Expected goals are a hot topic in the soccer community. What is expected goals and how can this stat be used to improve analysis? Read on to find out everything you need to know about the expected goals stat.
Expected goals glossary
Expected goals (xG) – the number of goals a team or player would be expected to score based on the quality and quantity of shots taken.
Expected goals per 90 (xG/90) – Expected goals per 90 minutes played by a specific player.
Non-penalty expected goals (npxG) – Total expected goals minus expected goals from penalties.
Expected goals for (xGf) - The number of goals a team is expected to have scored based on the quality and quantity of shots taken.
Expected goals against (xGa) - The number of goals a team is expected to have conceded based on the quality and quantity of shots they have taken.
Expected goals assisted (xA) – The number of assists a player is expected to have made based on the quality and quantity of the shots taken directly from their passes.
Expected points (xPts) – The number of points a team is expected to have won based on expected goals data.
What is expected goals? Expected goals explained
Expected goals is a metric which assesses the chance of a shot becoming a goal. It provides a good way to judge the quality of shots since a shot with a 0.4 expected goal (xG) value should be scored 40% of the time. An xG of 1 is the highest value a single shot can be, implying the player has a 100% chance of scoring.
How many factors are taken into account when calculating the xG of a shot depends upon the model.
-
Read: Expected goals models
What are non-penalty goal stats used for?
Non-penalty goal stats make comparing the goals and expected goals stats of players a more accurate reflection of their performance. Penalties have an xG of 0.76 so designated penalty takers will have their xG/90 substantially increased by any penalties they take.
Since they may not have earned the penalties themselves, including penalties taken in stats can distort the relative performance of such players when compared to non-penalty takers.
Why are expected goals useful?
Expected goals can be beneficial because they increase the sample size used when analysing soccer. Soccer is a low-scoring game and goals are a rare event. As a result, pure goals data can sometimes be misleading.
There are numerous examples every season where the team who created more chances ultimately loses the match. Basic goal data has trouble reflecting this and may be unrepresentative of the actual game as a result. Expected goals take chances into account by calculating the number of goals that, on average, are scored from each position.
Prior to expected goals, metrics like total shots or shots on target were used to attempt to analyse games. Like goals, these stats can be misleading. Total shots count an attempt from the halfway line as equal to a shot from inside the six-yard box.
During the 2014/15 Bundesliga season Borussia Monchengladbach’s Granit Xhaka and Mainz 05’s Yunus Malli took a similar number of shots per 90. Shots data would place the two players in the same bracket but expected goals can separate the two:
Xhaka vs Malli 2014/15 Bundesliga season
Despite taking more shots per 90 minutes played, Xhaka was unable to match Malli’s xG output. Malli scored 6 goals to Xhaka’s 0 despite playing fewer minutes. Malli primarily shot from close range with 70.7% of his attempts taken from inside the penalty area. In contrast, 64% of Xhaka’s were from outside the box.
Malli shot less than Xhaka but from better locations. Expected goals shows the higher value of these attempts and provides a quick and easy way to factor the players shot locations and types into the data. This allows for a better analysis of the difference between the two players.
What factors are used to calculate expected goals?
As discussed above, the location of the shot is a big part of a shot’s xG rating. However, as models have become more sophisticated, other factors have been taken into account in an attempt to make them more accurate.
Some models now factor in everything from the body part used to take the shot through to defensive positioning, attack speed and where the first possession of the attack started.
Further reading: Calculating expected goals
Which expected goals model is most accurate?
There is some debate as to which expected goals model is most accurate. Fortunately, Pinnacle have an article discussing the merits of different expected goals models.
Using expected goals as a predictor of future performance
A good example of how a basic goals analysis can be misleading is the August 2017 Premier League game between Arsenal and Stoke. The 1-1 result suggested the game was evenly matched. It is evident that using this draw as a predictor of future Premier League success would be problematic.
Here is how the game looks taking expected goals into account:
Arsenal vs Stoke xPts
Arsenal vs Stoke xPts
Using expected goals analysis we can see that, in the long-run, Arsenal could expect to win this game 55% of the time. This is much more useful for predicting future performance since we can minimise the effect of scoreline variance. Using this data it is clearer that Arsenal are likely to perform better than Stoke in the long-run.
Further reading: Using expected goals to predict the Premier League table
What are the limitations of expected goals analysis?
As with every soccer metric, expected goals cannot fully reflect reality 100% of the time. One common criticism is that since xG calculates the average shot taken it neglects to factor in the finishing skill of an elite striker or the reflexes of the world’s best goalkeepers. This is a fair criticism and definitely worth considering when analysing individual players with the ability of Harry Kane or David De Gea.
For more reading on the impact of elite ability on expected goals Marek Kwiatkowski discusses quantifying finishing skill here.
Harry Kane expected goals vs goals scored
Harry Kane expected goals vs goals scored
Some defensive systems also seem to be better at forcing attacks to underscore their expected goals than others. One such example is Sean Dyche’s Burnley side who regularly concede far below expectations due to a uniquely organised defence.
Equally, previous expected goals statistics are not always applicable when teams change style or personnel. When using expected goals to judge the ability of individual players it is also important to take their position and teammates into account.
It is no surprise, for example, that Lionel Messi’s xG numbers are greater than Sergio Busquets’. Both players bring different skillsets to the team. Likewise, a player who moves from a weaker team to a stronger one will often see an improvement in their expected goals numbers.
Further reading: Is one big chance better than multiple small chances
How to interpret expected goals for soccer betting
As discussed above, expected goals offer an incredibly useful metric for analysis but are not always 100% reflective of real situations. It is important that bettors factor in outliers like Burnley’s defence or Harry Kane’s shooting ability when analysing teams and players.
Expected goals in a betting context can help to identify outliers that may provide value in the future. A team that is over or underperforming their expected goals numbers in the short term are liable to regress to the mean. This could provide potential value if the market has not factored in the abnormal nature of the side’s short-run performance.