Simon Rolfes is Bayer 04 Leverkusen's Sporting Director, and between 2005 and 2015 he played 288 Bundesliga games and won 26 caps for Germany. In this weekly column, he will write about the new, innovative Bundesliga Match Facts powered by AWS that you can see during games.
It was on Matchday 28 last season when Bundesliga fans and TV viewers first met 'Expected Goals', or xG for short, and 'Goal Probability'. During games on TV and in live match blogs, the xG value was calculated, and the actual Goal Probability shown after one is scored. For a lot of viewers, as I soon noticed myself, it's unusual and — at first glance — also a little complicated.
What added value does xGoals give me?
In short, this measures the efficiency of a player or a team in his or their finishing. xGoals can, however, also reflect how a game is going and a team's chances of scoring.
Firstly, we need to distinguish between 'Goal Probability' and 'xGoals'. Goal Probability is actually the basis for xGoals. Every single attempt at goal in the Bundesliga is now given a percentage value. Many factors are taken into account in the calculation: what is the distance to goal? What is the angle for the shot? How many opponents are putting pressure on the player shooting? How is the goalkeeper positioned? How fast is the shooter running? Is he kicking or heading the ball? The quality of the player has no influence at all: whether the player shooting is Robert Lewandowski or an amateur footballer, it makes no difference to Goal Probability. Only when it comes to chance conversion does the individual quality of the player enter the equation.
If you think about it logically, a long-distance shot from an acute angle under pressure from defenders has a much lower Goal Probability than a tap-in into an empty net. The simplest example is a penalty, because it always occurs under the same conditions. The Goal Probability for a spot-kick is always 77 per cent: it's always from 12 yards, there's never any pressure from opponents, with a 37 degree shooting angle, and — based on past experience — 77 of 100 attempts end up in the net. This value is calculated by a 'Machine Learning Algorithm', a form of Artificial Intelligence, which bases its calculation on over 40,000 attempts at goal.
Bayern: by far the most clinical team in front of goal
Unlike Goal Probability, xGoals is not given as a percentage, but as a value between 0 and 1 — a penalty, for example, is given as 0.77. The xG values of a team or individual player are now added together. From this total, a statistical statement can be made of how many goals a team or a player — measured by their attempts at goal — should have scored. xG values can be calculated for teams and players within a game, but also for the whole season up to that point.
An example: FC Bayern München have an xG of 20.2 in the current season. But Hansi Flick's side have actually scored 34 goals, 14 more than would have been expected given their attempts at goal. This major discrepancy between xG and actual goals scored is by far the biggest in the Bundesliga. In other words, Bayern also score from seemingly impossible situations.
xGoals: Lewandowski's class has an impact here
This phenomenon can also be individual as well as collective. Lewandowski's 2020/21 xG is 6.8, meaning we could have expected nearly seven goals from him. But he's actually scored 12 times already. No wonder he's at the top of the scoring charts again. There's also our Leverkusen centre-forward Lucas Alario, who has himself come to the fore thanks to his clinical streak of late. With an xG of 2.8, he actually has seven goals, the second-best value league-wide in terms of efficiency in front of goal. But it doesn't surprise me, as I see every day in training what a machine Lucas is in front of goal.
Actually, at Bayer Leverkusen, we always look at our xG as it provides good feedback on how our attacking play is developing. Basically, it's good for us to have a high xG in comparison to our opponents, because it means we've taken a lot of chances. If we've been really clincal and scored more times than our xG, all the better! For fans and TV viewers, xG is a good reflection of how a game is going, something other stats such as possession, shots on goal, and duels won perhaps don't always do quite so well.
That fact that we at Bayer Leverkusen are currently second in the table is thanks in large part to our qualities in front of goal: 14.2 goals is our expected xG — we've scored 19. If we can maintain that sort of finishing, it's certainly going to be an exciting season.