Statistical Domination 2023/24

[hogesar 9,732]

Welcome to Statistical Domination 2023/24 Season! Some things:

1. You can view last seasons thread here (I guess the related picture is quite relevant, in the end)
2. We don't have fancy graphs this season. Neither experimental361 or FBRef or the Twitter accounts seem to be regularly producing them. So for this season we're going to be using https://www.sofascore.com/ for the stats for consistency throughout the season.
3. Lot's of you don't like stats and in particular xG. That is fine, footballs for everyone after all. So is this forum - so if you're not into it don't feel obligated to post how you're not into it.

 

Matches:

 

 

 

Discussion points so far:

1. Our total xG for (note we don't have the figures for League Cup fixtures) is 10.52 xGF (for) vs  6.2 xGA (against) after 5 league games.
2. Southampton is a clear outlier. It looks bad statistically but two debatable penalties equate to 1.52xG and our offside goal won't have counted towards xG at all. 
3. After 5 league games, 2 home 3 away (typically the away side has a lower xG in the championship, at a very basic rough level), we average an xGF of 2.1 per game vs xGA of 1.24 per game. For those interested in comparing to last season, the first 5 games we averaged an xGF of 1.32 per game vs xGA of 0.81 per game. 
4. From above I think it would surprise several to see that we were better defensively first 5 games of last season, but poorer going forwards.
5. Rotherham wasn't a great performance from us but no where near as bad as some made out - their first goal was a great strike but the chances of them scoring another like that this season is incredibly slim. We created more, but lots of half-chances in reality.

 

League Comparison:

 

Worth noting we've  created more big chances  than anyone else in the league so far. We are also the second best team behind Huddersfield for accurate crosses. 

Conversely, we are not in the top 10 for ball possession or accurate passes which shows the swing from Farke era. To reinforce that further, we are one of the better teams (7th) for interceptions, successful dribbles (10th), clearances (6th).

And like with what seems like every season we've had recently in the Premier League, we already top the table for amount of times we've hit the woodwork.

 

 

Edited September 4, 2023 by hogesar

[Robert N. LiM 4,476]

about f***ing time Hoggy.

Let battle commence

[Robert N. LiM 4,476]

More seriously, thanks for posting these, @hogesar. I think they provide a really good basis for some interesting discussion of how we're playing.

It's a great shame that the graphs aren't available this season. I thought they were really helpful - it's the biggest advantage in my view, of xG - that it attempts to quantify the quality of chances created in a game.

I was very struck at HT in the Rotherham game that everyone who'd been watching was slating the performance, but that the BBC stats (which are very, very basic) had us dominating possession and having had way more shots. I asked about it on the match thread and the response was unanimous that those stats flattered us, which I'm completely prepared to believe. Shots, and even shots on target, are a pretty useless measure when analysing a performance.

I assume what an xG graph would have shown was that the shots we had were low-quality chances (from the highlights I don't remember us missing a really good chance in the first half), and that Rotherham scored from two quite low-quality chances themselves. In other words, poorly as we clearly played in the first half, that sort of game of football might well have been 0-0 at HT. Obviously the first goal was a superb strike, and as you say the guy's unlikely to repeat it this season, but Hugoal's header was outstanding, too. Wasn't an easy chance, but he took it really well.

Edited September 4, 2023 by Robert N. LiM
typo

[Badger 2,412]

Thanks - good stuff. The Rotherham figures suggest we were not as bad as suggested.

[ZLF 274]

I think the other data item worth comparing, if available, is the no of big chances conceded, which i expect to see an even bigger improvement than seen in xGa.   The structure and press is significantly better for us than either managerial team last year.

Rotherham shows we need to be energised and focused for a full 90 mins, a half isn't enough but did show that we will be a threat with the right mentality.   And that it wasn't all idahs fault.

The data reflects what we have seen on pitch so far.  Hopefully we will see a repeat over the next 5 games. 

 

Edited September 4, 2023 by ZLF

[hogesar 9,732]

2 hours ago, ZLF said:

I think the other data item worth comparing, if available, is the no of big chances conceded, which i expect to see an even bigger improvement than seen in xGa.   The structure and press is significantly better for us than either managerial team last year.

Rotherham shows we need to be energised and focused for a full 90 mins, a half isn't enough but did show that we will be a threat with the right mentality.   And that it wasn't all idahs fault.

The data reflects what we have seen on pitch so far.  Hopefully we will see a repeat over the next 5 games. 

 

Yes, I'm looking for a reliable big chances conceded - Sofascore didn't offer that to me.

Rotherham can quickly be put down to "one of those games" - so long as our reaction and follow up performances match the earlier games.

It wasn't terrible from chance created / conceded point of view but our passing and ball retention with Sara and McLean was probably as poor as its ever been. The fact we still created more than them is probably testament to the system itself above the players performances that day.

[hogesar 9,732]

6 hours ago, Robert N. LiM said:

More seriously, thanks for posting these, @hogesar. I think they provide a really good basis for some interesting discussion of how we're playing.

It's a great shame that the graphs aren't available this season. I thought they were really helpful - it's the biggest advantage in my view, of xG - that it attempts to quantify the quality of chances created in a game.

I was very struck at HT in the Rotherham game that everyone who'd been watching was slating the performance, but that the BBC stats (which are very, very basic) had us dominating possession and having had way more shots. I asked about it on the match thread and the response was unanimous that those stats flattered us, which I'm completely prepared to believe. Shots, and even shots on target, are a pretty useless measure when analysing a performance.

I assume what an xG graph would have shown was that the shots we had were low-quality chances (from the highlights I don't remember us missing a really good chance in the first half), and that Rotherham scored from two quite low-quality chances themselves. In other words, poorly as we clearly played in the first half, that sort of game of football might well have been 0-0 at HT. Obviously the first goal was a superb strike, and as you say the guy's unlikely to repeat it this season, but Hugoal's header was outstanding, too. Wasn't an easy chance, but he took it really well.

Yes I think primarily it was chances for the likes of Sara from just outside the area. In 9 out of 10 games, you fancy him to score one of those over their RB smashing in a half-volley top-bins from outside the area. But we all know football doesn't quite work like that!

[Jimmy Raggatip 61]

[Captain Holt 522]

18 hours ago, hogesar said:

Welcome to Statistical Domination 2023/24 Season! Some things:

1. You can view last seasons thread here (I guess the related picture is quite relevant, in the end)
2. We don't have fancy graphs this season. Neither experimental361 or FBRef or the Twitter accounts seem to be regularly producing them. So for this season we're going to be using https://www.sofascore.com/ for the stats for consistency throughout the season.
3. Lot's of you don't like stats and in particular xG. That is fine, footballs for everyone after all. So is this forum - so if you're not into it don't feel obligated to post how you're not into it.

 

Matches:

 

 

 

Discussion points so far:

1. Our total xG for (note we don't have the figures for League Cup fixtures) is 10.52 xGF (for) vs  6.2 xGA (against) after 5 league games.
2. Southampton is a clear outlier. It looks bad statistically but two debatable penalties equate to 1.52xG and our offside goal won't have counted towards xG at all. 
3. After 5 league games, 2 home 3 away (typically the away side has a lower xG in the championship, at a very basic rough level), we average an xGF of 2.1 per game vs xGA of 1.24 per game. For those interested in comparing to last season, the first 5 games we averaged an xGF of 1.32 per game vs xGA of 0.81 per game. 
4. From above I think it would surprise several to see that we were better defensively first 5 games of last season, but poorer going forwards.
5. Rotherham wasn't a great performance from us but no where near as bad as some made out - their first goal was a great strike but the chances of them scoring another like that this season is incredibly slim. We created more, but lots of half-chances in reality.

 

League Comparison:

 

Worth noting we've  created more big chances  than anyone else in the league so far. We are also the second best team behind Huddersfield for accurate crosses. 

Conversely, we are not in the top 10 for ball possession or accurate passes which shows the swing from Farke era. To reinforce that further, we are one of the better teams (7th) for interceptions, successful dribbles (10th), clearances (6th).

And like with what seems like every season we've had recently in the Premier League, we already top the table for amount of times we've hit the woodwork.

 

 

Superb.

[Flying Dutchman 744]

Good stuff Hogesar, thanks for this.

One question, whilst I understand not every goal has an assist, I don't understand how the expected assists (xA) can occasionally be higher than the xG, Hull at home as an example.

Surely that isn't possible?

[king canary 7,664]

18 hours ago, hogesar said:

From above I think it would surprise several to see that we were better defensively first 5 games of last season, but poorer going forwards.

Interested in why you think this would be surprise people? I think the criticism last season early doors was how we struggled to create chances and gave away stupid goals through individual errors rather than people opening us up.

[Yelloow Since 72 54]

Excellent again and it's particularly useful to compare all the games played rather than a single game on its own. Brilliant

[hogesar 9,732]

50 minutes ago, Flying Dutchman said:

Good stuff Hogesar, thanks for this.

One question, whilst I understand not every goal has an assist, I don't understand how the expected assists (xA) can occasionally be higher than the xG, Hull at home as an example.

Surely that isn't possible?

It's because (I think, anyway) expected assists xA measures the likelihood that a given pass becomes a goal assists. It's based on the pass, and doesn't therefore require the player receiving the pass to take a shot - which is unlike xG, which of course requires the shot to happen.

So we've seen it pretty often in football where a player plays a dangerous, threaded pass to put the striker in on goal but he takes an extra touch and gets tackled last second. In that scenario, the xG would be 0 because he never shot, but the xA would get a value because the pass was deemed good enough to create a goal-scoring opportunity. 

[hogesar 9,732]

40 minutes ago, king canary said:

Interested in why you think this would be surprise people? I think the criticism last season early doors was how we struggled to create chances and gave away stupid goals through individual errors rather than people opening us up.

I think you're right re individual errors but a defender losing the ball on the edge of his own area and a player having a clear run and shot on goal would be a relatively high xG. I think people have been happy with Duffy (less so Gibson, although he has made less mistakes than Duffy so far).

Maybe I remember it wrong but very early on in the season (perhaps before the winning run) the general thesis seemed to be we can't attack well and we can't defend well, and certainly early home games felt nervous around the crowd.

[hogesar 9,732]

@Bethnal Yellow and Green shared this on twitter, which shows a rolling non-penalty xGF and xGA from Smith onwards:

The graph does show a very positive start this season but equally, we saw that at the start of last season without being convinced - styles of course playing a big part in fan emotion / involvement etc.

[hogesar 9,732]

4 minutes ago, hogesar said:

It's because (I think, anyway) expected assists xA measures the likelihood that a given pass becomes a goal assists. It's based on the pass, and doesn't therefore require the player receiving the pass to take a shot - which is unlike xG, which of course requires the shot to happen.

So we've seen it pretty often in football where a player plays a dangerous, threaded pass to put the striker in on goal but he takes an extra touch and gets tackled last second. In that scenario, the xG would be 0 because he never shot, but the xA would get a value because the pass was deemed good enough to create a goal-scoring opportunity. 

To reinforce how xA is used by football clubs, this little snapshot from "thanalyst.com" says:

Quote

Norwich City’s sporting director, Stuart Webber, discussed how they used xA in the recruitment of arguably their best player over the last few seasons, Emiliano Buendia. While his assist numbers weren’t substantial when he was playing in the Spanish second division, their analysts used expected assists as one of the metrics to recognise the underlying creativity of the Argentine.

 

[king canary 7,664]

1 hour ago, hogesar said:

I think you're right re individual errors but a defender losing the ball on the edge of his own area and a player having a clear run and shot on goal would be a relatively high xG. I think people have been happy with Duffy (less so Gibson, although he has made less mistakes than Duffy so far).

Maybe I remember it wrong but very early on in the season (perhaps before the winning run) the general thesis seemed to be we can't attack well and we can't defend well, and certainly early home games felt nervous around the crowd.

My memory was the 'we can't defend' stuff was based around the high amount of errors we were committing. We may not have been giving up high xG but we were top of the charts for 'errors leading to goals' for most of the season.

[Google Bot 3,299]

1 hour ago, king canary said:

My memory was the 'we can't defend' stuff was based around the high amount of errors we were committing. We may not have been giving up high xG but we were top of the charts for 'errors leading to goals' for most of the season.

It's just a take from a set of data, you can approach and argue in any number of ways. 

xG is really designed to work out if you're taking the chances available on average based on relative statistics.  Whereas figures of losing possession in key areas that lead to goals is much more of an absolute count. 

I'm sure the two would correlate somewhere along the line, but individual mistakes and their frequency can't really be portrayed in xG figures too well as it's all about the position that the attacking/defending team find themselves in after such events have occured.

Maybe there's some kind of equivalent possession based stat to xG?  Would be a nightmare to gather that though, as it would involve where passes were made, who the target was and what kind of space they had.

[Badger 2,412]

BYG posted this on Twitter. For those of us, who place some value on football statistics, I think it makes interesting reading and shows how well we have started both offensively and defensively, although still not quite as tight as early season Smith.

 

 

 

[hogesar 9,732]

Cheers @Badger - I'd posted that above!

Slightly different, here's a graph interpretation of playing style:

[Canaries north 151]

I do look at stats and do believe they are a good guide to state of play. I also think in game management has a big part to play. At 2 0 up against a team you are expected to lose to Rotherham stats are always going to be misleading with one team sitting back.

[hogesar 9,732]

On 07/09/2023 at 12:06, Canaries north said:

I do look at stats and do believe they are a good guide to state of play. I also think in game management has a big part to play. At 2 0 up against a team you are expected to lose to Rotherham stats are always going to be misleading with one team sitting back.

Yeah, 100%

There were games towards the end of Smiths' reign that didn't look bad on xG figures alone, but that was because we'd gone 1-0 down early and then had 10-15 shots which were all low xG chances but built up to make it look half decent.

[Bristol Canary 17]

No FBref or Project FiveThirtyEight stuff yet but Opta are still doing some interesting things here (As you've probably seen Hogesar, but for others that might not have):
https://theanalyst.com/eu/2023/08/championship-stats-2023-24-opta/

We're currently 2nd for xG for in the league in open play and 3rd for xG against which suggests we've been difficult to create changes against and good at creating our own so far. Interestingly they also have us top for creating turnovers that lead to shots which suggests the press has been working well too to help create those chances. 

what was the quote "No playmaker in the world can be as good as counter pressing... except Gabby Sara"

[repman 635]

 

Managed to put together the xG game graphs for all our league fixtures so far, if anyone has any ideas on what could be added let me know!

The data is from fbref.com who I think get their data from opta now.

Just want to add that although the graphs are a good visualisation they aren't perfect explainers either. Take the Rotherham game for example, yes we created much more xG than they did but the majority of it was after we were already down at least a goal. Hull on the opening day was similar, if you look it's clear that most of our chances that day came as we ramped up the pressure before Rowe's equaliser.

Also the Huddersfield graph is maybe the best reason as to why xG is a better metric than simple shot totals, Huddersfield outshot us but the quality of their shots to ours was night and day.

[hogesar 9,732]

Great work @repman, do you mind if I add them into the original post?

[Robert N. LiM 4,476]

1 hour ago, repman said:

Also the Huddersfield graph is maybe the best reason as to why xG is a better metric than simple shot totals, Huddersfield outshot us but the quality of their shots to ours was night and day.

Am I reading that graph right? The four goals came from just five attempts on goal in that period? Lethal.

[Robert N. LiM 4,476]

1 hour ago, repman said:

Just want to add that although the graphs are a good visualisation they aren't perfect explainers either. Take the Rotherham game for example, yes we created much more xG than they did but the majority of it was after we were already down at least a goal. Hull on the opening day was similar, if you look it's clear that most of our chances that day came as we ramped up the pressure before Rowe's equaliser.

Also the Huddersfield graph is maybe the best reason as to why xG is a better metric than simple shot totals, Huddersfield outshot us but the quality of their shots to ours was night and day

Just wanted to say that this is a great analysis of some of the pluses and minuses of xG for anyone struggling to understand it or suspicious of its value. Like all stats, xG stats need to be analysed and put into context if you're going to get full value out of them. But yes, xG is so much more useful than the simple shots/shots on target stats you get on the BBC site, for instance.

[repman 635]

1 hour ago, hogesar said:

Great work @repman, do you mind if I add them into the original post?

Go for it!

1 hour ago, Robert N. LiM said:

Am I reading that graph right? The four goals came from just five attempts on goal in that period? Lethal.

Yeah, it reminds me in some ways of Wagner's first few games in charge, where in total I think we scored 6 goals early on from a total of maybe 7 shots. It's good that you're taking your chances and especially when they're high quality like vs Huddersfield, but it may also be masking some deeper issues too. That being said I think the rest of our games suggest that we can create quality chances at a decent rate too.

[hogesar 9,732]

35 minutes ago, First Wazzock said:

Totally agree.

If we had been playing a team with their shooting boots on we could have lost that quite Easily. even though we looked really good to start off with there was no finish, and I have got to say, overall I felt it was a game low on quality. Our defence still worries me, at times it looked like it was going to part like the Red Sea (Exodus 14: 19-31 for our younger members!)  

But as everyone keeps saying, a win is a win

 

56 minutes ago, Indy said:

Yes we were lucky, they had two clear chances to equalise and should have both times their players missed the ball! It should have been 1-1 I’m not saying on the first half performance we didn’t deserved the win we certainly did deserve the win, but we were lucky both chances they wafted on the shot!

Sorry to the two I've used as examples above because there's plenty more but there's been a lot said about how lucky we were. I didn't think we were particularly lucky, we just defended well.

Whilst it was close overall we missed mode "big chances" than they did. Far from a top performance but not the lucky fluke style of win either!

[Indy 3,314]

Hogy I saw two clear chances both air shots by their players, one very questionable pull on the players shirt to put him off but both clear chances! No one is saying we didn’t deserve the win, and no one is saying lucky fluke! Where is that in anyone’s points? But it could have easily been 1-1…..I don’t care how much you try and say weren’t lucky I’ll take that luck with the three points!