Amateur Numbers: Expected Goals (xG)
A new delight for Soccer Aficionados.
Early October 2019, Tottenham Hotspur were demolished by Bayern Munich in their new stadium. If they only lost the game by a goal or two, that would be somewhat acceptable. However, Spurs lost the game 2–7 to the visitors; such a fantastic fashion for a home defeat. If we look at the statistics, Spurs had 15 shots compared to Bayern’s 19. Even if we delve a bit further to the attempted shots figures, both teams were not so different. The home side managed eight of their attempts on target, while Die Roten only two more.
In reality, statistics don’t stop there. Opta recorded a quite similar figure in both teams’ expected goals tally; Tottenham with 1.94 compared to Bayern’s 1.34. Looking at those numbers, last year’s UEFA Champions League finalist should have scored more goals than the German team; not conceded five more often. But wait, what is the expected goals?
Expected goals (xG) is a statistical measure to calculate the quality of a goal-scoring attempt, as well as the likelihood of scoring. The usage of xG is getting more common in today’s soccer as quite often the result of a game doesn’t reflect what really happened on the pitch.
An xG metric uses the scale between 0 (zero) to 1 (one) per shot. The closer a shot’s xG to 1, the higher the probability of scoring; vice versa. For example, generally, a penalty shot is rated with 0.76 xG. Such value occurs from a historical reason; where 76% of the penalty shots taken have resulted in a goal.
There are lots of variables are used by the data-and-analytics companies to calculate xG in a soccer game. From passing types, game situations, parts of the body used to make a shot, even data-proven history can be used to count the quality of chances. In this piece, we are going to take a look at the two most common variables used in xG calculation.
How to calculate xG?
To be fair… We don’t really know.
There is no, one true way of calculating xG. Each statistics provider have their own xG model, but most of them account these important aspects other than the aforementioned above:
- The number of defenders around the attacker
- The distance of defenders
- The type of pass to the attacker
- The positioning of the goalkeeper
There are also different shot-taking situations that we must take into account, like:
- Shot attempts from open play
- Penalties
- Free-kicks (direct or indirect)
- Corners
- Rebounds (from a save by the goalkeeper, hitting the woodwork, etc.)
But along the way of our journey finding the true value of xG, we find one of the simplest model of calculating xG with:
xG=0.10×shots
Where in this case, all kinds of shots/attempts are regarded as an equal chance, since on average there is a 10% chance of scoring a single shot in soccer. This model does not actually represent the real situation. Because it values every shot in any kind of situation or area of the pitch as equal.
Two common aspects that are most used in calculating xG will be explained below.
Expected Goals by shot distance
Let’s take a look at the shot chart above. Logically, a close-range tap-in attempt is more likely to result in a goal rather than a speculative long-range shot. Hence, the shot made from point A has bigger xG value than the shot made from point B.
For example, let’s take a look at James Ward-Prowse’s goal against Manchester City back in November 2019. That time, Ederson went to save the ball but somehow the ball slipped from his hand to Ward-Prowse’s path. The Southampton midfielder then tapped the ball in from the goalmouth to score the first goal of the match. This very-close-range attempt was rewarded a sky-level 0.87 xG.
Now let’s move to mainland Europe. Precisely in France, when Stade Brestois defeated Toulouse away from home. In that mid-January match, Hiang’a Mbock scored a spectacular goal in the 79th minute. Mbock’s 35-yard strike is only rewarded with an xG score as low as 0.02. To simplify, that value shows such particular attempt made by any soccer player has only a 2% chance of seeing the back of the net.
Furthermore, a relatively high xG value doesn’t always results in a goal. For example, Brandon Williams’ miss when Manchester United defeated Norwich City four goals to nil in January. Williams’ one-yard-from-goal attempt is rewarded by a soaring-high 0.71 xG. Unfortunately, the youngster’s shot didn’t end up in goal; instead, it went into Row Z.
In ‘normal’ statistics, Williams’ effort is counted as a worthless missed shot; whereas, the reality shows the opposite. United should have added more to their goals tally from this huge opportunity. Luckily, the Red Devils were having a three-goal lead; making Williams’ miss somewhat acceptable. The advantage of translating the quality of a shot more accurately is the main reason why xG metric is more common in today’s soccer.
Expected Goals by shot angle
Another chart, another question. Which player will come up with a better xG value on his shot? By logic, an attempt made from an open-angle is more likely to end up in the goal rather than from narrow ones. Hence, the shot made from point Y in the image above has bigger xG value than the shot made from point X.
Let’s turn back time a little to February 2016. That time Mohamed Salah — then an AS Roma player — successfully made a goal from a narrow-angle against Palermo. Such goal was a bit similar to the one he scored against RB Salzburg for Liverpool in the UEFA Champions League last year. Salah’s attempt against Palermo was rewarded by 0.13 xG; equal to a relatively-low 13% goal-scoring chance.
On the other side, Son Heung-min’s winning goal against the Citizens two weeks ago was rewarded by an xG number of 0.30. This means, Son’s strike was 17% more likely to end up in the goal compared to Salah’s goal, even though the Korean made his shot from a further range. Son kicked the ball just inside the box, probably 17-yard from the goal. It is further than Salah’s shot, which was made from as far as 13-yard. Such significant xG disparity is resulted by both shots’ angle difference; Son’s from an open angle, while Salah made his shot from a narrow viewpoint.
Player and Team xG?
You might have come across statistics provider websites like understat or footballxg and noticed that the xG value is not in the form of percentage (0-to-1). What the heck is all this?
Let’s take a look at this table which shows the top ten goalscorers in the Premier League:
Jamie Vardy, Premier League’s current best goalscorer, have 17 goals under his belt with 12.94 xG. The xG shown comes from the accumulation of every xG value from the shot Vardy took up until today. Looking at the numbers, with all the shot he took he should’ve only scored 13 goals but in fact he has 17. It means that Vardy able to convert his chances into goals, even in difficult or unprecedented situations.
To get a better understanding on how shot distance and angle affect the value of xG in a match, let’s look at these examples below:
First, from the year-opener game when Arsenal defeated the Red Devils at their home. Both teams equally made 10 shots in the game; even a similar tally in shots on target with four each. However, Arsenal achieved a much higher xG value than the visitors. The Gunners got 1.85 xG value, yet United only 0.72; respectably justified the 2–0 scoreline. Can you guess how could this occur?
Looking at the graphic above we can see Arsenal (blue) attempted their shots more often from inside the 16-yard area. In total, they made eight shots from there and only two from outside the box. Hence, they were able to convert two of the penalty-area shots into their two goals (stars). Now, take a look at United’s (yellow) shot locations. The visitor made five attempts from a long-range distance; three more than Arsenal. This is the reason why United’s shot quality are lesser than the home side.
Next, let’s look at the numbers from Liverpool’s last win in Anfield against Southampton a few weeks ago:
Liverpool scored four goals with 3.66 xG value from a total of 16 shots. Oppositely, Southampton made 17 shots; one more than the home team. What did the Saints get? Only 0.72 xG value with … no goals. How did this happen? Let’s see both teams’ shot location chart:
From the chart above we can see that Liverpool (blue) made more shots from closer distance. They made 11 shots from inside the box; even better, from goalmouth area. Meanwhile, Southampton (yellow) made 11 shots from the 16-yard area, yet they took it from tighter angles. Furthermore, only one from the six-yard box. The Saints’ incapability of taking better shots; especially from better angle says about the low xG value from the match.
Nowadays, the usage of advanced statistics is getting more common as it helps us to analyze football games. The reason is that such type of data is more empirical and factual compared to bias-filled opinions from the fans. The presence of expected goals metric and its diverse variables can help all football aficionados to view the game more rationally.
Indeed, more time is needed for this relatively-new metric to be understood. But, if it allows us — the fans — to judge the game more wisely, why not?
Source: Cartilagefreecaptain, Bundesliga, Dribble9, Fantasyfootbalfix, Footballxg, Ligue 1, Opta, Serie A, Thelastlibero, Understat, Wyscout, Youtube
Written by Petrick Sinuraya and Ammarsha Rewindra Ridwan
