Sean Dyche keeps referring to it. Everton keep winning on it. But what exactly is xG?
Football is fundamentally a results-driven business, and when all is said and done, it’s goals for and against that matter most.
Yet relatively speaking, football is a low-scoring sport. On average, Premier League matches last season produced an average of 2.85 goals per game. This means variance can often play a decisive role in results. A good set piece, a bad set piece, a wonder goal, an own goal, precise finishing, inaccurate finishing… numerous factors can influence the final scoreline.
So how do you analyse beyond the result to better capture the qualify of a performance?
Underlying metrics in football have steadily increased in value over recent decades with the purpose of doing just that. Measuring factors such as possession, passes, shots and much more have all become standard practice within clubs, and a lot of that information is also freely used and accessible in mainstream media.
Even so, for so long counting shots for along with shots against has remained the primary way to measure whether a team deserved more or fewer goals than what they scored.
But we know from watching football regularly that all shots are not the same – far from it. That’s why the Expected Goals (xG) metric was created. 
There are many different xG models out there, and each utilises different characteristics. But the bulk usually take into account similar traits: which part of the body a shot is taken with (it is often easier to score with your foot than head), distance to goal, angle to goal and the type of assist (cross in the air, a through ball etc). Some also add key information such as the position of the opponent’s goalkeepers and defenders to give a more accurate picture of chance quality.
Utilising historical shot data taken from hundreds of thousands of attempts, the effort is compared against others with similar characteristics and then attributed a value of between 0 and 1 to indicate the probability of it producing a goal.
To give you an idea of how the values relate to real-life recent Everton goals, Dominic Calvert-Lewin’s close-range strike against Luton Town was assigned a value of 0.98 by Opta and would be expected to be scored around 96 in 100 times.
Meanwhile, Jack Harrison’s stunner against Bournemouth was assigned a value of just 0.02, an effort that would have been predicted to lead to a goal on just two of 100 occasions.
It’s rare to see an exact marrying up between a team’s xG and goals, because when it comes to goals, a shot is either converted, or it isn’t, while xG values are assigned a probabilistic value between 0 and 100.
Yet a study from Lars Maurath, shared via football data specialist Statsbomb, suggested that between 79 per cent and 93 per cent of teams should expect to see their goals and xG roughly match. In short, it’s not an exact science, but it should be fairly close on most occasions.
Therefore right now, it’s fair to say that xG models are the most accurate resource we have to assess the underlying quality of performances from teams and players.
With that in mind, let’s look at the Toffees this season.
According to Statsbomb’s data, after eight games, the Blues have created the second highest xG in the Premier League.
A large portion of that xG (6.54) has come from set pieces, which illustrates Everton’s efficiency in terms of creating from dead-ball opportunities.
But there are also only five sides in the division who have created more shots from in-play passes (as per Opta), illustrating the Blues have been well-rounded in terms of their chance creation. Dyche’s side also rank second in terms of overall shots attempted this season.
At the other end of the pitch, there are only five sides that have accumulated a lower xG-against than Dyche’s side, and again only five who have faced fewer shots.
Combining the two to work out Expected Points (xP) – how many points a team could have expected to have won based on the scoring opportunities that they both created and conceded – Everton rank seventh in the division this season (as per Understat).
Results have not followed suit in this regard so far, in part as a result of those aforementioned variables that can often have such a sizeable impact on the result of a match.
Yet Everton’s performances, for the most part, have been trending in an upward trajectory for most of the season, and it felt inevitable that Dyche’s side would sooner rather than later enjoy a dominant victory both in terms of display, and indeed scoreline.
That came this past weekend when Everton registered an xG over 2.0 for the fourth time this season – having done so just twice across the whole of last season – en route to their impressive 3-0 success over Bournemouth.
While we know it’s goals, not xG, that win football matches, continuing to dominate in the latter is a positive sign for the Toffees moving forward this season, and a clear indicator of the progress made under Dyche.
Leave a Reply