# Expected Goals | How To Use xG For Football Betting

Football is a game where all goals are valued equally at 1, but not all attempts on goal are equally likely to result in a goal. Basic statistics like ‘shots on target’ do not provide sufficient insight into the performance of teams. That’s where Expected Goals (xG) comes in.

Expected Goals (xG) is a statistical metric that assigns a “quality” value to every attempt on goal, based on past data. This enables us to better analyse games and determine whether results are in line with the expected outcome. Expected Goals has the potential to revolutionise football prediction, team and player evaluation, and even sports betting.

Article Contents

## What Is Expected Goals (xG)?

Expected goals (xG) is a statistical metric used in football to quantify the quality of scoring chances in a match. It is a measure of the likelihood of a particular shot resulting in a goal.

xG is typically calculated using machine learning models that analyse large amounts of data on historical shots and goals, in order to identify the factors that are most predictive of goal-scoring success. These models can be trained on a variety of variables, such as the distance from the goal, the angle of the shot, the body position of the shooter, and the number of defenders between the shooter and the goal.

Once a model has been trained, it generates an xG value for each chance in a particular match. This value represents the probability of the chance resulting in a goal, on a scale of 0 to 1. For example, a shot with an xG value of 0.1 would be expected to result in a goal 10% of the time, while a shot with an xG value of 0.9 would be expected to result in a goal 90% of the time.

xG has become a popular metric in football analysis, as it provides a more objective way to evaluate the quality of scoring chances than traditional statistics like shots on goal or goals scored. It can also be used to identify players or teams that are overperforming or underperforming relative to their expected scoring output.

## An Example of Expected Goals (xG)

You might have heard of “expected goals” or seen it used on Match Of The Day. Here’s an example from an old Premier League fixture. What this shows is an xG score of 2.2-0.4 to Man United after the game.

###### Man United v Stoke Expected Goals Stats (Opta)

On average, 9.7% of shots in the Premier League are converted into goals. Some are long-range, some are headers, and others are straight forward ‘tap-ins’. Importantly, not all shots have an equal chance of hitting the back of the net. Expected Goals accounts for this, and determines how many goals teams ought to have scored in a match. This gives an expected scoreline.

In the above graphic, Man United’s chances are shown in red/pink, and Stoke’s in dark grey. The bigger the square, the better the chance was. It shows that Man United created more clear-cut chances than Stoke — some of which were inside the six yard box. In this case, the final (actual) score of 3-0 isn’t too far away from the xG estimation of 2.2-0.4. But ultimately Man United scored 0.8 more goals than the xG model expected. This suggests they were fairly clinical, or perhaps slightly fortunate — but nonetheless, they deserved to win the game.

The accuracy of the model largely depends on what factors are used to calculate the xG rating for individual chances. For instance, some basic xG models will only account for the distance of the shots. Other more complex models will account for the positioning of several influential players in relation to the ball to more accurately define how difficult it was to score from different situations. I discuss this in more detail later on in this article.

## The Appeal of xG for Football Analysis

The xG statistic is growing in popularity because it gives fans, pundits, and even those involved in the sport, a much better way of justifying who was superior in a game — regardless of the actual score.

Previously we might have relied on basic stats, like shots on goal to justify our opinions. But with Expected Goals we have the ability to delve deeper, and account for the quality of those shots. Now we can say “that game should have been 3-0″, with enough detail to justify the statement we’re making.

I’ve spoken before on this site about the limitations of using basic statistics for football analysis. It’s important to recognise that some results simply don’t tell the full story. The scoreline isn’t the only stat that matters for your betting analysis. It’s important to have some sort of context to a past fixture, and even better to be able to answer questions, such as:

• What was the most likely final score?
• Were [Team X] fortunate to have won?
• In total, how many “clear-cut” chances did [Team X] have in the match?
• Should [Player A] have scored from his/her opportunities?
• This season, how much below or above average has [Player A] or [Team X] performed?

Now, thanks to Expected Goals, we can mathematically answer these questions and make better judgements on football matches going forward. From a betting perspective, it has the potential to vastly improve on the basic football prediction models I detailed in my post: The Basics Of Creating a Football Prediction Betting Model.

## xG for Analysing Runs of Form

Runs of form often deceive us throughout the season. It’s is something I’ve touched on before in my guide to football hype & noise.

Expected Goals (xG) is a powerful metric that can help bettors to maintain realistic expectations. By comparing a team’s xG with the actual outcome, bettors can calculate the xG goal difference and assess whether the team’s current performance is sustainable.

This measurement can be particularly useful for mid-table clubs, which may experience ups and downs during a season but generally lack the depth and class of top Premier League teams. By using xG, bettors can gain a clearer understanding of a team’s true potential and make more informed betting decisions going forward.

In the next section, I’ll delve deeper into how xG goal difference works.

## xG for Individual Players

The Expected Goals (xG) figure has a lot more meaning for individual strikers than it does for other players who aren’t necessarily expected to score (e.g. defenders).

I was intrigued by something:

Are the top scoring strikers in the Premier League flattered by the fact their team creates more chances for them?

I looked into it and found that during the 2016/17 season, the player who outperformed the number of goals he was expected to score by the biggest margin was in fact… top goalscorer Harry Kane.

Harry Kane is deservedly regarded as the best, most prolific striker in the Premier League 2016/17 season (not that I’m biased or anything). He was expected to have scored 18.59 goals, but managed to score an extra 10.41.

###### These statistics came from the BBC.
Player Goals Expected Goals Expected Goals Difference
Only includes players with 50+ shots
Kane (Tottenham) 29 18.59 10.41
Lukaku (Everton) 25 15.32 9.68
Llorente (Swansea) 15 7.09 7.91
Son (Tottenham) 14 6.73 7.27
King (Bournemouth) 16 9.56 6.44

This table also sheds light on why the top Premier League clubs were interested in Swansea’s Llorente (#3) at the end of last season. He performed well despite playing for a struggling team.

Provided the level of detail is high enough, Expected Goals could be used to compare and even value strikers, based solely on their ability to convert chances.

## How Accurate Is xG?

That’s the big question.

xG can include as many or as few factors as you like in order to calculate the average likelihood of each shot being scored. In theory the more the model knows about past cases, the more accurate it is. But, of course, this relies on the model being fed with relevant data only.

I’ve seen sports bettors forming their own xG prediction models by manually assigning a score to each individual chance in games. The principle is great, but in practice there are many ambiguous cases — such as accidental goals — that can potentially weaken the accuracy of the model.

Football data experts Opta create their own Goal Expectation figures. They’ve analysed over 300,000 shots (a large sample) to calculate the likelihood of an attempt being scored, given a specific position on the pitch, during a particular phase of play. Opta’s model accounts for factors such as:

• Distance from goal
• Angle of the shot
• 1v1 situations
• Quality of the assist (the type of ball put in)
• Passage of play (e.g. open play, direct free-kick, corner kick)
• Player’s energy (has he just beaten an opponent to create the chance)
• Rebounds

To obtain an accurate Goal Expectation figure, advanced data gathering and statistical skills are required. However, it is still possible to gain an insight into football matches using fairly simple expected goals systems.

Understat is a great site for monitoring the expected goals from past fixtures. But unfortunately I haven’t yet found a free source that enables visitors to download xG figures for all historical fixtures.

## xG — The Future Of Football Betting?

To summarise xG, it’s worth reiterating that the scoreline isn’t the only stat that matters.

OK — in terms of points, it is. But Expected Goals reveals a number of things about a game, series of games, and individual players such as:

• Whether chances were really as good as we thought.
• What the scorelines should have been, given the chances.
• A team’s goal-scoring record in comparison the chances they’ve created.
• How many goals individual players should have scored from the chances they’ve had.
• Whether ‘runs of form’ are likely to continue (using the xG difference).

Expected goals enables us to better judge teams and players, reduces bias, and helps to more accurately predict future results. The applications to betting are certainly there — and it doesn’t need to stop at football. In fact, the logic behind xG can be applied to any sport. For now it’s being used for football — a complex game that’s very difficult to predict, and requires equally complex stats to derive something useful from past results.

However, Expected Goals isn’t an all-knowing metric that’ll instantly turn your betting model into a money-making machine, mind you. In fact, in the future it could just help make the markets more accurate. But for now it is a step in the right direction for football prediction and value hunting.