xG statistics is a method of football match analysis that calculates an expected number of goals in a game (xG - Expected Goals) scored by each team, where each shot is given a value of 0.01 to 1.
Adding up xG values of all chances gives us an alternative game outcome which more accurately reflects the events on the field. xG makes it possible to evaluate the effectiveness of chances created, without taking luck, player’s form and mood into consideration.
It is worth mentioning that a victory based on xG values does not always guarantee a winning outcome, but it will always show which team put more effort towards winning. Let's take the 2022 UEFA Champions League final between Liverpool and Real Madrid as an example. The match ended with a minimal win for the Spanish team with a score 0-1. Real Madrid’s goalkeeper Thibaut Courtois earned final man of the match award. And for good reason: Liverpool won that final 2.9 - 0.7 based on expected goals. Statistics of shots was 25 against 5, and all 10 Liverpool’s shots on target were saved by the Belgian goalkeeper.
As we see, xG statistics of a single game doesn’t necessarily tell us much about a overall team’s plays and its strength, but a thorough xG analysis over a span of several games and its correct interpretation can be very beneficial, especially for creating game predictions.
The use of xG metric in the world of football is becoming more and more common. Coaches and managers use xG stats to analyze games, rate their players’ effectiveness and determine the best scoring positions. Worldwide media and TV channels display xG stats during their broadcasts. Top European football clubs hire teams of professional analysts to create game strategies and tactics based on modern xG metrics.
How did xG come to be?
For many football fans xG is a commonplace thing and it is hard to imagine that some 10 years ago it didn’t exist.
The title of the first xG prototype creator is rightfully attributed to Egil Olsen, a legendary coach of the Norwegian national football team, which famously showed an outstanding performance and outcomes during world championships of 1994 and 1998.
As a manager Olsen was famous for analysing game tapes to improve the team’s performance. As part of his analysis, Egil divided the penalty zone and the area around it into squares and assigned each of them a “danger” score.
Egil’s idea was to have his players make their way to specific zones and shoot from the more “dangerous” positions. At the same time, the team’s defence would allow the rivals to shoot only from the “medium” and “low danger” areas.
The Norwegian coach’s model, however, had a few cons. His method could only apply to analyze one specific team, for example. It is also quite subjective in determining a shot’s value.
Present day process of calculating xG statistics had been developed by Opta, a British sports analytics company. They’ve analyzed hundreds of thousands of shots to find out the likelihood of any given chance to result in a goal, taking into account a wide range of factors, which we will go over in more detail below.
How is xG calculated?
To put it simple, a shot from a short distance in front of the goal gives you the highest chance of scoring. If an xG value is equal to 1.00 it means that the likelihood of a goal is 100%. The most basic example of calculating xG is a penalty, with a goal rate of 76% where you only take into account a static player and a set distance to the goal. This means that if the only attack a team executed was one penalty, their xG score will be 0.76, whether they scored it or not.
Open play shots are much more difficult to interpret as numbers, here are the key variables to consider:
- Distance to the goal.
- Forward’s position.
- Goalkeeper’s position.
- How many defenders are in front of the ball.
- Body part used to shoot.
Some additional factors that could be taken into account are:
- Angle, height and direction of a shot.
- Player’s and goalkeeper’s skill.
- Type of assist (cross, pass, rebound, etc).
- Number of touches before the shot.
The more advanced xG models might also consider the following factors:
- Time a player spent on the field.
- Forward’s form and motivation.
- Weather and field conditions.
- Stage of the tournament.
Where does xGscore get its data?
There are currently two main ways to get xG stats. The first one is to keep your own statistics, by hiring a group of experts and developing appropriate software. The second option is to use an existing service, and buy your data from specialized websites.
Because the main service of xGscore is match predictions, our developers’ main focus is on prediction systems and algorithms, and the statistical data is purchased from third-party providers. Moreover, we use data from 4-5 different services for each match, we analyze it and provide the end result. Such method of data utilization from several streams gives our xG model a significant advantage, bringing calculation errors to a minimum.
It is worth mentioning, that the xG value of a match found on our website might slightly change within 24 hours after the match. This is due to the fact that some websites provide their data instantly as a livestream, whereas others need more time to analyze every moment of a match replay.
Aside from xG metrics, our resource calculates a value of «xG Fairness» for each match. It reflects the ratio between the goals scored by the team and their xG values on a 0 to 100% scale. The fairness value also helps to determine how successful or unsuccessful a particular team is.
How to implement xG in predictions?
There are many strategies and methods of using xG statistics in betting. Let’s take betting on an underperforming team for example. Here you need to find a team that creates a lot more xG than they score. Usually the winning coefficient for a team like that would be higher for the next match, because most bettors rely solely on real results and scored goals, and underestimate the team’s play. With each new match, the possibility that the team will finally realize its potential grows because of mathematical deviation of their prior results. These bets work even better against a team that overperforms.
xG statistics are often analyzed to determine bets on totals and both to score. Search criteria here is not the only highest scoring teams, but everyone who generates more chances and has higher xG and xGc (conceded xG) values is taken into account.
The xG model performs especially well on long term bets like the ones on a tournament champion or making it into top three. Another useful xG value is xPts (Expected Points) which can help predict the final ranking of the team based on the first 5–10 games of the season.
One of the most effective ways to use xG is to figure out an expected number of goals in an upcoming match based on the numbers from previous games. This method is the most difficult one because the bettor has to correctly interpret the initial data and convert it into the end result. This process makes it possible to create your own betting line based on an outcome probability. The last thing left to do is to compare your line to bookmakers' betting line before making the most beneficial bet.
This is precisely the method that our service utilizes. We analyze each upcoming match and present prediction of its outcome. Our betting statistics shows that xG data analysis for match outcome prediction increases profitability and ROI of betting by 2-3% on average, which is a great result for long term betting.
Moreover, our service has created a new term for evaluating our predictions named «xG Predictability». This indicator, like a fairness, is expressed as a percentage, it shows how the post-match xG matches the score prediction from xGscore. Due to this value, we able to determine teams and tournaments that are best predictable, and as a result, the most profitable for a bets.
xG predictability and fairness indicators available in the post-match statistics for each game individually, as well as their average season value on the match’s prediction page.