What Is Expected Goals (xG) In Football

Mathematics, football and football fans are unlikely combinations, yet with the sport advancing more and more in rent years, somehow mathematics and sports go hand in hand.

Football has incorporated data and statistics much more in recent years. Expected goals, or “xG” for short is now a vital role in the sport. But what does this strange term “xG” mean? Who invented it and where did this term come from? Find out all about it in our article guide.

What Is “xG” or Expected Goals In Football? How Is xG Calculated?

Sam Green of Opta invented “xG” in 2012, and since then it has grown to be one of the most well-known and useful metrics in football analytics.  

“Expected goals” was one of the first formal metrics to catch on with football fans, therefore it was inevitable that it over time it would come to face some sort of critique, especially between the conventional method of watching the game and the impending field of data analytics. But before we make a decision about whether it contributes positively or negatively to the game, it’s critical to understand how the “xG” metric functions and how we should use it. 

Expected goals / xG, measures the quality of a goal chance by estimating its chances of success based on data from previous attempts at comparable shots. On a scale from 0 and 1, where 0 means an impossibility to score and 1 means an expectation that a player would score every time, around one million shots are used from OPTA’s historical database to calculate the “xG”.

It’s obvious that a shot from within the penalty area is more likely to result in a goal than one from the halfway line. We can quantify these possibilities using xG. Assume, for instance, that the chance coming from within the box has an xG value of 0.1. In this instance, a player should score one goal out of every ten shots, or 10% of the time.

How Are XG Stats In Football Calculated? Why Is XG Model Important in Football Leagues?

In order to calculate correctly XG stats, usually we have an instinctive sense of which opportunities are more or less likely to result in a goal when watching a football game.

• How near the goal was the shooter?
• Was the angle of their shot good?
• Did it have a header?

There are often 25 shots towards the goal net every game, which presents a challenge. In most matches, that’s 250 shots over the course of a weekend. Even the most skilled football fans would need a lot of time to precisely determine the chances of scoring in each game. 

For each of the 9,609 shots that were taken during the 2022–23 Premier League season, a simple xG formula model can determine the probability that a goal will be scored. With the help of almost a million shots from OPTA data, OPTA’s xG model shows a digital approach called XGBoost (unrelated to the name of the statistic). 

The xG formula takes into account a number of factors up until the precise instant the shot was fired. The impact of more than 20 variables on the probability of a goal being scored is analysed. The following list includes a few of the most crucial elements:

• The distance from the shot to the goal net
• How likely it is that the player would be expected to score from where he was placed
• The angle of the shot
• The position of the goalkeeper and how likely he is to save the goal/the likelihood of a goal happening
• Depending on where other players are positioned, how well the player who hit the ball is
• How much pressure the other team’s defenders put on their opponents
• Shot characteristics, such as what body part the player used or if it was a volley, header, or one-on-one.
• The game-playing style; includes open play, quick breaks, direct free kicks, corner kicks, and throw-ins, among others.
• Details about previous actions on the pitch; i.e the type of assist (such as a through ball, cross, etc.).

The GK (goalkeeper) position component in the xG goals model which enables us to calculate the likelihood that a goalkeeper will make a save, is another feature. How close the goalkeeper was to the shot and his location in relation to the shot’s line of sight to the goal are taken into account, as well as whether or not the GK was inside the penalty box and able to use his hands.

The xG model predicts where the player shooting is likely to aim in the goal and how this influences the chance that the shot will be saved by the GK, in addition to other things. With the use of these elements, xG is able to assess the GK placement and determine the ideal spot from which to make a save.

Football penalties are the most reliable shot and are assigned a constant value that reflects their historical conversion rate (0.79 xG).

Typical Misconceptions About xG In Football

• xG Overperformance: An Understanding – A football player or team that has been exceeding its xG, later on, isn’t required to play below expectation in order to revert to expectation. It is possible that a player will still outperform their predicted goal total if they have already scored five goals more than expected at the beginning of the season. Erling Haaland is a perfect example of this.  

Erling Haaland has scored two goals in one game so far in the 2023–2024 Premier League campaign. Haaland scores 2.25 goals overall per 90 minutes. Additionally, he has a season total G/A (goals + assists) of 2 goals. His involvement in the goal amounts to 2.25 per 90 minutes. He has a 1.47 Non-Penalty xG per 90 minutes. He now ranks in the top 99 percentile of Premier League players with an npxG production of 1.31.

• Game-level xG – The first objection frequently takes the form of instances in which the measure isn’t being used properly. The most prevalent of which is at the top-flight level. It’s not always true that a team with a greater overall xG total in a game should have won. xG does not account for the expected result of the game; it just measures chance quality.
• Expected goals – The probability of an expected goal does not “expect” goals to happen precisely as it predicts. We also understand that goals cannot be scored partially. The term “expected goals,” which refers to a measure of the chance that a goal will occur, is taken from the mathematic notion of “expected value.” I.e, a fair coin toss has a 50% chance of landing on heads and a 50% chance of landing on tails (both the expected heads or tails is 0.5). The same applies to the expected goals. Unavoidable variation from the predicted value is important data that we take into account when calculating xG in football.

Depth of Expected Goals

Football isn’t generally classified as a high-scoring sport. By providing data analysts and football commentators with projected goals, we can provide them with another instrument to quantify the narrative that every football fan loves to hear. “Which striker is having trouble with their goal-scoring? Which team’s performance would indicate that they belong higher in the league table?”

Although xG is one of the advanced metrics that goes beyond the standard shot counts, it’s vital to keep in mind that it is only an analytical metric. Although it may be used to assess underlying performances, actual goals are what decide football games and cause so much excitement for the fans. 

We now have almost 4.5 million shots supplemented with xG values for more than 100,000 players thanks to OPTA’s unmatched depth of data, allowing us to compare and comprehend the performances of individuals and teams throughout the world.

How do xG models treat penalties?

The majority of models assign a static value of 0.76 xG to penalty kicks since they all share the same attributes, which corresponds to the historically observed conversion rate of penalties. This static value is changed to 0.78 xG in the 2022 update of the StatsBomb xG model. When analysing performance, it’s common to subtract goals scored and xG from penalties from player and team totals.