xG under Ralf

[Bubz27]

Are penalties a bit of an outlier for xG?

The cross Spurs put in that led to the pen would not have resulted in a high xG, if any at all. But obviously a penalty will lead to a higher xG.

 

[sullydnl]

Are penalties a bit of an outlier for xG?

The cross Spurs put in that led to the pen would not have resulted in a high xG, if any at all. But obviously a penalty will lead to a higher xG.

There was actually a piece in The Athletic recently on that touched on this:

 

[Bubz27]

There was actually a piece in The Athletic recently on that touched on this:

Thanks, will give that a read.

 

[Pyroblazer]

There was actually a piece in The Athletic recently on that touched on this:

That's a great read and difficult to disagree with any of that. Especially considering how many players just try to get into the penalty area and dive with a small chance of actually scoring or creating a chance.

 

[cyberman]

People generally underestimate how often the average player fecks up "must score" chances.

Ronaldo's finish on the second actually brushed the goalkeeper's leg on the way in. An inch or so more off and it isn't a goal, despite seemingly being a must score chance.

But it’s the perfect ball across and the keeper is rooted to his line? If the ball is fizzed across or the keeper is rushing out the man fair enough, there’s moving parts in play but Sancho leaves it on a plate. The ball is almost stopped directly in front of goal for Ronaldo to run on to.
That’s where XG falls down for me. I understand the argument but someone missing a glorious chance in another instance doesn’t lessen what the chance was. If Ronaldo scoring or missing doesn’t affect his XG then how does seeing someone else miss a similar chance mean anything?
Explaining the process of XG doesn’t really change that imo.

 

[DomesticTadpole]

The only stat that matters is how many you put in the net, not the ones that you might have.

 

[elmo]

But it’s the perfect ball across and the keeper is rooted to his line? If the ball is fizzed across or the keeper is rushing out the man fair enough, there’s moving parts in play but Sancho leaves it on a plate. The ball is almost stopped directly in front of goal for Ronaldo to run on to.
That’s where XG falls down for me. I understand the argument but someone missing a glorious chance in another instance doesn’t lessen what the chance was. If Ronaldo scoring or missing doesn’t affect his XG then how does seeing someone else miss a similar chance mean anything?
Explaining the process of XG doesn’t really change that imo.

XG doesn't care who takes the shot, it's supposed to be the average chance of the goal being scored from that position and plenty of players feck up every game.

It's why top players tend to overperform XG, they're supposed to be better than the average.

 

[cyberman]

XG doesn't care who takes the shot, it's supposed to be the average chance of the goal being scored from that position and plenty of players feck up every game.

It's why top players tend to overperform XG, they're supposed to be better than the average.

But my argument is that fecking up such an easy chance doesn’t lessen how great of a chance that was. That’s too result orientated.

 

[elmo]

But my argument is that fecking up such an easy chance doesn’t lessen how great of a chance that was. That’s too result orientated.

And my point is that great chances aren't as great as you think it is because statistically, most chances don't turn into goals no matter how clear cut you think it is.

 

[cyberman]

And my point is that great chances aren't as great as you think it is because statistically, most chances don't turn into goals no matter how clear cut you think it is.

But they’re still expected to score? Again you’re too result orientated here. I’m saying the science behind it doesn’t make sense but the counter argument coming back to me is using that science to make a point.
I think we’re at a stalemate here so we’ll just leave it at that.

 

[bosnian_red]

XG doesn't care who takes the shot, it's supposed to be the average chance of the goal being scored from that position and plenty of players feck up every game.

It's why top players tend to overperform XG, they're supposed to be better than the average.

That 2nd part isn't actually true. Top players rack up high xG but for example, Lewandowski and Ronaldo are pretty much dead on their xG since 2013/14. Overperforming xG doesn't necessarily mean you are a top player, just means you finish clinically but that doesn't make someone the better striker.

 

[Zehner]

But my argument is that fecking up such an easy chance doesn’t lessen how great of a chance that was. That’s too result orientated.

xG exposes the illusion of threat, so to speak. Is it really a big chance only because people get excited?

 

[Zehner]

That's a great read and difficult to disagree with any of that. Especially considering how many players just try to get into the penalty area and dive with a small chance of actually scoring or creating a chance.

But penalties are part of the game. Their mere presence means that you have to defend differently in the box which leads to a more open and attacking friendly game.

 

[bosnian_red]

But it’s the perfect ball across and the keeper is rooted to his line? If the ball is fizzed across or the keeper is rushing out the man fair enough, there’s moving parts in play but Sancho leaves it on a plate. The ball is almost stopped directly in front of goal for Ronaldo to run on to.
That’s where XG falls down for me. I understand the argument but someone missing a glorious chance in another instance doesn’t lessen what the chance was. If Ronaldo scoring or missing doesn’t affect his XG then how does seeing someone else miss a similar chance mean anything?
Explaining the process of XG doesn’t really change that imo.

XG is literally the average rate that a similar chance is scored based on historical data. Penalties are the easiest to explain because they are scored roughly 75% of the time, so it gets a 0.75 xG. Technically, yes, every goal scored "improves" the xG of that chance and every shot missed reduces it. But it's by a miniscule amount because there are thousands of samples that they use.

Nothing is ever going to be 1.00 xG because a player can technically miss any chance. It's statistics. Nothing is 100%. He's missed easier chances and Lloris was close to saving that one. Understat has that goal at 0.53 xG which is pretty fair. First time shots have a higher rate of feck ups than a chance when you are moving calmly with the ball or a penalty where you are set.

 

[UNITED ACADEMY]

Are penalties a bit of an outlier for xG?

The cross Spurs put in that led to the pen would not have resulted in a high xG, if any at all. But obviously a penalty will lead to a higher xG.

I think most chances that came from 6 yard box will have high xG because you would be expected to score from that if you are given chances to shoot. That includes penalty. Spurs had good chances from 6 yard box like the Dier's header (cleared by Dalot), penalty, and Son's chance in 2nd half. Other than those three, their chances weren't that high xG.

Ronaldo's first goal is from long shot so the xG is lower than those three chances I mentioned. Ronaldo's winning header is probably the same one as Dier's one in term of xG while Ronaldo's 2nd goal is probably slightly higher than Son's chance in 2nd half (may be) since Ronaldo had easy tap in while Son had defenders closer to him.

 

[cyberman]

xG exposes the illusion of threat, so to speak. Is it really a big chance only because people get excited?

It not about excitement, it’s the fact it’s almost as good as an open goal. The perfect ball across, Loris being rooted to his line. Ronaldo being able to run onto it 2 (?) yards out and having his choice of finishes.
The stars lined up perfectly, if ever there was a perfect chance that was it. Missing it doesn’t make it lesser of a great chance, it just means you missed a great chance

 

[sullydnl]

It not about excitement, it’s the fact it’s almost as good as an open goal. The perfect ball across, Loris being rooted to his line. Ronaldo being able to run onto it 2 (?) yards out and having his choice of finishes.
The stars lined up perfectly, if ever there was a perfect chance that was it. Missing it doesn’t make it lesser of a great chance, it just means you missed a great chance

Fecking up a great chance doesn't stop it from being a great chance. But the point is that we know "great chances" regularly get missed. So if you're trying to assess the value of great chances in terms of the actually liklihood of them resulting in a goal, you have to take all those misses that "shouldn't" happen but obviously do into account.

This was a great chance. But what we often describe as "great chances" are chances that typically get scored 5-8 times out of every 10. By the standard of most chances, that is great. Even penalties which are generally the best individual chance to score you'll get in the vast majority of games still get missed 1 in 4 times.

You're saying this was a near perfect chance but even ignoring all the other ways it could have gone wrong, Ronaldo was still only a few centimeteres from missing it given it the shot went in off the keeper's foot. Even without looking at stats at all, the fact that there was such a narrow margin between this supposedly near-perfect chance being scored or missed highlights exactly why its statistical value isn't higher.

 

[Withnail]

Thanks. So it is complicated to work out then. I don’t see how it helps, maybe only for the xG for the whole team. Rather than individuals.

It's really to be used over a longer period of time than one game.

 

[arnie_ni]

But they’re still expected to score? Again you’re too result orientated here. I’m saying the science behind it doesn’t make sense but the counter argument coming back to me is using that science to make a point.
I think we’re at a stalemate here so we’ll just leave it at that.

But they aren't "expected" to score, otherwise it would be 1

 

[Classical Mechanic]

Interesting article on attempting to put a value on the 'art of backwards passing'

https://analyticsfc.co.uk/blog/2022/03/31/measuring-the-art-of-backward-passing/

It does a scouting exercise on young players that score highly on the metric in The Championship. Jimmy Garner is mentioned but they pick out Alex Scott as the one to watch.

 

[Rossa]

If an attack ends up with eight shots on target, does that mean theres an accumulated xg? If say the average chance is an xg of 0,3, that means that attack alone will accumulate an xg of 2,4? You could still only score once considering it’s the same attack… same with a penalty; if you finish one on one with the goalie, and the goalie saves but the attacker is fouled for the rebound thus being awarded a penalty, is there an accumulated xg? Again, the attacker could only score once.

 

[TheRedDevil'sAdvocate]

If an attack ends up with eight shots on target, does that mean theres an accumulated xg? If say the average chance is an xg of 0,3, that means that attack alone will accumulate an xg of 2,4? You could still only score once considering it’s the same attack… same with a penalty; if you finish one on one with the goalie, and the goalie saves but the attacker is fouled for the rebound thus being awarded a penalty, is there an accumulated xg? Again, the attacker could only score once.

No, they'll try to decipher how likely is it that the chance doesn't get converted, instead.

1-0.3=0.7xG with 8 shots. So, 0.7x0.7x0.7x0.7x0.7x0.7x0.7x0.7=0.057. Which tell us that there's a 5.76% (6%) chance that the team in question doesn't score from these 8 shots. Subsequently, the attack will get 0.94xG

Same with penalties or other similar situations.

 

[Adam-Utd]

Interesting article on attempting to put a value on the 'art of backwards passing'

https://analyticsfc.co.uk/blog/2022/03/31/measuring-the-art-of-backward-passing/

It does a scouting exercise on young players that score highly on the metric in The Championship. Jimmy Garner is mentioned but they pick out Alex Scott as the one to watch.

It's an interesting discussion, but I think you could only look at each individual passage of play and ask, was passing backwards better than passing forward in that moment?

Sometime players pass backwards when pressured or 'pass the buck' and put their other mates into trouble. In that moment it might be better to try and dribble out of pressure or make a quick turn and find a team mate in a forward area.

That's why football is such a difficult game to analyse with stats, but also makes decision making on the pitch so vital. Being fast in the brain will always give you a great chance even in a running sport like football.

I would say that backward passing from forward areas can be a great thing, as long as you have forward runners from midfield. That's why target men became so popular in the first place.

 

[VidaRed]

It's an interesting discussion, but I think you could only look at each individual passage of play and ask, was passing backwards better than passing forward in that moment?

Sometime players pass backwards when pressured or 'pass the buck' and put their other mates into trouble. In that moment it might be better to try and dribble out of pressure or make a quick turn and find a team mate in a forward area.

That's why football is such a difficult game to analyse with stats, but also makes decision making on the pitch so vital. Being fast in the brain will always give you a great chance even in a running sport like football.

I would say that backward passing from forward areas can be a great thing, as long as you have forward runners from midfield. That's why target men became so popular in the first place.

Maguire does that repeatedly with our full backs and mcfred and we get caught out more often than not.

 

[Classical Mechanic]

It's an interesting discussion, but I think you could only look at each individual passage of play and ask, was passing backwards better than passing forward in that moment?

Sometime players pass backwards when pressured or 'pass the buck' and put their other mates into trouble. In that moment it might be better to try and dribble out of pressure or make a quick turn and find a team mate in a forward area.

It's about finding players that consistently start progresssive moves with back passing. If the player passes to their team mate and puts them in trouble then it's unlikely that any progression will be achieved.

I think the idea is that the players who score consistently highly on the metric have intelligent movement when showing for the ball which draws opposition players to them and then opens up other avenues for ball progression through their team mates.

 

[Chesterlestreet]

Same with penalties or other similar situations.

Was thinking about this the other day.

NN shoots from the penalty spot (xG 0.76), misses but scores on the rebound (say, xG 0.69). The xG for that situation will be (as I understand it) 0.9256.

Which seems to make sense at first glance: the chance of scoring from a penalty + rebound (from a good position) is considerably higher than scoring from the penalty itself (alone). This is in line with how most fans will perceive the situation - and I'm sure that if you analyzed a large sample of such situations (missed pen + rebound taken from good position), the numbers would certainly be in the ballpark.

I'm less certain about a situation where a single attack includes, say, three finishes from xG positions 0.18, 0.21 and 0.29. In that situation the official xG might be artificially high - or?

To be precise, I suspect that a specific analysis of similar (enough) situations where an attack results in 3 finishes from similar (enough) positions may actually yield a (significantly) lower number.

The xG in the situation in question (three finishes: 0.18, 0.21 and 0.29) will be calculated from overall/general data (positions and other relevant factors) - not from a statistically meaningful number of similar situations (as such).

(The same positions (and other relevant factors), but not the same situation - the latter being a very particular combination of positions and other relevant factors).

 

[bosnian_red]

Maguire does that repeatedly with our full backs and mcfred and we get caught out more often than not.

I know it's shit on Maguire season, but the last problem he has is passing the buck with a shorter pass to someone else instead of trying the pass himself. I'd go as far as to criticize him for trying to be overly progressive, all the time. He sits on the ball for ages and looks for a pass and just forces them through to the wingers, which counts as a progressive pass but quite often doesn't lead to much (as a simpler, quicker pass like what Varane does would've worked better).

 

[TheRedDevil'sAdvocate]

Was thinking about this the other day.

NN shoots from the penalty spot (xG 0.76), misses but scores on the rebound (say, xG 0.69). The xG for that situation will be (as I understand it) 0.9256.

Which seems to make sense at first glance: the chance of scoring from a penalty + rebound (from a good position) is considerably higher than scoring from the penalty itself (alone). This is in line with how most fans will perceive the situation - and I'm sure that if you analyzed a large sample of such situations (missed pen + rebound taken from good position), the numbers would certainly be in the ballpark.

I'm less certain about a situation where a single attack includes, say, three finishes from xG positions 0.18, 0.21 and 0.29. In that situation the official xG might be artificially high - or?

To be precise, I suspect that a specific analysis of similar (enough) situations where an attack results in 3 finishes from similar (enough) positions may actually yield a (significantly) lower number.

The xG calculated from the situation in question (three finishes: 0.18, 0.21 and 0.29) will be generated from overall/general data (positions and other relevant factors) - not from a statistically meaningful number of similar situations (as such).

I don't have the statistics to tell you whether you're right or wrong. What i do know, is that they're trying to factor the reaction of the keeper in the variables. Which i suspect can play a role in this type of situation.

Nevertheless, in your example, we get an 0.54xG. That's a chance with >50% probability of being converted. It's not even an excellent chance, it's a glorious chance to score a goal. It's not low, at all.

It's a work in progress. I'm sure they'll add to it. But the basic model has already proved to be a very useful tool.

 

[Chesterlestreet]

But the basic model has already proved to be a very useful tool.

Oh, yes - I'm not anti-xG at all (to be clear).

Just find it interesting to look at possible flaws, etc.

 

[Rossa]

No, they'll try to decipher how likely is it that the chance doesn't get converted, instead.

1-0.3=0.7xG with 8 shots. So, 0.7x0.7x0.7x0.7x0.7x0.7x0.7x0.7=0.057. Which tell us that there's a 5.76% (6%) chance that the team in question doesn't score from these 8 shots. Subsequently, the attack will get 0.94xG

Same with penalties or other similar situations.

Cheers mate! That makes much more sense!

 

[sullydnl]

Stolen from twitter.

 

[The Corinthian]

Anyone have the xG for the game today?

 

[bucky]

 

[Iker Quesadillas]

xG model looking real bad with these comparisons.

 

[sullydnl]

xG model looking real bad with these comparisons.

Really? It's pretty much what I would have expected from watching us in those periods.

 

[sullydnl]

Sorted by xG difference/90. Very little difference between that and the actual table.

Not sure we've ever been better than 4th in xGD or xPTS post-SAF in any models I've seen? Which seems a pretty accurate reflection of our general performance levels over the last hell decade.

 

[Grande]

Really? It's pretty much what I would have expected from watching us in those periods.

Likewise.

 

[sullydnl]

On the subject of xG generally, check out the difference between Bayern and the rest.

Not what you'd call competitive.

 

[The Corinthian]

Thank you sir!

 

[Iker Quesadillas]

Really? It's pretty much what I would have expected from watching us in those periods.

It says United should have gotten the 2nd most points under Ole when he got the least amount of points. Kind of a whiff.
(Not really the model being bad, just that the differences between managers aren't substantial enough)