Stats Perform’s expected goals on target (xGOT) builds upon the original expected goals model (xG) by crediting on-target shots based on a combination of their underlying chance quality and the quality of their execution.

With Stats Perform’s expected goals (xG) model we can measure the quality of a chance and the likelihood of it becoming a goal, but we know that players can score more or less from these chances. So, how do measure what happens next?

A chance may be valued at 0.1 xG, where we’d expect an average player to score one in ten times, but players can execute this chance very differently. A shot from this chance that is heading into the top corner is far more likely to result in a goal than a shot that is hit straight down the middle of the goal. This is where Stats Perform’s expected goals on target model (or xGOT) can measure that next level of context.

What are Expected Goals on Target?

Expected goals on target (or xGOT) measures the likelihood of an on-target shot resulting in a goal, based on the combination of the underlying chance quality (xG) and the end location of the shot within the goalmouth. It gives more credit to shots that end up in the corners than shots that go straight down the middle of the goal.

xGOT is measured on a scale between zero and one, where zero represents a shot that will never result in a goal and one represents a shot that is expected to be scored every single time. This model is only for on-target shots given that, if you don’t get your shot on target, there’s a 0% chance that it will result in a goal.

Let’s see how this looks for Manchester City’s Raheem Sterling’s freekick against Crystal Palace at the start of 2021. The pre-shot chance quality measure is relatively low here (0.1 xG) but, in our post-shot xGOT model, we will also account for the finishing location of his shot in the goal.

Raheem Sterling freekick xG
Manchester City’s Raheem Sterling’s free kick against Crystal Palace was worth 0.1 xG

Sterling’s shot was heading into the top right corner of the goal, making it incredibly hard to save for Crystal Palace’s Vicente Guaita. Ultimately it did prove to be too hard to save and, despite the difficulty of the chance, the high quality of Sterling’s placement is reflected in the post-shot xGOT model (0.81 xGOT).

Raheem Sterling free kick placement xGOT

How Are Expected Goals on Target Calculated?

Stats Perform’s expected goals on target (xGOT) model is calculated using a logistic regression model. It is built on hundreds of thousands of on-target shots from our historical Opta data and includes both the original xG of the shot and the goalmouth location of where the shot ended up.

The coordinates are taken at the point at which the shot was expected to cross the goal line to determine the goalmouth location of the shot and the interaction effect between the visible angle of the goal and the point that it crossed the line.

How Can We Use Expected Goals on Target?

Quantifying Finishing Ability Using xGOT

Now that we can value the execution of a player’s shots, we can see how this shot placement compares to the underlying quality of their chances. A player whose xGOT is exceeding their xG is executing better quality shots than the quality of the chances he has attempted shots from. We call this difference between xGOT and xG, shooting goals added (or SGA).

There is an important caveat to using xGOT for attackers in this way though. Given that you receive an xGOT value of zero for blocked shots, the outcome of a shot can be heavily influenced by a defender’s intervention. A shot may be destined for the top corner, but the defender may block it and so we cannot isolate the outcome only to the shooter. While it still useful context to understand, shooting goals added is rather a measure of on-target execution than purely being indicative of a player’s finishing ability.

Player NameExpected GoalsExpected Goals on TargetShooting Goals AddedGoals
Harry Kane10.914.5+3.516
Pierre-Emerick Aubameyang13.216.1+2.920
Alexandre Lacazette8.611.3+2.710
Mason Greenwood2.54.7+2.36
Jonjo Shelvey2.95.2+2.310
Premier League 2019-20 shooting metrics (excluding penalties)

One of the most notable performers in the Premier League 2019-20 season for shooting goals added was Tottenham’s Harry Kane, whose sixteen goals exceeded the underlying quality of his chances (10.9 xG). xGOT enables us to credit some of this overperformance to the quality of his shot placement (14.5 xGOT). The placement of Kane’s shots improved the quality of his pre-shot chances by 3.5 goals, more than anyone else in the Premier League last season. This indicates that, from his given chances, he is both getting his shots on target and hitting them in good locations.

Measuring Goalkeeper Performance Using xGOT

Evaluating goalkeeping performances using traditional metrics like clean sheets or save percentage have their limitations as these numbers can be heavily biased by team and defensive strengths. If you’re in goal behind the best defence in the world, you would expect to perform highly in these metrics regardless of your abilities as a goalkeeper.

Given that xGOT measures the probability of shots on target resulting in goals, the only factor preventing them being scored is the goalkeeper. A shot that is worth 0.3 xGOT has a 30% likelihood of being scored but it is also a shot that has a 70% chance of being prevented by the goalkeeper.

This means that we can predict how many goals a goalkeeper would be expected to concede, based on the quality of the shots that they faced. It allows us to directly credit to goalkeepers for their ability to prevent goals, irrespective of their team’s defensive strengths.

During the Premier League 2019-20 season, Sheffield United’s Dean Henderson was one of the standout performers. Based on the quality of the shots on target that the Manchester United loanee faced (39.4 xGOT), the average goalkeeper would have been expected to concede over 39 goals. Given that he only conceded 32 goals (excluding penalties and own goals), we can credit the young English goalkeeper with preventing more than 7 goals with his saves. This was the fourth highest in the Premier League that season.

Dean Henderson goals prevented xGOT

In terms of his traditional metrics Dean Henderson had a remarkably similar season to his incumbent at his parent club, David de Gea. Both goalkeepers kept the same number of clean sheets (13), conceded the same number of goals (32) and David de Gea (128) only faced two more shots than Henderson (126). The notable difference was that Henderson faced higher quality shots (39.4 xGOT) than de Gea (33.0 xGOT). Consequently, Henderson actually prevented more goals for Sheffield United (7.4) than David de Gea did for Manchester United (1.0).

Dean Henderson David de Gea expected goals on target comparison

While goals prevented is an intuitive measure of goalkeeper performance, the inevitable rebuttal here is that goalkeepers who face more shots have the opportunity to ‘prevent’ more goals. To allow for a fair comparison, we can standardise for the number of shots each keeper faced by looking at their goals prevented rate. Goals prevented rate is the number of goals that a goalkeeper was expected to concede as a proportion of the number of goals they actually conceded.

For example, Sheffield United’s Dean Henderson and Crystal Palaces’ Vicente Guaita both had the same goals prevented rate (1.23) in 2019-20, despite The Eagles’ goalkeeper ‘preventing’ more goals (9.4). Normalising for the volume of shots allows us to see that both goalkeepers were expected to concede 1.23 goals for every goal that they actually conceded.

Measuring Shot Placement and Prevention

With the context of Stats Performs xG and xGOT models, we are now able to analyse shots in far more detail. xG measures the quality of the chances that a side creates while xGOT builds on this to tells us what a team actually managed to do with these chances. Both models measure chance quality but while xG only takes into account factors before the shot was taken, xGOT is a post-shot measure.

Using expected goals on target, we can isolate and evaluate goalkeeper performances independently of their team in a way that traditional metrics can’t. It enables analysts and pundits to distinguish between the goalkeepers who are making high quality saves and those whose save counts may have been inflated by easier, low quality shots. While the frequency and quality of shots that a goalkeeper faces are still determined by their teammates, we can now see who is the most effective at preventing these from becoming goals.