xG stats

[whelk]

I take the p1ss a bit about the new xG stat but actually is one of the more meaningful stats in the days of saturation. Certainly more interesting than passes completed, even possession.

 

This table https://understat.com/league/EPL

shows we should have expected to be 8pts ahead of Burnley.

Arguably that isn’t really on the manager as down to player opportunism.

 

I am not saying they are the be all an end all of stats but do highlight over and under performance. We are also only side to have out xG’d Man City this season.

[doddisalegend]

I take the p1ss a bit about the new xG stat but actually is one of the more meaningful stats in the days of saturation. Certainly more interesting than passes completed, even possession.

 

This table https://understat.com/league/EPL

shows we should have expected to be 8pts ahead of Burnley.

Arguably that isn’t really on the manager as down to player opportunism.

 

I am not saying they are the be all an end all of stats but do highlight over and under performance. We are also only side to have out xG’d Man City this season.

 

so if I'm reading that right Crystal palace would actually be above us in the table if they had scored all their xG

[Noodles34]

There is only one stat that counts.

[Mattio]

https://www.youtube.com/watch?v=w7zPZsLGK18https://youtu.be/w7zPZsLGK18

[verlaine1979]

so if I'm reading that right Crystal palace would actually be above us in the table if they had scored all their xG

 

A statistical model that suggests a team with a goal difference of -17 and only 10 goals scored should actually have a goal difference of -1 and 23 goals scored would appear at first glance to have some kinks that still need working out.

[Cartman]

A statistical model that suggests a team with a goal difference of -17 and only 10 goals scored should actually have a goal difference of -1 and 23 goals scored would appear at first glance to have some kinks that still need working out.

 

Or that their finishing is beyond abysmal.

[simo]

I'm still baffled by expected goals , can someone show their working out ?

[GGalpin]

I'm still baffled by expected goals , can someone show their working out ?

 

It's basically quality of chances. So a 25 yard shot from a tight angle is 'less likely' to go in than a tap in in the middle of the goal two yards out. If a chance is 10% 'likely' to go in, it's got a rating of 0.1. Ten of those a game, and the xG for those chances is 1.0.

 

It's basically a stat that shows the "we created the better chances". So Swansea at home opening game was 0-0, but the chances we created should be expected to have brought 2 goals. When you think Yoshida, JWP and Tadic missed big chances that game, that metric shows it. https://understat.com/match/7122

[The9]

so if I'm reading that right Crystal palace would actually be above us in the table if they had scored all their xG

 

Palace against us is the second biggest difference between xG and actual goals in Prem history - they should have scored 3 and scored none. The "record" was Arsenal v Man U recently, their xG was around 5.

Edited 12 December, 2017 by The9

[GGalpin]

This table is quite interesting- https://understat.com/league/EPL/2016

 

Saints scored 41 goals, but the xG was 50.31, meaning we scored nine less than what the chances would suggest we should have scored, and the 43.21 for the goals against expected meant we conceded 4 more than we should have expected.

 

You look at our 'expected points' and we were above Everton, who scored 12 more than their chances should yield.

 

Plenty have pointed out the difference being Lukaku but these stats show it even more, and with him not being replaced and someone not taking the chances Everton shouldn't reasonably expect to have taken, it's little wonder they've regressed.

[GGalpin]

so if I'm reading that right Crystal palace would actually be above us in the table if they had scored all their xG

 

Possibly. But looking at why they've not taken those chances (Benteke being their only real striker and hasn't scored since May) you could argue if they keep up the quality of chances created and they get a decent striker in January who can take those chances, then they should be safe.

[The9]

Possibly. But looking at why they've not taken those chances (Benteke being their only real striker and hasn't scored since May) you could argue if they keep up the quality of chances created and they get a decent striker in January who can take those chances, then they should be safe.

 

Sound logic, (the kind of stuff I was spouting about us last season when we ended up 8th somehow), though even last night when they scored against Watford to equalise they took 3 shots to score the first goal - the initial one was from outside the box so fairly low xG, but then Sakho missed from 6 yards right in front of goal before scoring the rebound of his own shot. So even that single goal was probably low compared to the combined xG of the three shots. They did then score with their next chance too, which will have improved their differential a bit.

[simo]

It's basically quality of chances. So a 25 yard shot from a tight angle is 'less likely' to go in than a tap in in the middle of the goal two yards out. If a chance is 10% 'likely' to go in, it's got a rating of 0.1. Ten of those a game, and the xG for those chances is 1.0.

 

It's basically a stat that shows the "we created the better chances". So Swansea at home opening game was 0-0, but the chances we created should be expected to have brought 2 goals. When you think Yoshida, JWP and Tadic missed big chances that game, that metric shows it. https://understat.com/match/7122

Cheers for the breakdown , Think I get it .

[The9]

It's actually quite an interesting stat. Be interesting to see the difference between rich top teams and bottom in terms of the average XG to average goals, compared to likes of us.

 

Would show whether it's creating chances of scoring them that's the biggest issue. Think we still have a worrying lack of both at the moment. But might also stop the over praising of forwards who score regularly for big clubs, if it's clear they're getting 3 or 4 good chances a game.

 

Always wondered what would happen if you stuck an average forward like Austin, Wood, Vokes etc up front for Man Utd or Chelsea.

 

There's a reason the top strikers are at the top clubs. The amount of time it takes for the meritocracy to kick in at the top level is shorter than ever before. Long's xG compared to goals scored has been rubbish for a season or two, but that's not the stat we signed him for anyway.

 

This is a nice graphical tool for looking at player (and chance) xG for 2017/18, there are previous season versions too - shows every chance this season (not sure how up to date) along with the location, type of shot, xG, player who took the chance, player's xG, and the opposition. There's also a summary of "best players" by pure xG rather than actual goals (Kane, Lukaku, Jesus, Salah, Lacazette).

 

https://public.tableau.com/profile/paul.riley#!/vizhome/PremierLeaguexGandShotMap2017-18/PremierLeaguexGShotDashboard2017-18

[The9]

Current Saints players' xG (the Bournemouth game is on here) - which only show how many they SHOULD have scored from the efforts on goal taken, actuals in brackets and calculation of goals scored compared to xG on the right. I think the amount with less than xG says a lot.

 

Name;xG;(Goals ["G"]); G-xG

 

Davis 0.772 (3) +2.228

Austin 2.259 (4) +1.741

Gabbiadini 1.57 (3) +1.43

Pied 0.03389 (0) -0.03389

Hojbjerg 0.037 (0) -0.037

Boufal 1.099 (1) -0.099

Stephens 0.3397 (0) -0.3397

Lemina 0.364 (0) -0.364

Ward-Prowse 0.4904 (0) -0.4904

Bertrand 0.613 (0) -0.613

Cedric 0.619 (0) -0.619

Hoedt 0.656 (0) -0.656

Tadic 3.040 (2) -1.040

Romeu 2.156 (1) -1.156

Redmond 1.338 (0) -1.338

Van Dijk 1.368 (0) -1.368

Long 1.432 (0) -1.432

Yoshida 2.506 (1) -1.506

[The9]

FWIW

Abdoulaye Doucore 1.776 (5) +3.8224

Chris Wood 2.160 (4) +1.94

Sam Vokes 1.158 (3) +1.842

Harry Kane 8.2 (10) +1.80

 

Steven Davis = better finisher than Harry Kane.

Edited 13 December, 2017 by The9

[Pamplemousse]

I take the p1ss a bit about the new xG stat but actually is one of the more meaningful stats in the days of saturation. Certainly more interesting than passes completed, even possession.

 

This table https://understat.com/league/EPL

shows we should have expected to be 8pts ahead of Burnley.

Arguably that isn’t really on the manager as down to player opportunism.

 

I am not saying they are the be all an end all of stats but do highlight over and under performance. We are also only side to have out xG’d Man City this season.

 

Came across that website the other day. The one thing that stood out to me was Crystal Palace's xG.

[nta786]

Came across that website the other day. The one thing that stood out to me was Crystal Palace's xG.

 

Tbf even with our game against us they could have scored a couple.

 

My only query is what xG do they give for a penalty?

Is it 1? Is it 0.8? How do they constitute that for penalties

[GGalpin]

Tbf even with our game against us they could have scored a couple.

 

My only query is what xG do they give for a penalty?

Is it 1? Is it 0.8? How do they constitute that for penalties

 

0.76 or a 76% chance, judging by this. https://understat.com/match/7130

[whelk]

0.76 or a 76% chance, judging by this. https://understat.com/match/7130

 

Surely less for Benteke?

[Saint-Fred]

What a crock of ****. XG ignores the fact that part of the game is actually putting the ball in the net. How many chances or how good they were is irrelevant,

The Kelvin game against Leeds would have been one of the most one sided XG's in history...however who cares..we got the three poinrs and that's all that matters.

[whelk]

What a crock of ****. XG ignores the fact that part of the game is actually putting the ball in the net. How many chances or how good they were is irrelevant,

The Kelvin game against Leeds would have been one of the most one sided XG's in history...however who cares..we got the three poinrs and that's all that matters.

 

It doesn’t ignore it you simpleton. Clearly how many and what type of chances you create and concede matters.

But you have been smart and realised people are really hoping the traditional points system will be replaced by xG eventually. Don’t forget to campaign against them with the only one thing matters argument.

[Dangermouth]

It doesn’t ignore it you simpleton. Clearly how many and what type of chances you create and concede matters.

But you have been smart and realised people are really hoping the traditional points system will be replaced by xG eventually. Don’t forget to campaign against them with the only one thing matters argument.

 

Does the system factor in e.g. if it falls to CA rather than SL? If it's on the wrong foot? If he's tackled from behind or the side, etc? It might do some of that, but not all and it isn't detailed enough and arguably a better measure is how many goals does a player score per game e.g. Jesus/Aguero - a lot; Steven Davis not a lot. But we know that anyway without this; which really only measures (based on what it's been decided is the measurement) whether a chance is 'good' or not and as we well know a player (e.g. JRod, SL) can quite happily miss an open goal from 3 yards out or PEH can continue with his rugby conversion training. Really it's an attempt to be clever that fails as much because it doesn't tell you much more than what you can see when you watch a game and might actually confuse matters because it will give a higher 'rating' to a chance that has a statistically (ish; I'm still unconvinced this is proper stats) low chance of conversion e.g. an 80 yard chip that is scored and doesn't recognise other factors e.g. that when Tadic steps up to take a penalty you're as likely to wonder if it will reach the goal line as whether or not he can put it where the keeper isn't.

 

It all still boils down to who has the best (and most confident) players.

[Saint-Fred]

It doesn’t ignore it you simpleton. Clearly how many and what type of chances you create and concede matters.

But you have been smart and realised people are really hoping the traditional points system will be replaced by xG eventually. Don’t forget to campaign against them with the only one thing matters argument.

 

Way to put words in someone's mouth! Clever!!!

if XG matters so much that you really believe changing the person on the end of the chance will turn chances into goals, then You are ignoring the fact that changing that individual will change any build up of the play they are involved in ( it's a team game after all) and therefore change the XG and the chances created.

[The9]

Does the system factor in e.g. if it falls to CA rather than SL? If it's on the wrong foot? If he's tackled from behind or the side, etc? It might do some of that, but not all and it isn't detailed enough and arguably a better measure is how many goals does a player score per game e.g. Jesus/Aguero - a lot; Steven Davis not a lot. But we know that anyway without this; which really only measures (based on what it's been decided is the measurement) whether a chance is 'good' or not and as we well know a player (e.g. JRod, SL) can quite happily miss an open goal from 3 yards out or PEH can continue with his rugby conversion training. Really it's an attempt to be clever that fails as much because it doesn't tell you much more than what you can see when you watch a game and might actually confuse matters because it will give a higher 'rating' to a chance that has a statistically (ish; I'm still unconvinced this is proper stats) low chance of conversion e.g. an 80 yard chip that is scored and doesn't recognise other factors e.g. that when Tadic steps up to take a penalty you're as likely to wonder if it will reach the goal line as whether or not he can put it where the keeper isn't.

 

It all still boils down to who has the best (and most confident) players.

 

The stats show the xG of the average chance in similar circumstances, whether it is then scored or not by the player gives a player rating in comparison to it. I used a very simple pure goals vs xG rating above, before Yoshida scored and probably moved himself up the list a bit. Suffice to say Tadic airkicking from a yard out was not entirely unpredictable either.

 

It's fair to say that the models can be developed further, but the entire point of it is putting a value on each of those chances, and each of those players, and each of those goals precisely because it rates the things that people can see and gives a tangible score to them. It is based on a massive database of efforts on goal dating back years, so an 80 yard chip will be rated as a very low xG precisely because evidence has shown it is very unlikely to score.

 

xG of an "average" penalty is 76% (0.76), but Tadic has his own penalty score/miss stats to show how likely he is to score a penalty. Obviously because a penalty is a fixed match position, it's easier to compare than the wide variety of other chances in a match.

[The9]

Way to put words in someone's mouth! Clever!!!

if XG matters so much that you really believe changing the person on the end of the chance will turn chances into goals, then You are ignoring the fact that changing that individual will change any build up of the play they are involved in ( it's a team game after all) and therefore change the XG and the chances created.

 

There are models for calculating how well players move the ball from low positions of xG towards positions of high xG as well. Though as xG only measures the chance itself, your argument about the creation of that chance is moot. xG also isn't about changing who is on the end of the chance, that's just an activity that might happen as a result of the data existing.

 

Man City have been using this data for years (they even shared their data from 2011/12 to encourage amateurs to build statistical models), and the first time I heard about it was when Aguero's title-winning goal was used as evidence - he took the ball an extra couple of steps towards the goal-line (even though he was wide of the goal), which hugely increased the likelihood of him scoring compared to his original shooting opportunity.

 

City's metrics guys were keen to put that down to Aguero's understanding of xG and the concepts of goal probability being coached into him rather than just that "striker's instinct" stuff which often leads to inefficient shots, like most of Redmond's, especially when Austin is in the box.

Edited 15 December, 2017 by The9
Added URL to the City data-sharing article

[Saint-Fred]

The stats show the xG of the average chance in similar circumstances, whether it is then scored or not by the player gives a player rating in comparison to it. I used a very simple pure goals vs xG rating above, before Yoshida scored and probably moved himself up the list a bit. Suffice to say Tadic airkicking from a yard out was not entirely unpredictable either.

 

It's fair to say that the models can be developed further, but the entire point of it is putting a value on each of those chances, and each of those players, and each of those goals precisely because it rates the things that people can see and gives a tangible score to them. It is based on a massive database of efforts on goal dating back years, so an 80 yard chip will be rated as a very low xG precisely because evidence has shown it is very unlikely to score.

 

xG of an "average" penalty is 76% (0.76), but Tadic has his own penalty score/miss stats to show how likely he is to score a penalty. Obviously because a penalty is a fixed match position, it's easier to compare than the wide variety of other chances in a match.

 

Obviously Benteke didn't read his own XG factor then...lol

[Saint-Fred]

There are models for calculating how well players move the ball from low positions of xG towards positions of high xG as well. Though as xG only measures the chance itself, your argument about the creation of that chance is moot. xG also isn't about changing who is on the end of the chance, that's just an activity that might happen as a result of the data existing.

 

Man City have been using this data for years (they even shared their data from 2011/12 to encourage amateurs to build statistical models), and the first time I heard about it was when Aguero's title-winning goal was used as evidence - he took the ball an extra couple of steps towards the goal-line (even though he was wide of the goal), which hugely increased the likelihood of him scoring compared to his original shooting opportunity.

 

City's metrics guys were keen to put that down to Aguero's understanding of xG and the concepts of goal probability being coached into him rather than just that "striker's instinct" stuff which often leads to inefficient shots, like most of Redmond's, especially when Austin is in the box.

 

Sorry in one word are you saying Aguero took extra touches because of his understanding of XG?

[Saint86]

Sorry in one word are you saying Aguero took extra touches because of his understanding of XG?

 

It is obviously not because of his understanding of expected goals though!! Its because he is one of the worlds best strikers and knows instinctively how to score. Expected goals is in effect a statistical average of all players abilities to score. Aguero does it as one of the best, ergo his play fits their pattern. The end.

[Saint-Fred]

It is obviously not because of his understanding of expected goals though!! Its because he is one of the worlds best strikers and knows instinctively how to score. Expected goals is in effect a statistical average of all players abilities to score. Aguero does it as one of the best, ergo his play fits their pattern. The end.

 

I am not sure that is what The9 said?

[Pamplemousse]

On a slightly related note, I remember a chart from last season which showed which teams created chances and compared them to how many were taken by teams. We were one of, if not the most wasteful team. It would be interesting to see it for this season, if one exists.

[Charlie Wayman]

A statistical model that suggests a team with a goal difference of -17 and only 10 goals scored should actually have a goal difference of -1 and 23 goals scored would appear at first glance to have some kinks that still need working out.

 

or put in another way... complete and utter tosh!

 

We are where we are because of who we are. The world is full of what-ifs and that includes each one of us.

 

Unless you believe in a mutiverse where every permutation of happenings is feasible the hear and now is all that matters in our solitary universe. We are all stuck with this one and this Southampton team and Redmond & Ward Prowse.

[The9]

Sorry in one word are you saying Aguero took extra touches because of his understanding of XG?

 

That's what the (slightly biased) Man City metrics guys said when asked about it.

[The9]

I am not sure that is what The9 said?

 

That's because it's not what I said. What I said was his natural skills have been enhanced by coaching based on xG (and other stats), which Man City's coaches have confirmed. It was years ago though. See a couple of posts below for a link that mentions it.

Edited 15 December, 2017 by The9
Found a link

[The9]

Oh, and I found the article with the stats that Tadic was the most wasteful Prem player last season: http://www.bbc.co.uk/sport/football/40699431

 

Player Goals Expected goals Expected goals difference

Tadic (Southampton) 3 6.66 -3.66

 

Only includes players with a minimum of 50 shots

[The9]

Also, if "Redknapp hates it" isn't a good enough reason for you, what is?

 

It's an article about the development of ProZone into xG, also has lots of Saints references, Woodward, Wilson etc and ends up with this little snippet:

 

"Two minutes later, City's attacker Sergio Aguero received the ball on the edge of the box, in a position to shoot. According to Prozone's goal-expectation model, he had a 12 percent chance of scoring. Instead of shooting, he went around a defender to a corner of the penalty area and, from a spot where he had a 19 percent chance of scoring, slotted the ball past the keeper. By the time Wilson landed at Gatwick, the news ticker running across the TV screens was saying that Manchester City were the new champions"

 

There's enough in the rest of the article about affecting decision-making to support a reasonable assumption that the Aguero xG stuff above is accurate, though I've seen an actual interview about it somewhere.

Edited 15 December, 2017 by The9

[shurlock]

Sad!

[The9]

Sad!

 

[Saint-Fred]

Also, if "Redknapp hates it" isn't a good enough reason for you, what is?

 

It's an article about the development of ProZone into xG, also has lots of Saints references, Woodward, Wilson etc and ends up with this little snippet:

 

"Two minutes later, City's attacker Sergio Aguero received the ball on the edge of the box, in a position to shoot. According to Prozone's goal-expectation model, he had a 12 percent chance of scoring. Instead of shooting, he went around a defender to a corner of the penalty area and, from a spot where he had a 19 percent chance of scoring, slotted the ball past the keeper. By the time Wilson landed at Gatwick, the news ticker running across the TV screens was saying that Manchester City were the new champions"

 

There's enough in the rest of the article about affecting decision-making to support a reasonable assumption that the Aguero xG stuff above is accurate, though I've seen an actual interview about it somewhere.

 

But surely the crux of it is whether Sergio did it naturally, as that's what he does, or if someone "coached" it into him because of knowledge of XG..Surely if he is coached to get in a more likely position to score every time he gets the ball, he will never shoot from outside the area, for example?

[Convict Colony]

The question I want to know is if xG's stats are better than MLG's stats ?

 

Who is the most trustworthy source :-)

[Dangermouth]

But surely the crux of it is whether Sergio did it naturally, as that's what he does, or if someone "coached" it into him because of knowledge of XG..Surely if he is coached to get in a more likely position to score every time he gets the ball, he will never shoot from outside the area, for example?

 

You know full well he did it naturally. And going from 12 - 19% still doesn't make it a good chance. That's some bumpf to try and make XG look good/valid/useful/something more than a mere plaything.

 

Because otherwise we could coach SL to put his 85% XG 'bound to score' ratings from 3 feet away and the drought of however many games it is now would be gone.

[whelk]

Just seen on MOTD that Huddersfield had xG of 0.08 against Burnley.

[Suomi Saint]

The league table doesn't lie......