*** OMT - Tottenham vs Wolves ***

[alekaras]

That’s where I’m struggling with the methodology then. That suggests that that chance only goes in 4 times out of 100. I’d have serious doubts about that.

I can understand how one would think that is a chance that should go in more often, but the xG is based on thousands of data points that indicate that's the probability. So it's pretty objective in that regard.

That being said, a world class player may outperform the xG and get it in the net more than 4 times out of 100, but it's not going to be 20 or 30 times. More like 6 or 7. So still low.

 

[Pirate55]

In the first three games of last season unfpder Nuno, didn't we outperform our XG? Didn't we all therefore know we couldn't keep that up. And so it proved.

 

[Bullet]

That’s where I’m struggling with the methodology then. That suggests that that chance only goes in 4 times out of 100. I’d have serious doubts about that.

It’s because most headers from 14 yards out in the centre of the goal are hotly contested by defenders.
XG doesn’t take into account whether it is a rubbish chance from a corner with huge challenges piling in… or a free header on a plate

 

[braineclipse]

Might seem an odd question re: XG but does it take into account shots not taken? When someone should have shot but takes too many touches or squares the ball badly and the opportunity is lost?

It doesn't.

And it's one of the reasons why looking at xG numbers for a single game should be done with caution.

I think xG definitely has a place. It certainly adds some context to the “Shots” stat which sometimes leaves me wondering if I’m watching the same game when I see it in isolation.

However, I’m amazed that header that went just wide was less than .1. That looked a really good chance to me.

It was a really good header imo, more than really good chance.

For context, that was about as close as Wolves got. Kane's header into the crossbar was closer and imo a better chance.

Imo we created better chances than them. That looks the case looking at xG too.

 

[braineclipse]

In the first three games of last season unfpder Nuno, didn't we outperform our XG? Didn't we all therefore know we couldn't keep that up. And so it proved.

I remember quite clearly arguing on here that it was unsustainable.

The contrast to what we're doing now is clear as day imo.

 

[DeanoAustin]

It’s because most headers from 14 yards out in the centre of the goal are hotly contested by defenders.
XG doesn’t take into account whether it is a rubbish chance from a corner with huge challenges piling in… or a free header on a plate

That I can buy then because you have a decent challenge on you and I can imagine only 4 of 100 go in. However the wolves guy had almost a free header because our lad (Davies I think) did little to put him off. That made it a better chance than 4 in 100 IMO.

 

[Spur of the moment]

That was a training ground set piece I believe. I thought our best player was Perisic. Kane fluffed a good chance, as did son, in the penalty box. Both should have scored imo.



Sent from my SM-T865 using Fapatalk

So you believe strikers should score every decent chance they get? Tell me one striker who does. FFS get real!

 

[alekaras]

That I can buy then because you have a decent challenge on you and I can imagine only 4 of 100 go in. However the wolves guy had almost a free header because our lad (Davies I think) did little to put him off. That made it a better chance than 4 in 100 IMO.

I believe xG does take into consideration whether someone is unmarked, or has 5 defenders around him. Even unmarked, however, it's a chance that will not have much success. It just feels like it should, but if you look at a similar chance across hundreds or thousands of other games the result is no goal 96 times out of 100. As I said, what is not factored in is player quality, so a world class player will outperform the chance's xG. But not by a huge amount.

 

[alekaras]

It doesn't.

And it's one of the reasons why looking at xG numbers for a single game should be done with caution.

I'm sure that this is something that will be coming eventually. For example, player A squares it to player B who is unmarked in the 6-yard box, but player B swings and completely misses the ball. Won't show up in the xG numbers, but you have to count it as a high probability scoring opportunity somehow. Same with a slightly misplaced pass (a bit in front or a bit behind a player in scoring position). The algorithms to calculated these types of chances will probably be much more complex, but someone will figure them out.

And don't forget that all of the above rely on the availability of a wealth of data points from past games. As the years go by and the data sets get richer, even player quality could be factored in to the xG of a chance. Still a ways to go, but exciting to have more metrics to gauge a game in addition to the "eye test."

 

[DeanoAustin]

I believe xG does take into consideration whether someone is unmarked, or has 5 defenders around him. Even unmarked, however, it's a chance that will not have much success. It just feels like it should, but if you look at a similar chance across hundreds or thousands of other games the result is no goal 96 times out of 100. As I said, what is not factored in is player quality, so a world class player will outperform the chance's xG. But not by a huge amount.

There's something missing here. There is no way that chance, in identical circumstances, doesn't get converted more times than 4 in 100 IMO. I don't doubt the stats, just the parameters under which those stats are compiled which would suggest to me that you can't definitively say that a chance is "x% of an xG" because there are a lot of variables in the equation.

 

[johnola]

I'm sure that this is something that will be coming eventually. For example, player A squares it to player B who is unmarked in the 6-yard box, but player B swings and completely misses the ball. Won't show up in the xG numbers, but you have to count it as a high probability scoring opportunity somehow. Same with a slightly misplaced pass (a bit in front or a bit behind a player in scoring position). The algorithms to calculated these types of chances will probably be much more complex, but someone will figure them out.

And don't forget that all of the above rely on the availability of a wealth of data points from past games. As the years go by and the data sets get richer, even player quality could be factored in to the xG of a chance. Still a ways to go, but exciting to have more metrics to gauge a game in addition to the "eye test."

Circular error: using xg to assess player quality then using player quality to finesse xg.

 

[braineclipse]

There's something missing here. There is no way that chance, in identical circumstances, doesn't get converted more times than 4 in 100 IMO. I don't doubt the stats, just the parameters under which those stats are compiled which would suggest to me that you can't definitively say that a chance is "x% of an xG" because there are a lot of variables in the equation.

I agree. Even looking at xG numbers for a single game as a factual thing about chances created is imo not a good idea. For a single shot it will definitely not be that accurate.

Circular error: using xg to assess player quality then using player quality to finesse xg.

I think it would be plausible. Some kind of adjusted xG that takes player quality into account. So xG numbers and adjusted xG numbers in two different, but related models.

However to really get to grips with that statistically you'd want to see how players have done in similar chances rather than overall. Quite a few players are really good at some types of shots, not so much at others.

That would however take a sample size well above what we can realistically get for almost all players.

 

[Bishop]

I agree. Even looking at xG numbers for a single game as a factual thing about chances created is imo not a good idea. For a single shot it will definitely not be that accurate.



I think it would be plausible. Some kind of adjusted xG that takes player quality into account. So xG numbers and adjusted xG numbers in two different, but related models.

However to really get to grips with that statistically you'd want to see how players have done in similar chances rather than overall. Quite a few players are really good at some types of shots, not so much at others.

That would however take a sample size well above what we can realistically get for almost all players.

It would kind of defeat the purposes of xG if you adjusted as per the players finishing ability. The main thing you can gain from looking at xG as per an individual player is it will show in general how good their finishing is relative to the mean.

We know that Kane and Son are excellent finishers using our eyes but looking at their xG number we can see they finish a significantly higher percentage of their chances versus many other PL strikers.

Sent from my XQ-BC72 using Fapatalk

 

[SpurMeUp]

I didn't actually think we were that bad first half. After a decent start against chelsea we lost control because we lacked some accuracy in the face of their press. Against Wolves we lost control too at a similar point (around 20 mins, Chelsea a few minutes earlier). In this game we lost control due to a number of fouls conceded. We need a little more cool - both tackling and retaining the ball while attacking. On that note Perisic is cool personified. One of those players who always seems to have time, even when he doesn't. An awesome debut. And Richarlison too was impressive in his home debut. A better player than I gave him credit. Both just missing scoring a goal.

The difference in the second half was the intensity of our press. First half we were fast running to close them down, second half sprinting to get the ball back as though lives depended on it. And it made the difference. Our press isn't as effective as Chelsea or Southampton. We're not as committed moving as a unit. Its more reactionary. But it keeps our back lines intact.

Credit Wolves, fresh and organised they played well. I think teams are actively smothering Benancur and there is a loss of quality playing out from the back when they do. Would like to see Lenglay, Davies for me doesn't quite offer enough. Thought we edged both halves, but need a bit more.

 

[braineclipse]

It would kind of defeat the purposes of xG if you adjusted as per the players finishing ability. The main thing you can gain from looking at xG as per an individual player is it will show in general how good their finishing is relative to the mean.

We know that Kane and Son are excellent finishers using our eyes but looking at their xG number we can see they finish a significantly higher percentage of their chances versus many other PL strikers.

Sent from my XQ-BC72 using Fapatalk

"in two different, but related models"

 

[Bedfordspurs]

Wolves splayed with NO striker so had more bodies in midfield but lacked a goal threat
Chelsea plays with NO strikers so had more bodies in midfield but no goal threat… a striker would have hurried one of the crosses they had
If teams want to play like that, they have to win the midfield battle. If they don’t win that battle they lose the game

 

[Bedfordspurs]

"in two different, but related models"

It doesn’t take into account weather, time of the shot in the match, what end the team are attacking…
That’s just a few BOG variables that matter