG has an accuracy level of 59% and it is at 57.5% this season, which is a one, or two-result swing from being on trend.

Most importantly, xG is not trying to be a barometer of individual results, rather a metric which assesses every chance a team creates and how effective they are at converting them.

Aston Villa, Sunderland and Tottenham would be in the bottom four if their results were based on xG , yet in the real world actual goals have them in fourth, seventh and ninth respectively.

Villa have been exceptionally good at scoring from distance (more on this later), while for the other two teams goalkeepers Robin Roefs and Guglielmo Vicario have been excellent at keeping out high-quality chances from the opposition.

So, you might ask, what is the point of even measuring xG in a game?

Here’s a better way to think of it: If you leave a game saying “I can’t believe we lost that” or “if we keep playing like that we’ll be alright” then you will most likely have a higher xG than your goal return.

If you’re saying “I can’t believe we won that”, then it’s going to be a low xG.

The idea is, eventually, a team should start performing to their xG, good or bad. That cannot take into account various issues about general player and team form. Just look at Liverpool: on xG this season, they should be fifth rather than 12th.

So, what influences xG? And is it really a reliable way of assessing a team’s performance?