What is expected goals (xG)
The metric of expected goals first appeared in sports newspapers in America after a man named Brian McDonald made a presentation at a sports conference in 2012, and expected goals is data collected over 4 years from games in their NHL. i.e. National Hockey League.
It has now been exactly 10 years since expected goals have been a staple of modern sport. All teams at the highest level in soccer, and not just football but many other sports, invariably use this analytics data. I say analytical data, not statistics, for a reason. Don’t confuse the terms, as there are still many people who consider expected goals to be a statistic, and this has nothing to do with statistics. Rather, it could be considered quantitative analysis.
If you watch English football, very often in post-match interviews of some coaches and managers you will hear them commenting on expected goals and how they analyse them. However, the metric also suffers a lot of criticism. More generally, it is a clash of the traditional way of looking at sports games and on the other hand the total modernization and imposition of science in the world of sports.
Essence
But what exactly are the expected goals? Although first used in hockey, in this video we will mainly talk about how it applies in football. Expected goals (or xG) measures the quality of a chance by calculating the probability that it will be scored from a certain position on the field during a certain phase of the game. This value is based on several factors from before the shot was taken. The xG is measured on a scale between zero and one, where zero represents a chance that is impossible to score and one represents a chance that the player would be expected to score every time.
We know that a goal attempt from midfield is not as likely to result in a goal as a chance from goal. With xG, we can actually quantify how likely a player is to score from each of these situations. Suppose, for example, that the chance from the small penalty area with a given set of pre-shot characteristics is worth 0.1 xG. This means that on average a player is expected to score one goal out of every ten shots in this situation, or 10% of the time.
The terminology may be new, but these phrases have been used by football fans and commentators for years before xG was introduced. For example “he scored that nine times out of ten” or “he should have had a hat-trick already”.
Not understanding the concept
The main criticisms of xG often occur in scenarios where the metric is not actually applied correctly. The most common is at the game level. The team that has the higher xG in a game doesn’t necessarily mean they should have won the game. The xG only measures the quality of the chance, not the expected outcome of the game. Just as the old adage suggests, goals change games and the result affects the teams play. If a team takes an early lead, they don’t necessarily “need” to generate more scoring chances and we often expect to see the opponent make more scoring attempts the rest of the game in pursuit of a better result.
Another misconception is in the literal interpretation of the name of the indicator. No one expects goals to be exactly as predicted by probability. Some of the goals may not be scored. The name “expected goals” is derived from the mathematical concept of “expected value” and is a measure of the probability of achieving a result. The expected value of a fair coin flip is a 50% probability of coming up on the tongue and a 50% probability of coming up on the turn. But realistically, by flipping the tongue and the tour, you don’t expect the tongue and the tour to fall exactly half the time, but rather that with a larger number of coin flips, the statistics are likely to stick to that balance. The same goes for expected goals. Deviation from the expected value is inevitable and this is valuable information that we can analyse in football.
A player or team that has outperformed their xG does not need to underperform to return to expectations. This is a concept known as the gambler’s fallacy. Although we would expect them to get back to scoring goals in line with expectations with their future shots, they have already “built up” that superior performance and therefore no one has said they won’t just up their xG from here on out rather than lower their goals. Similarly, if a coin flip lands on Eazy ten times in a row, future coin flips are still equally likely to be Eazy and Tura, and it doesn’t matter that it’s been Eazy 10 times in a row so far.
How to calculate xG
While watching a match, we can intuitively tell which chances were more or less likely to be scored. How close was the striker to scoring? Did he try to shoot from a good angle? Was it one-on-one? Was it a header?
The problem is that there is an average of 25 shots per game for which you have to work out the metric against all the unique situations. The advantage of this model of expected goals is that they can now take the variables to quantify how each affects the likelihood of scoring a goal. This allows them to assess the quality of chances for all 9398 shots taken in the 2019-20 Premier League season in seconds.
Stats Perform’s xG model, for example, is built using a logistic regression model that is fed from hundreds of thousands of shots from Opta’s historical data and incorporates a number of variables that affect the likelihood of scoring a goal, some of the most important of which are:
- Distance to goal
- Angle to goal
- One-on-one
- High chance of goal
- Body part for the shot (e.g. head or foot)
- Type of assist (e.g. forward pass with ball, centering, feint, etc.).
- Pattern of play (e.g. open play, fast break, direct free kick, corner kick, tap in, etc.)
No doubt some situations are particularly unique and are therefore modelled independently of others. Penalties are given a constant value corresponding to their overall conversion rate (0.79 xG); direct free kicks have their own pattern; and header chances are valued differently to those from static positions and those from open play.
Since the start of the 2017-18 season, Stats Perform’s detailed event data includes qualifications for opponent pressure on the ball and visibility of the shot, which explicitly measure pressure and positioning of defenders and the goalkeeper. These will be a prerequisite for upcoming improvements in new versions of the model.
How to use xG
Suppose two players take exactly 100 shots in a season (excluding penalties), but score 14 and 8 goals respectively.
By quantifying the quality of the 100 chances for each player, xG adds additional context to their shots that goes beyond traditional metrics such as shots on goal or average distance per shot. That is, it can measure the quality of the situations each player has.
From the scoring chances of the first player, it was expected to score 18 goals (17.7 xG). On the other hand, from the 2nd player’s chances, we would expect him to score only 7 goals (7.0 xG). We can immediately see why their goal totals are so different. Although the 1st player had a super performance and the 2nd player not so much, the quality of their shots were different and their results reflect that.
Here we have focused on an example of an individual player, but the expected goals metric can be applied to teams or matches in a similar way. Of course, here we can see that a player or team may score more or less often than their xG value suggests, but that is exactly the variation we can now analyze. Does a player score less than they should and who gets chances in high xG situations?
Depth of expected goals
Football is a relatively low-scoring sport and so our ability to measure the likelihood of a goal being scored is an important context. With expected goals, we arm pundits and analysts with another tool to quantify the situations that every football fan wants to hear. Which striker should improve his finishing, which team’s form suggests they should be higher in the league table, etc.
Stats Perform’s depth of data means they now have over 2,500,000 shots enriched with xG values for more than 66,000 players, allowing them to compare and understand the performance of players and teams around the world.
xG is a metric that goes beyond traditional shot counting, but it’s important to remember that it’s still just a metric. We can use it to evaluate basic performance, but it’s the actual goals that will win you football matches. Football is unpredictable and goals can come from any number of unexpected outcomes, but with expected goals we can explain exactly how likely they were.
Conclusion
xG is also a very important part of betting, because in this way you can judge more realistically the level and qualities of certain players and teams, so more and more tipsters use it as an additional important tool in their analysis.
If we missed something, add it in the comments