In this piece, the following will be explained:
- What Expected Goals is, its positives, and its negatives.
- Which Newcastle players are performing better or worse than “expected”.
- What Expected Goals tells us about Newcastle’s attack, defense, and form during this campaign.
What is Expected Goals?
Tifo Football (a wonderful YouTube channel that visualizes myriad football concepts) made an excellent video that tidily summaries Expected Goals.
If you can’t watch that video, the gist is this: expected goals (or ‘xG’) is the probability a given shot will end up in goal.
The way it’s calculated is a based on a couple of variables: where the shot was taken, the proximity of defenders, and the type of attack (penalty, free-kick, open-play, etc.).
xG ranges from 0 (attempting a shot from the parking lot) to 1 (attempting a shot from inside the goal). The highest actual possible number is 0.99, as there is no such thing as a certain goal (as any Newcastle fan can testify).
If a player scores on a shot with an xG of 0.04, they have ferociously outperformed the historical expectation. And if they miss on a shot with an xG of 0.95, then they have fiercely underperformed.
What Does xG Do?
All shots are equal in the statistic ‘shots on target’, and yet we know that not all shots are equal. By being able to give different weights to every attempted shot, more context can be given to attacking performances.
For example, in the 24th Feb 2018 Bournemouth-Newcastle game, Ayoze Perez passed a ball underneath Asmir Begovic to Dwight Gayle, who one-timed it into the back of the net. The xG for that shot was 0.94, which means that given the location (roughly a meter away from the goal line, and roughly centered) and that it was Gayle’s dominant foot, the chances of the goal going in happened 94/100 times in the past. If Gayle missed that, it would be hugely frustrating.
On the other end of the spectrum, in the 63rd minute of that same game, Matt Ritchie attempted a shot outside the 18-yard box on the right side of the goal, with his left foot. The xG rating for that was 0.02, meaning that given the conditions of that shot, historically that shot finds the back of the net 2 out of every 100 attempts.
What Does It NOT Do?
It only tells us about the quality of the shot opportunity itself; it’s not predictive.
Also, xG won’t tell us if the result of a shot attempt should be attributed to exceptional technique, exceptional goalkeeping, luck, weather conditions, how long its been since “Blaydon Races” was sung, and other variables.
In other words, xG tells has very a narrow and specific utility, in terms of football analytics.
How Can xG Be Applied?
xG values are determined for every single shot taken. Add up all those values for a player, then subtract how many goals they scored. That number, which we’ll call “Net xG”, can inform how clinical or poor an attacker was.
For example, take Mo Salah. His Net xG this season was 21.79, and he scored 29 goals (as of 21st April 2018).
In other words, the historical average suggests he was expected to score about 22 goals but, in reality, he scored 29. This means he has a Net xG of -7.21. (Negative Net xG values are a good thing.)
On the other side of the coin is Christian Benteke – he has 2 goals against an xG of 9.50. This means his Net xG is +7.50.
Again: If a Net xG value is negative (Salah’s -7.21), this is desirable. If it is positive (Benteke’s 7.50), this is undesirable.
What does Net xG say about Newcastle players?
To keep things simple, center backs, full backs, and goalkeepers were omitted from this study.
All other players were separated into three groups: strikers, wide attackers/midfielders, and attacking/central midfielders.
Also, the minimum appearances for this study were set at 16, since we want to exclude small sample sizes.
Unsurprising to those that have watched Newcastle this season, our players are not clinical.
What does ‘Expected Goals’ say about the Newcastle attack?
Talk to any fan of any club, odds are that they’ll claim their team would be higher up the table if they’d only “finish their chances”. Fortunately (and unfortunately), Newcastle fans have data to back these claims:
Newcastle United have the 3rd lowest Net xG ratio in the league. If they merely finished at the average, expected rate, they would’ve added 5 more goals to their season tally (through 32 games as of this writing).
Could definitely be worse – Crystal Palace is 14 goals below the historical average for their scoring chances this year.
What does ‘Expected Goals Allowed’ say about the Newcastle defense?
‘Expected Goals Allowed’ is the inverse of Expected Goals. The sum of all expected goals is added up, and that number is assigned to the defense that allowed those goals.
Much like how Net xG was formed, a Net xGA can be calculated by removing the ‘goals allowed’ from the ‘expected goals allowed’ sum (called xGA).
For example, when Manchester United played Newcastle on 11 Feb 2018, their total xG was 1.72, and yet they scored 0 goals.
So, to calculate that Newcastle’s Net xGA, the xG of Man U (1.72) minus the actual goals they scored (0) results in a Net xGA of 1.72.
For Man United’s Net xGA, take the xG of Newcastle (0.68) minus the goal scored (1), which results in a Net xGA of -0.32.
Note: for Net xGA, positive values are desirable and negatives values are undesirable.
Using ‘Expected Goal Difference’ to Examine Newcastle’s Season Form
There is a lot to unpack here:
- Dummett’s return was significant for the team’s form. The only games after his return where the Net xGD was negative was the following: at City, home to Man United, away to Palace, and away to Liverpool.
- Lascelles’ injury was absolutely devastating to the team’s form. The only point won in his absence was the Leicester draw.
- Dubravka’s arrival has had a dramatically positive impact on the team. Since his debut, only 5 points have been dropped, and the Net xGD has been consistently positive.
xG is far from a robust, well-rounded stat. And as with any stat used in isolation, it can create a skewed picture of reality. When attackers fail to bury “easy chances”, xG can’t provide all the answers.
However, if its limitations are understood, it can help create a context for both attacking and defensive performances as well as trends in a club’s performance over time. Its strength lies in the “long view” rather than isolated, individual moments.
All xG stats were sourced from understats.com. All visualizations were generated in Tableau Public, after a bit of Excel tomfoolery.