It's an assessment of the quality of the positions a team is shooting from. In one game it's a sideshow, over a longer period it has value.
Chris Wood massively outperformed his xG last season. He's a good finisher, who had the season of his life. xG confirms what we all could say anyway. But if someone we weren't watching every week had a similar season, it would indicate they were worth having a good look at.
Who determines the relative quality of shooting positions, and how?
I don't recall 'who determines the relative quality of shooting positions' in the Stonecutters song in The Simpsons, so it's not them.
A big database of whether or not people have scored from those positions in the past.
The clearest example is penalties. About 79% (i think) of penalties are scored, so a penalty has an xG of 0.79.
Now extend the same principle to shots from the edge of the area with 5 defenders between you and the goal (id guess about 0.05) and shots from the edge of the six yard box with no goalkeeper (about 0.98?).
OK. Now imagine the same shot taken twice from the exact same place with the exact same defensive configuration of one defender and the goalkeeper. In scenario 1 the defender is Marquinhos and the keeper is Donnarumma. In scenario 2 the defender is Muswell with his dodgy knees and the keeper is all 5'7" of me. Are we making a case that the likelihood of scoring is exactly the same in both scenarios?
No. And xG isn't saying that it is. It's giving a number based on a large volume of data at a professional level.
But if it's not taking into account opposing action, then what possible value can it have?
Well it is. But it's taking a volume of data, so comparing against an average defence.
Look at the forwards and goalkeepers who outperform their xG. They will, almost without exception, be players that you think are good at finishing and saving respectively. Now, maybe that means it tells you nothing, or maybe it means for games and players you haven't seen, it does say something.
So in the game tricky is talking about, either SheffUs finishing was bad and/or Brizzles was good and/or Sheffu were bad at stopping shots from going in and/or Brizzle were good.
And id guess someone who watched the whole game would agree with one of those.
This does not sound like a remotely insightful stat. "This stat tells us that any of a bunch of things, or any combination of them, could be true, but you'd have to watch the game to know."
Again, over a longer period of games it's more useful.
If Sheffield United are consistently underperforming their xG, they have a problem. If, over the next 5 or 6 games it evens out, it was just a bad day.
I like it, as a data point to combine with others when you can't see all the games. If someone scores 10 goals on an XG of 15 then I'd look at them rather differently than I'd they had an XG of 7, and also think about who they were playing for and against.
This is one of the things that makes me think it's a pretty blunt tool... Because not all positions are the same. There's a world of difference in your scoring chances if you shoot from a position where there has been stability for a period of time, versus if everyone has run half the length of the pitch to arrive in the same positions. You might consider this an extreme angle, but this is exactly what different ways of playing aim to achieve. Different ways of arriving in scoring positions. Which definitely have different percentage outcomes. You only get good correlation if everyone is setting up to play the same way. You are going to get very different predicted statistical spread in a basketball game, than in a game where both teams focus on not giving the ball away. Not to mention the variance if you are playing with a team getting run over all over the park, versus a team with a stable platform, and always being backed up.
But I could accept that it's just a pretty unsophisticated form of averaging if fucking idiots weren't constantly banging on about it telling them something that it doesn't.
Ultimately if xG is constantly at odds with the actual goals scored, on a match by match basis, it's not very much use. Except to give a reassuring number to people whose eyes don't work.
It has been created, like sadly all too many things in football, to appease the fantasy sports crowd. I have been firmly of the view for some time that fantasy sports are the worst thing to happen to sport in a long time, because it makes people look at and value the wrong things. Matt Forte was at one point the most valuable player in fantasy NFL leagues despite being a mediocre yardage accumulator who had little impact on actual games, and Matt Carkner was a value pickup in fantasy hockey because he was guaranteed to spend minutes in the penalty box despite this objectively being a bad thing for his team.
If you read “How To Win the Premier League”, Ian Graham, who was Liverpool’s “Director of Research”, used it heavily. So this is just untrue.
It’s part of football’s attempt to find a way to do Moneyball.
I absolutely do not accept that xG is used in any meaningful way to scout players, any more than FIFA videogame ratings are. They're probably based in the same level of subjective "science".
Even if that was the original purpose for xG, it's clearly used primarily as a tool for gamblers and gamers now. There is a wealth of actual science and scouting technology available to clubs these days, any player acquisition team using obviously flawed stats like xG as a primary tool to evaluate players isn't going to last long.
Not saying I'm shitfaced, but I just spent a couple of minutes trying to figure out why my key fob wasn't working on my apartment door. Was about to go get the concierge to fix it when I realized that the fob is for the front door of the building, my apartment requires me to use the key. I've only lived here 11 months.
May be watching the match on catch up.
