14 years ago
1 day ago

Expected Goals
As well as having more data and analysis at your fingertips before a match, there are now a number of new stats that you can track during a game too. We’ll cover most of these in the coming weeks but there’s one we want to focus on here – xG or ‘expected goals’, to give it its full name.

xG has become more widely used in recent years but, for a long time, we felt that none of the current systems would work effectively with the FM match engine. This is why, this year, we’ve created our own xG system, with the help of our friends at football data intelligence specialists, SciSports, who work with real clubs across the sport.

Our xG system is tailor-made to work with the FM match engine which allows us to go beyond what current expected goals models are capable of in real life. Among other factors, our model considers the shooting player’s distance from goal and the angle and speed of their shot to determine xG, we’ve also accounted for the position of the other players on the pitch relative to the shooting player too.

You can view in detail xG in both the half-time review and full-time summary where you’ll be able to see the xG match story and how each team’s xG has fluctuated over the course of the game. You can also view xG shot maps for both teams that show each shot’s xG rating. Post-match, xG will be included in the analysis you receive from your data team.

As well as adding xG, the full-time summary has been expanded to include more of the stories that matter in relation to the match. Your Press Officer will include the best media clippings from the immediate aftermath of the final whistle and you’ll see reactions from fans on social media too as well as the updated league tables and final scores coming in from around the grounds.
14 years ago
1 day ago
I love that xG has finally made it into Football Manager!

But I thought FMs own introduction was a little short on the details. So here's a quick primer on what xG is in real life.

When a match finishes 2-2 and two shots were unlucky to ricochet off the post, fans often walk away from the game saying something like "we were unlucky! We should have scored 4". xG is similar to this, but much more sophisticated.

When a player takes a shot from inside the six yard box with no defenders around him and an open goal you'd say he has a very high chance of scoring, probably close to 100%, lets say 99%, this would be called 0.99xG. If a player takes an off balance shot from 40 yards out with a keeper well positioned you'd say that had a very low chance of scoring, lets say 10% or 0.1xG. 0.1xG in this case means that from this position we would expect a goal 10% of the time.

The probability of a goal is calculated using a variety of sophisticated formulas which take into consideration the location of the shooter, the body part used, the type of pass that he received the ball from and the type of attack, e.g. is it a fast counter from a defensive corner or a quick change over of possession in the final third.

Thousands of goals from around the world have been analysed to create formulas which will tell you how likely a goal is in a certain situation. A header on the penalty spot from a cross following a 5v3 fast counter attack? If the data shows that happened 1000 times in football matches around the world and resulted in a goal 800 times, the expected goals for that chance is 0.8xG.

That's a simplistic example, in reality the most sophisticated xG tools use thousands of data points to try and give the most accurate predictions for every chance.

In fact the most advanced xG models do far more than look at specific shots, they can analyse entire passages of play. Even if a chance doesn't lead to a shot because of a last minute intervention, the data might show that such an intervention was very unlikely and the probably of scoring of a goal from a particular position was very high, so you may still say that a team had more xG even without actual shots being shown in the stats.

So how can xG inform our decision making?

If we have a striker who has scored 5 goals in his last 10 games but when analysing each of his chances we predict that he should have actually scored 20 of them we would say that he is not meeting his xG.

Similarly if we had a striker who had scored 20 goals in 10 games but his xG was 5 goals we'd say he was beating his xG.

This is a great way to analyse the performance of striker, since we're not just looking at the number of goals scored, but whether he's actually converting the chances he's receiving.

But there are different ways to analyse xG. A player who is getting 0.25xG (only scoring 25% of the goals you'd expect him to score given the chances he's had) is that a reason to drop him? Or is it a reason to think that over time he should start scoring more given the opportunities he's finding himself in.

Similarly if a player is scoring well above the number of goals we'd expect, this could be reason to give him a mega new contract as a star player, or it could be reason to assume his luck will soon run out and you can't rely on him scoring so many over a longer period.

I lean more towards thinking that everyone is going to regress to the mean. You're never going to get a player scoring twice his expected goals for an entire season and if you're relying on him doing so then you probably need to figure out how to create more and better chances to guarantee you're scoring goals in the future.

For a more in depth look at Expected Goals check out this article:

i'm very interested to see what affect xG will have on the game. For instance, statisticians have shown that Jurgen Klopp was incredibly unlucky when his Dortmund side found themselves second bottom of the Bundesliga in the 2014/15 season. Their xG for and against actually had them much further up the table. Will the board take into account your xG and your "Expected League Position" before deciding to sack you? We'll have to wait and see I suppose.
13 years ago
22 hours ago
Brilliant explanation of xG. Thanks!
7 years ago
2 months ago
So we now blame xg instead of shit match engine? Thats briliant.

You'll need to Login to comment