We’re all guilty of yelling when we see a football player miss what we believe to be a ”sure goal” or a ‘sitter,’ but is our rage really justified? Perhaps not as much as you might imagine. Fans, according to Opta or Sportec Solutions’ ‘Expected goals’ model, may need to control their expectations.
InjuriesandBans has for you the insights on the statistical tool that is revolutionizing how experts, bookmakers, punters and fans alike analyze the beautiful game!
Traditionalists oppose academics invading sport, as expected goals are mostly a creation of academia. You don’t have to be a statistician to record the stats that are normally presented in a post-match graphic — possession, shots, shots on goal, corners, fouls – but you clearly do to figure out each team’s expected goals.
What exactly are Expected goals (xG)?
Expected goals (xG) is a prediction model that is used to evaluate each goal-scoring opportunity and the possibility of scoring.
A xG model computes the likelihood of scoring for each opportunity based on what we know about it (event-based variables).
The higher the xG – with 1 being the maximum because all probabilities vary between 0 and 1 – the higher the probability of scoring.
In reality, this means that a chance with 0.2xG should be scored 20% of the time. If it has a value of 0.99xG, it should be converted 99.9% of the time, and so on.
When determining the quality of a chance, a typical xG model considers the following event-based variables:
- The distance from the goal
- Angle to the goal
- Was that a header or did the player strike it with his feet?
- What area did it take place in? (For example, open play, direct free-kick, corner kick, or counter-attack) Has the player recently beaten an opponent?
Assuming all other factors remain the same, a close-range shot from a central position will have a higher xG value than a header from an acute angle.
The approach described above is widely used by xG providers around the world. Sports Data Collection and Mining companies have also added fresh new variables to current xG models to boost their prediction capability by using their unique algorithm to mix event and tracking data.
These new factors include, for example, goalkeeper’s position and the pressure applied to the player attempting the goal-scoring attempt.
What are the origins of xG ‘’Expected Goals’’?
The origins of xG are unspecified. Vic Barnett and Sarah Hilditch created the term in their 1993 study of the influence of artificial pitches on home team performance. However, the first known usage of a metric that may be considered as a forerunner to the one we use today occurred in the 1960s, thanks to the work of Charles Reep and Bernard Benjamin.
Reep and Benjamin were both statisticians, but their relationship with football couldn’t have been more distinct. Charles Reep was a football fanatic to the point of neurosis. In his early years, he devoured Plymouth matches from the terraces at Home Park before going to London after joining the RAF in 1928. Reep’s fixation increased in London. He got charmed by the offensive wing play of Herbert Chapman’s Arsenal after attending a series of seminars delivered by Arsenal skipper Charles Jones.
Reep began to watch football with a new, more scientific eye after that. He created play-by-play charts, tracked the number of attacking plays in a game, and looked for links between different passing combinations and goals. His discoveries made him a long-ball champion and the first genuine performance analyzer in football. Bernard Benjamin, on the other hand, was a seasoned professor, health scientist, and statistician.
Reep and Benjamin collaborated on a massive data mining project. They saw 667 matches between 1952 and 1967, traveling to four World Cups in the process. They discovered that it required around 10 shots to score a goal, with astonishing consistency. This research, together with Reep’s almost fanatical support for the long-ball, resulted in the birth of a very basic version of the anticipated goals metric.
The advancement of football analytics
While analytics have all but eliminated the long-ball game at the top level in today’s game, the study was one of football’s driving factors in the late 1960s, 1970s, and early 1980s. Charles Hughes developed Reap’s ideas into a little more complex football philosophy known as “POMO,” or Position of Maximum Opportunity.
For decades, sophisticated metrics have been standard practice in American sports. Baseball sabermetrics, made popular to a non-American audience by the film Moneyball, is likely the most high-profile example of establishing objective truths regarding in-game performance through the use of statistics. Similar concepts were created in basketball, gridiron, and ice hockey before statisticians wondered whether they might be used to football as well.
Developing the xG Expected Goals model
Long before Opta popularized xG, various researchers sought to develop a model that explained the aspects that may influence the effectiveness of a goal-setting effort.
In 2004, Richard Pollard – a great friend of Charles Reep’s – wrote a study outlining a number of these elements with the support of Samuel Taylor and Jake Ensum – who would go on to become Head of Analysis at Tottenham. The distance from goal, angle, space enjoyed by the striker, number of touches before attempting the shot, type of build-up play, and location of the assist were all factors considered by their model when determining the quality of a goalscoring opportunity.
Since then, innovative advancements in anticipated goal models have poured as freely as water. The flow of academic research into advanced metrics coincided with the meteoric rise of the football statistics business.
Modern Football and the xG Expected Goals
Data collection in sports has suddenly become a multi-million pound industry. The spark was new technology made accessible to data collection firms as machine learning and optical spatial data tracking grew in popularity. Rather of relying on anecdotal evidence, they now had cameras capable of capturing millions of pieces of data every game. As a result, xG became and continues to become increasingly credible.
Goals are football’s currency, but we rate the quality of a team or individual based on other criteria as well. No one who witnessed Celtic’s 2-1 triumph against Barcelona in 2013 could argue that they were the superior team on the night despite having just 11% possession and recording 166 passes in compared to Barcelona’s 955. Whenever we see a game in which the superior team clearly loses, we can practically guarantee that xG will look more favorably on them.
Expected goals are one of the most precise measures for predicting the goal expectation of each player’s shots based on pretty advanced statistical data. The metric is becoming increasingly popular, making its way to TV analysts’ desks and being used more and more by football clubs, bookmakers, odds analysts and football bettors around the world.