Thanks James. I can’t believe despite being a long time listener that I hadn’t quite picked up the xG range (0-1). I happily listen to you guys talk about this stuff and the .1 and .2’s but this helps 😀
In the St Mirren example for instance, Celtic get the combined xG for those chances, but you could argue, given they all happen in the same short sequence, we were only ever going to score one goal there, so the xG from missing all of the shots is an inflated representation of our probability of scoring said goal.
James has answered this but it allows me to bring up one of my favourite in-game data anomalies. In an otherwise unremarkable early Scottish Cup tie at home to Airdireonians in January 2019, in the same passage of play, Scott Bain made 5 consecutive saves. I have never seen the like before or since. The total xG of the 5 shots was 2.49 xG. At the end of the match, you might think, with Celtic winning 3-0 and the xG "score" being 4.46 - 2.49, you'd missed one of the great cup ties. it is the exceptions that excite us nerds not the patterns!
Hello Liam - all important points. Statsbomb actually calculated 2 xG figures per match to account for the sequencing issue you reference - i.e. a cluster of shots in one continuous phase of play could theoretically go well over 1.0
Thanks James. I can’t believe despite being a long time listener that I hadn’t quite picked up the xG range (0-1). I happily listen to you guys talk about this stuff and the .1 and .2’s but this helps 😀
Tks. Good article
Is total xg at the end of the game just the total of individual xGs?
E.g if Celtic had an xG at the end of a game of 1.5 that could be ten shots with an xG of 0.15 or 2 shots with an xG of 0.8 and 0.7
If that's the case should xG at the end be divided by no of shots?
Tks again
Hello Conor - you are correct. A main distinction people will make is total non-penalty xG vs inclusive of penalty.
xG per shot is the simple arithmetic calculation you describe and is generally considered as average chance quality.
Another main distinction people will make is xG from open play vs set pieces.
Thanks for this.
Is context and game state not important here too?
In the St Mirren example for instance, Celtic get the combined xG for those chances, but you could argue, given they all happen in the same short sequence, we were only ever going to score one goal there, so the xG from missing all of the shots is an inflated representation of our probability of scoring said goal.
James has answered this but it allows me to bring up one of my favourite in-game data anomalies. In an otherwise unremarkable early Scottish Cup tie at home to Airdireonians in January 2019, in the same passage of play, Scott Bain made 5 consecutive saves. I have never seen the like before or since. The total xG of the 5 shots was 2.49 xG. At the end of the match, you might think, with Celtic winning 3-0 and the xG "score" being 4.46 - 2.49, you'd missed one of the great cup ties. it is the exceptions that excite us nerds not the patterns!
Hello Liam - all important points. Statsbomb actually calculated 2 xG figures per match to account for the sequencing issue you reference - i.e. a cluster of shots in one continuous phase of play could theoretically go well over 1.0