Analytics in Football - Official Thread

[Philby]

A quick one on xG. Yesterday was a good example of a few interesting aspects.

1. Influence of game state on xG

City were utterly dominant but some late token attacks from Utd, at 3-0 down, meant the match xG (City ahead by 0.7 via Opta and 1.1 vs sofascore) makes it look like a vaguely competitive game in theory.

2. Two of those big xG chances were weird

84mins - Mbeumo 0.58 (Opta) / 0.64 (Understat)
85mins - Casemiro 0.47 (Opta) / 0.45 (Understat)

Both were flying in at high speed and each player had limited hope of accurately re-directing the ball.

Does the xG model accurately reflect just how hard those two chances were?

3. xG vs post-shot xG

Because Man Utd only had 2 shots on target their post-shot xG was tiny (0.14). This is arguably closer to the truth than the actual xG.

4. Mbuemo's fantastic volley

Went from 0.02 xG to only 0.08 post-shot xG.

I'd wager around a third of PL keepers wouldn't typically save that shot. An increment of only 0.06xG feels stingy for how much his contribution made that chance better than it was before he made contact

 

[Cloudy O'Ranieri]

I've been trying to tell folk around here about point one for ages. I'm still not sure they get it tbphwy. Although, Nollaig will act dumb when I know deep down he gets it but still uses the contextless stats for his schtick.

 

[Philby]

I've been trying to tell folk around here about point one for ages. I'm still not sure they get it tbphwy. Although, Nollaig will act dumb when I know deep down he gets it but still uses the contextless stats for his schtick.

He knows enough to be dangerous, the cute hoor.

On a related note, I think there was some analysis done a few years back that indicated that it's only after 11 games of a season that you can start to "trust" xG (...i.e. from then on there's more signal than noise with the use of 11 games worth of xGD to predict the remainder of the season)

 

[Cloudy O'Ranieri]

He knows enough to be dangerous, the cute hoor.

On a related note, I think there was some analysis done a few years back that indicated that it's only after 11 games of a season that you can start to "trust" xG (...i.e. from then on there's more signal than noise with the use of 11 games worth of xGD to predict the remainder of the season)

Even then, you'd need to be careful. There'd be some shit going on with some teams under the bonnet that the raw xG isn't capturing. Spurs last season were a great example. They were top of the pile for xGD after around 11 or 12 games but it never added up. It was because they were habitually shite at level game state and then went on the rampage at one or two goals down.

 

[Handy Run]

He knows enough to be dangerous, the cute hoor.

He's the Ronaldo of cherry pickers. And not the fat one either.

 

[KolaKubes]

A quick one on xG. Yesterday was a good example of a few interesting aspects.

1. Influence of game state on xG

City were utterly dominant but some late token attacks from Utd, at 3-0 down, meant the match xG (City ahead by 0.7 via Opta and 1.1 vs sofascore) makes it look like a vaguely competitive game in theory.

2. Two of those big xG chances were weird

84mins - Mbeumo 0.58 (Opta) / 0.64 (Understat)
85mins - Casemiro 0.47 (Opta) / 0.45 (Understat)

Both were flying in at high speed and each player had limited hope of accurately re-directing the ball.

Does the xG model accurately reflect just how hard those two chances were?

3. xG vs post-shot xG

Because Man Utd only had 2 shots on target their post-shot xG was tiny (0.14). This is arguably closer to the truth than the actual xG.

4. Mbuemo's fantastic volley

Went from 0.02 xG to only 0.08 post-shot xG.

I'd wager around a third of PL keepers wouldn't typically save that shot. An increment of only 0.06xG feels stingy for how much his contribution made that chance better than it was before he made contact

View attachment 44616 View attachment 44608

Interesting. The other thing not captured by that stat is when Utd break four on four (or better) and conspire to overhit a pass, run offside, turn back to pass it to Bruno rather than the sort of two pass movement that used end with Rooney crashing the ball past the keeper. We miss situations that most other Premier League teams translate into goals. They are creating these more routinely now though. They also won ball back high up the pitch but then mess up simple passes to turn same into high quality chances.

There’s no point talking down the chances due to context if going to ignore city being 2-0 up, Utd pressing up the field exposing the defenders they have that can’t play a high line that resulted in simple breakaway situations. Could take them off as well as didn’t require city to do much to create them.

What stat covers the ongoing capacity of Bruno, Dalot, Shaw and others to have utter brain farts that might lead to a goal directly or more insidiously cause the corner or loss of possession that led to the goal ultimately

 

[KolaKubes]

I've been trying to tell folk around here about point one for ages. I'm still not sure they get it tbphwy. Although, Nollaig will act dumb when I know deep down he gets it but still uses the contextless stats for his schtick.

City were not “utterly dominant” before their goal. It was pretty much a case of Haaland being an intimidating presence up front and great quality from Doku being the differentiator.

It’s a good post by Philby but if there’s one thing worse than Liverpools success it’s Liverpool fans talking to us like we haven’t the much longer period of time watching dominant teams play better football.

I will say the younger lads maybe subscribe to the Ronaldo model that if you’ve a couple of world class players it’s ok if they’re cunts. That was not a good thing and led to the destruction of 15 years of club culture with Ferguson’s blessing.

 

[Cloudy O'Ranieri]

City were not “utterly dominant” before their goal. It was pretty much a case of Haaland being an intimidating presence up front and great quality from Doku being the differentiator.

It’s a good post by Philby but if there’s one thing worse than Liverpools success it’s Liverpool fans talking to us like we haven’t the much longer period of time watching dominant teams play better football.

I will say the younger lads maybe subscribe to the Ronaldo model that if you’ve a couple of world class players it’s ok if they’re cunts. That was not a good thing and led to the destruction of 15 years of club culture with Ferguson’s blessing.

Regardless of any of that, a massive portion of United's xG came at three down. City were clinical and didn't really face much threat at the other end until late on when the game was over.

 

[Philby]

It’s a good post by Philby but if there’s one thing worse than Liverpools success it’s Liverpool fans talking to us like we haven’t the much longer period of time watching dominant teams play better football.

Cheers. I genuinely wasn't trying to have a go at Utd, if that's how it was interpreted. I just spotted some interesting aspects from an xG perspective.

Not sure what your point is around watching other dominant sides in the past though. Assume it's not analytics related?

 

[KolaKubes]

Cheers. I genuinely wasn't trying to have a go at Utd, if that's how it was interpreted. I just spotted some interesting aspects from an xG perspective.

Not sure what your point is around watching other dominant sides in the past though. Assume it's not analytics related?

xG covers the number of shots from a zone that turn into goals from what I can understand but it tends to overestimate the likelihood of scoring if you get a chance to hit it from a corner. The number of players around you when you’re trying to do that has to be factored in.

My point was I think overloads that really only need a couple of well executed passes are better chances than shots from distance. I’d liken it to three rugby guys breaking away but the first lad fucks up the pass. Utd do it constantly.