Let’s Talk About Lukaku

During the 2022 World Cup in Qatar, Belgian striker Romelu Lukaku had what can only be described as a nightmare in their game against Croatia. He missed chance after chance after chance, three of which could easily be described as “sitters”. One goal could have broken the deadlock in the match and seen Belgium progress to the next round.

He didn’t convert any, and Belgium’s “Golden Generation” crashed out.

Many have described this as the worst individual performance by a striker in World Cup history…and it’s hard to argue when you see the chances he missed.

So, how does this relate to xG? Well, Lukaku’s total xG for the match was 1.98, meaning that he could have easily scored two of these chances. This shows us that Lukaku severely under-performed during the match.

We’ve peppered James’ extract with KlipDrawn examples from the Belgium vs Croatia game, and James himself has kindly provided us the xG score for each chance.

With that being said, here’s the book extract…

Chapter 1: Expected Goals

What Is the Expected Goals Method?

Fans make several subconscious judgements whenever they watch a football match. Most routinely, they form an opinion of which team is playing better. Does the scoreline accurately reflect the performance level of each side? Should one team have scored more goals than they actually have? In other words, has luck had a considerable impact on the outcome of the match?

Football is a sport riddled with randomness. Bad teams will often defeat good teams, bad players will often go through patches of playing well and bad managers will often end up in charge of high profile clubs. The damaging effects of chance make football an incredibly

hard sport to understand. The Expected Goals method is a tool which we can use to separate the skillful from the lucky and more accurately gauge performance levels.

Put simply, xG tells us the quantity and quality of chances that each team creates from a match. When we look back on results, we often have strong opinions over which side played better. ‘If we had scored that penalty we would have won’. ‘They were so lucky to have scored that deflected shot from long-range’. ‘We created so many good chances, how did we lose?’ We base our view of who should have won on the scoring opportunities that were created during the match. Expected Goals data is simply a way of quantifying these scoring chances.

Football is heavily centred around goals. Match reports generally focus on the opportunities that each team created. Highlight packages centre around the openings that each team carved out. Commentators get most excited when a team is about to score. Football revolves around goals, and goals can only occur when teams create scoring opportunities. Indeed, every tactic ever created represents a coach’s attempt to develop his team’s ability to create chances, whilst at the same time improving their ability to not concede chances. A football match is essentially a series of attacks from each team on the other team. The sides who have the highest quantity and highest quality of attacks are clearly the best sides. This is what the Expected Goals method measures.

However, goals are almost as rare as they are important. Thousands of actions take place over the course of each match – in fact, Opta’s data collectors suggest that an average of 3,000-4,000 events such as passes, tackles, duels, saves and so on happen over a ninety-minute period. Only a handful of these actions will be shots that result in goals (the average number of goals per match is around 2.7). Each one of the thousands of in-play events is geared towards one thing: chance creation. Assessing the nature of the chances that each team created will reveal which side has deserved to win, regardless of the actual scoreline.

The Expected Goals method cuts right through to the core of football thinking. The ability of sides to threaten the opposition goal, whilst simultaneously preventing danger to their own goal, is what separates good teams from bad teams.

How Is Expected Goals Calculated?

Essentially, xG indicates how many goals a team could have expected to score based on the quantity and quality of chances that they created in a match. Fans often come away from football matches thinking, “We created much better chances then the opposition, we definitely should have won”. The Expected Goals metric is a way of quantifying these scoring opportunities, allowing a better insight into the ability of teams than the actual scoreline does.

Teams are often judged by the quantity of shots that they have in a match, or indeed in a season. Media companies will show the stats for how many attempts at goal each side has taken in a game. The central premise of xG is that the quality of those shots is of equal importance as the quantity. Analysts can work out the number of goals that a team would have expected to score from a certain amount of shots of a certain quality. Similarly, analysts can work out which players have scored more chances than they would be expected to.

Expected Goals data is collected by several different data companies, football clubs and betting firms. The main provider of xG stats to media companies is Opta Sports, who claim to collect the most complete dataset for the Premier League, English Football League, Scottish Professional Football League and many other divisions across the globe. Opta’s data experts have analysed over 300,000 shots to calculate the likelihood of an attempt being scored from a specific position on the pitch during a particular phase of play.

Expected Goals works by measuring the likelihood of each shot resulting in a goal. Each effort at goal which takes place in a match has a “Shot Probability” value. For instance, a shot from 30 yards out through a crowd of players may only have a 2% chance of hitting the back of the net, giving it a value of 0.02(xG). On the other hand, a shot into an open goal from six yards out might have a 95% chance of being scored, resulting in a value of 0.95(xG).

At the end of any given match, the Shot Probabilities from either side are added up to reveal the Expected Goals scoreline from a match. For example, suppose that Arsenal play against Manchester City. The London side have six shots over the course of the match, but they are all long shots from distance with a Shot Probability value of 0.1(xG). The Gunners will have amassed a total Expected Goals score of 0.6(xG). Over the course of the same match, Man City only have two shots at goal, but they are both from close range. Suppose that one shot is worth 0.3(xG) and the other is worth 0.4(xG). Man City’s Expected Goals score over the course of the ninety minutes is 0.7(xG). Thus, the xG scoreline from the match would be Arsenal 0.6(xG) – 0.7(xG) Man City. The scoreline would reflect the fact that Manchester City performed narrowly better than their London counterparts.

There is a natural question which follows: how do you determine the probability of a shot’s success? The location of a shot has a large bearing on how likely it is to result in a goal. A shot which is taken from a wide position, thirty yards out from goal, will only have a small chance of going in. On the other hand, a close-range shot from a central position will have a high probability of scoring.

An analyst could look at a large sample of past shots taken from an exact position and find how many beat the goalkeeper. Say an analyst looked at 1,000 shots taken from the exact position at the right-hand corner of the penalty area (we are assuming the analyst has a large database of shots spanning across several divisions and several seasons. Thousands of shots are taken each season, each one being recorded by companies like Opta. These companies can draw upon vast databases of past shots in order to determine Expected Goals probabilities). Suppose only 50 of these shots ended up in a goal. He could conclude that future shots from this location have a 5% chance of beating the goalkeeper (as 50/1000 previous shots from this position hit the back of the net). Thus, the Expected Goals value from this position is 0.05(xG).

Whilst the location of a shot forms the main basis of its danger level, other factors also play their part. Shots which come from crosses are considerably harder to convert than shots which take place when the ball is standing still. Whether the shot is headed, volleyed or hit from the ground also affects its chance of success. So too does it matter whether the effort is taken on a player’s weaker foot. Analysts account for a whole range of such factors in their Expected Goals models.

The above description is a very brief introduction into how to collect Expected Goals data, but hopefully the reader can already get a sense of how it can be used to analyse football. Teams who are consistently creating high value chances are clearly dangerous opposition, whilst those who are only clocking low xG Shot Probabilities evidently lack potency going forward. Similarly, an analyst can work out how many goals a player would have expected to score based on the quality of chances that he has received. If a striker is scoring lots of goals from difficult positions, we might applaud him. Conversely, if a player is drastically underperforming his Expected Goals output, we might question his ability.

Why Is the Expected Goals Method Important?

Every judgement, opinion or prediction that we make about the beautiful game is grounded on our assessments of past performances. But how can we expect to accurately analyse the sport when it is so considerably determined by luck? This is where the Expected Goals method steps in, providing an antidote to the disease of randomness which permeates football. It is at the forefront of a smarter, more analytical football philosophy.

In recent years, Expected Goals has been used by various establishments in order to facilitate a better understanding of the sport. The more innovative football clubs have begun using xG in several regards. Most obviously, they use it to measure the performances of their own team and players. Has their side been achieving the results expected of them? Do they actually deserve to be where they are in the league table? How efficient have they been at creating chances and preventing the opposition from creating scoring opportunities?

The scouting teams at clubs also use xG data to uncover and sign hidden gems; players who are undervalued by the rest of the footballing world. Certain English clubs have managed to consistently sign great players for low prices because of their Expected Goals tools. This analytical style of recruitment has allowed teams to enjoy great success on shoe-string budgets. We will study the methods of these teams later in the book.

A different type of institution has also utilised the immense predictive ability of the Expected Goals method. Professional gambling syndicates have used xG to calculate accurate probabilities of events occurring. These companies use Expected Goals data to generate odds, which they compare to the bookmakers’ odds. The success of their businesses depends on their ability to make more accurate forecasts than the bookies. The Expected Goals method has allowed them to do this. Later, we will more closely examine the top secret gambling cohorts who have turned over millions of pounds through utilising xG.

Fans have finally begun to pay an increasing amount of attention to Expected Goals stats. Slowly but surely, the media have taken notice of this increase in interest. Football supporters are insatiable consumers of facts. In a sport where knowledge is power, Expected Goals is slowly emerging as the most authoritative form of data. In later sections, we will see how the media have entered the early stages of the xG revolution, what problems they have faced in incorporating it into their broadcasts and where the future may lie for supporter interaction. I am confident that, in a few years’ time, it will be impossible to read a match report without finding a reference to the Expected Goals scoreline from the game.

The increased exposure which the metric has seen over the last couple of years is simply a drop in the ocean of what is to come. The reason why is simple: Every footballing judgement ever made is based on an analysis of the performance of teams or players. And the Expected Goals method offers by far the most advanced, profound and accurate gauge of performance.

Soon, those who do not understand or pay attention to xG data will be left behind. The Expected Goals method allows you to speak about football in a more intelligent language.

“Isn’t X an incredible manager?!”

No, xG shows that his team have incredibly lucky to get the results that they have.

“How could Y’s defence play so badly?!”

Actually, the Expected Goals data shows that they played very well.

“Why does Z keep missing absolute sitters?!”

Sorry to correct you, but xG suggests that this player is actually scoring more goals than would be expected of him.

Too often in football, the result dictates the narrative. A team who plays badly and wins has “a great mentality” and is able to “grind out results even when not playing well”. However, a team who plays badly and loses will be deemed to have obvious flaws. Both of these teams have performed at the same level (i.e. badly), but notice how our analysis has been changed dependent on their result. In order to avoid being fooled by randomness, we should direct more attention to the Expected Goals totals amassed from each game. This will allow us to assess performances, rather than results.

Only when we fully embrace the Expected Goals method can pundits begin to more accurately comment of football. Only then can managers give more reasonable post-match interviews. Only then can the fans select the best players for their fantasy teams. Only then can we haul football out of the dark ages and into a more intelligent era of analysis.

And that's it! To continue reading and learning more about the philospohy behind the xG metric, you can buy a copy of the book by clicking here. We'd like to thank James for allowing us to publish this extract of his book and for participating in the xG analysis of Lukaku. If you have questions about anything you have read in this extract, please don't hesitate to get in contact with us and we'll be happy to answer your questions, or pass them on to James himself. You can also check out his Twitter channel, The xG Philosophy.

Finally, the video extracts in this blog where created using KlipDraw Animate. If you would like a FREE 30-day trial of this software, click below.

Until then, thanks for reading!