German Football 22/23 | 2. Bundesliga returns | Hamburg vs Schalke 20:30 |

[hasanejaz88]

Damn, that Gravenberch injury looks bad. Just 24 hours after letting Sabitzer go.

Edit: Completed the rest of the match, looked better after a few minutes.

 

[kaiser1]

Three at the back might be a viable tactics vs PSG because we have 2 full backs who are not very comfortable defending and vs Mbappe Neymar and Messi you need every body.
In the Bundesliga we don't need that type of tactic though its what Naggelsman played at Leipzig

------Pavard---Upa--Deligt
Cancelo--Kimmich--Goretzka--Davies
----Coman--Chupo--Musiala

 

[Zehner]

The problem with using xG to judge individual matches is that it doesn't tell you anything about context or why the stat looks like it does. It doesn't capture at all if someone for example fails to square the ball for an open goal, it didn't capture Tabsoba's two attempted own goals on Sunday, it doesn't capture game-state related strategy, e.g. if a team drops back, because they lead or if a team gets a bunch of counter attacks in late in the game, because they happened to be in the lead.
The more games you take into consideration the less these factors matter, because they usually balance each other out over time.
But reading one value and applying it indisrimminately doesn't make an analysis and there's a reason why the creators of the model are looking for ever more advanced metrics, such as xT for example, to account for the aforementioned shortcomings.

How many games have you seen in which players fecked up so many square passes to change a xG difference of around 1.0 to their favor? You probably would have to fail in 3 seperate occasions at least. This is ultra specific and almost never the case. Plus the further you go down the "chain of plays" before a finish to an attack, the more you diminish the model. I mean, where do you draw the line? What if a player fecks up the pass that sets a player up so that he only needs to square for an open goal? What if a player fecks up the pass that sets a player up so that he only needs to play one through ball so that the receiver only needs to square it for an open goal? Leverkusen against Cologne for instance last year, the game in which we gave a 2:0 lead out of hand. We had countless counter opportunities but kept fecking up the last pass that would set up 1 on 1s. We only had ourselves to blame for that - we would not deserve a higher goal probability than what we got in the end.

We're talking optimization here. The bad side of pareto so to say. And if you worked with data models such as this, you know that the magic lies in its simplicity and the selection of the training data and considered attributes rather than projecting reality perfectly. Things like xT definitely sound interesting and I'm looking forward to implementations of it but in its current form, xG already gives you a great idea which team had the better chance quantity * chance quality.

 

[do.ob]

How many games have you seen in which players fecked up so many square passes to change a xG difference of around 1.0 to their favor? You probably would have to fail in 3 seperate occasions at least. This is ultra specific and almost never the case. Plus the further you go down the "chain of plays" before a finish to an attack, the more you diminish the model. I mean, where do you draw the line? What if a player fecks up the pass that sets a player up so that he only needs to square for an open goal? What if a player fecks up the pass that sets a player up so that he only needs to play one through ball so that the receiver only needs to square it for an open goal? Leverkusen against Cologne for instance last year, the game in which we gave a 2:0 lead out of hand. We had countless counter opportunities but kept fecking up the last pass that would set up 1 on 1s. We only had ourselves to blame for that - we would not deserve a higher goal probability than what we got in the end.

We're talking optimization here. The bad side of pareto so to say. And if you worked with data models such as this, you know that the magic lies in its simplicity and the selection of the training data and considered attributes rather than projecting reality perfectly. Things like xT definitely sound interesting and I'm looking forward to implementations of it but in its current form, xG already gives you a great idea which team had the better chance quantity * chance quality.

Expected goals is a function that assigns expected value to a shots. It doesn't try to say anything about how or why, about which side was better, which side was in control or which side was more dangerous in possession. And since football is full of little events that shape the outcome of one particular attack and consequently the game, it greatly diminishes the value of xG for individual matches, beyond it's core function: stating "these shots had that value". And it's not just failing to square a ball, that was just one example. Own goals, premature refree calls, last ditch tackles, misplaced passes (and probably many more) all of them can turn dangerous attacks into events that aren't even recognized by xG. As is strategic decisions, teams minimizing or maximizing their risk taking in order to chase goals or keep a lead. That's why you shouldn't make grand statements based on single game xG, because it's only when you take a large enough sample size that this variance usually evens itself out.

And even then you can't blindly use xG or xPts as a predictive measure. First of all, because teams already have banked their "excess" points, no one can take them away, they aren't forced to underperform to account for their overperformance and secondly, because the assumption that the past predicts the future ignores the possibility of (significant) development. For example someone, who had a leaky defence may have their star defender return, someone, who couldn't create might have their midfield maestro return and suddenly the past is largely irrelevant.

 

[Zehner]

Expected goals is a function that assigns expected value to a shots. It doesn't try to say anything about how or why, about which side was better, which side was in control or which side was more dangerous in possession. And since football is full of little events that shape the outcome of one particular attack and consequently the game, it greatly diminishes the value of xG for individual matches, beyond it's core function: stating "these shots had that value". And it's not just failing to square a ball, that was just one example. Own goals, premature refree calls, last ditch tackles, misplaced passes (and probably many more) all of them can turn dangerous attacks into events that aren't even recognized by xG. As is strategic decisions, teams minimizing or maximizing their risk taking in order to chase goals or keep a lead. That's why you shouldn't make grand statements based on single game xG, because it's only when you take a large enough sample size that this variance usually evens itself out.

And even then you can't blindly use xG or xPts as a predictive measure. First of all, because teams already have banked their "excess" points, no one can take them away, they aren't forced to underperform to account for their overperformance and secondly, because the assumption that the past predicts the future ignores the possibility of (significant) development. For example someone, who had a leaky defence may have their star defender return, someone, who couldn't create might have their midfield maestro return and suddenly the past is largely irrelevant.

I think in English you say that you're looking for the fly in the ointment. Everything you say is correct but it is insignificant overall. If you simply and strictly judge by xG values without even taking a single glance at the game, you probably already judge 80-90% of games correctly, probably even more.

In the end, most things you describe correlate with xG anyway. If you enforce lots of "own goal chances", you usually need to be in the box a lot for that meaning you'll have your finishes and collect xG. If you have a low xG, it means you force few of those situations. When you have to do lots of high risk tackles, sooner or later one will go wrong and lead to a high xG chance. And so forth.

Where xG actually has it's weaknesses is if a team plays with a system that creates situations that aren't captured by xG overaveragely often. Like for instance Favre who wants his team to defend passively (= many shots for the opponent from relatively close ranges) and also play for high percentage chances instead of taking shots too early (= fewer attacks leading to actual shots). Or Klopp who optimized a typically low probability attacking pattern (crosses) at Liverpool. Or if you have a player such as Valverde or Messi who are ridiculously good at taking shots from typically low percentage areas, meaning that their xG from those should be much higher than the average. But those are methodical "exploitations" of the model, meaning that they don't just mess with the xG values of a single game but with the entirety across the whole season as long as they stick to systemizing attacking patterns which aren't captured correctly by xG.

On top of it, it is debatable whether some of the things you mentioned should even be considered in terms of "goal probability". A last ditch tackle means the attack was defended in the end. If all of your attacks were defended that way, you didn't create chances. If you misplace passes, it's on your team that it didn't lead to a clear goal scoring chance. Of course it is interesting to capture those things for coaching teams in order to derive at which areas the team can improve but when you want to judge the performance of the team, they aren't necessarily positive actions. And when it is not even clear whether they should be valued positively or negatively, they shouldn't be included in the computation. As said, the magic lies in simplicity. I believe you generally make the mistake many make when they work with large data sets and predictive models for the first time. You have to keep it simple and focus on some key attributes - if you include everything that could affect the outcome then chances are that you won't find enough relevant data to train the model for all the different combinations of attributes. Entropy and all that.

 

[do.ob]

I think in English you say that you're looking for the fly in the ointment. Everything you say is correct but it is insignificant overall. If you simply and strictly judge by xG values without even taking a single glance at the game, you probably already judge 80-90% of games correctly, probably even more.

In the end, most things you describe correlate with xG anyway. If you enforce lots of "own goal chances", you usually need to be in the box a lot for that meaning you'll have your finishes and collect xG. If you have a low xG, it means you force few of those situations. When you have to do lots of high risk tackles, sooner or later one will go wrong and lead to a high xG chance. And so forth.

Where xG actually has it's weaknesses is if a team plays with a system that creates situations that aren't captured by xG overaveragely often. Like for instance Favre who wants his team to defend passively (= many shots for the opponent from relatively close ranges) and also play for high percentage chances instead of taking shots too early (= fewer attacks leading to actual shots). Or Klopp who optimized a typically low probability attacking pattern (crosses) at Liverpool. Or if you have a player such as Valverde or Messi who are ridiculously good at taking shots from typically low percentage areas, meaning that their xG from those should be much higher than the average. But those are methodical "exploitations" of the model, meaning that they don't just mess with the xG values of a single game but with the entirety across the whole season as long as they stick to systemizing attacking patterns which aren't captured correctly by xG.

On top of it, it is debatable whether some of the things you mentioned should even be considered in terms of "goal probability". A last ditch tackle means the attack was defended in the end. If all of your attacks were defended that way, you didn't create chances. If you misplace passes, it's on your team that it didn't lead to a clear goal scoring chance. Of course it is interesting to capture those things for coaching teams in order to derive at which areas the team can improve but when you want to judge the performance of the team, they aren't necessarily positive actions. And when it is not even clear whether they should be valued positively or negatively, they shouldn't be included in the computation. As said, the magic lies in simplicity. I believe you generally make the mistake many make when they work with large data sets and predictive models for the first time. You have to keep it simple and focus on some key attributes - if you include everything that could affect the outcome then chances are that you won't find enough relevant data to train the model for all the different combinations of attributes. Entropy and all that.

Well I tend to disagree, what I described can be seen in most games to some degree. Case in point today's match:

This is infogol's live display for the second half:

It shows a combined xG of 0.09 for Leverkusen, three pitiful long shots. Looking at xG I would have to say that Augsburg's defense was inpenetrable, I'd be tempted to ask whether Leverkusen even had any attackers on the pitch. Yet watching the game I saw a bunch of somewhat promising attacks. Nothing too impressive, but clearly a lot more than is shown here. If Augsburg were to produce a tap in from a counter attack, because Leverkusen threw everyone forward, I'd even be tempted to see total domination from their part.

And "sooner or later" or "most things correlate with xG anyway" are factors that come into play with larger sample sizes than one single game or even less.

 

[ForEverEleven]

Who couldve known that sitting deep against every opponent and hoping Diaby and Frimpong are fast enough to constantly run after long balls on the counter isnt a tactic that works over a longer period of time??? Alonso must be flabbergasted.

 

[stefan92]

Who couldve known that sitting deep against every opponent and hoping Diaby and Frimpong are fast enough to constantly run after long balls on the counter isnt a tactic that works over a longer period of time??? Alonso must be flabbergasted.

Yes, Alonso made Bayer stable again but now he has reached the point where he has to prove that he is above the level of the typical firefighter and actually is a capable manager.

 

[do.ob]

Who couldve known that sitting deep against every opponent and hoping Diaby and Frimpong are fast enough to constantly run after long balls on the counter isnt a tactic that works over a longer period of time??? Alonso must be flabbergasted.

It's not like the squad building has made it easy for him, is it? I won't give him lots of praise, but I don't think it's fair to blame him too much either. His one fully fit "quality" attacker is a pace merchant, who is lost in tight(er) spaces. Much of this is forced by Leverkusen's transfer strategy.

 

[Piratesoup]

Koan Neuer mehr?

 

[ForEverEleven]

It's not like the squad building has made it easy for him, is it? I won't give him lots of praise, but I don't think it's fair to blame him too much either. His one fully fit "quality" attacker is a pace merchant, who is lost in tight(er) spaces. Much of this is forced by Leverkusen's transfer strategy.

Apart from the fact that I dont think thats a fair assessment of Diabys qualities, who is underrated in tight spaces, youre right partially. Midfield doesnt have any creativity and apart from Diaby our attacking players lack consistency, due to their age. But lets not pretend this team isnt good enough to convincingly beat Augsburg. The complete absence of any attacking patterns or structure in possession is the managers responsibility.

 

[stefan92]

Koan Neuer mehr?

At least FC Hollywood is back.

 

[do.ob]

Apart from the fact that I dont think thats a fair assessment of Diabys qualities, who is underrated in tight spaces, youre right partially. Midfield doesnt have any creativity and apart from Diaby our attacking players lack consistency, due to their age. But lets not pretend this team isnt good enough to convincingly beat Augsburg. The complete absence of any attacking patterns or structure in possession is the managers responsibility.

This will probably sound much more offensive than I mean to, but your post reminds me of a certain meme:

Step 1 shitty attackers
Step 2 central midfield devoid of creativity
???
better beat Augsburg convincingly!

When I hear Leverkusen vs Augsburg, sure, my mind expects Leverkusen to win, but when you actually look at that lineup, I'd ask: why should they? Frimpong and Diaby get neutralized by a lack of space. Are Hlozek and Adli supposed to counterbalance that in other areas? Are Palacios and Andrich supposed to compensate for impotent and inconsistent attackers? It's not exactly like dropping points was inevitable, but a coach can only work with what the club gave him and what he has right now is pretty mediocre and mediocre teams sometimes lose to opponents like Augsburg.

 

[ForEverEleven]

This will probably sound much more offensive than I mean to, but your post reminds me of a certain meme:

Step 1 shitty attackers
Step 2 central midfield devoid of creativity
???
better beat Augsburg convincingly!

When I hear Leverkusen vs Augsburg, sure, my mind expects Leverkusen to win, but when you actually look at that lineup, I'd ask: why should they? Frimpong and Diaby get neutralized by a lack of space. Are Hlozek and Adli supposed to counterbalance that in other areas? Are Palacios and Andrich supposed to compensate for impotent and inconsistent attackers? It's not exactly like dropping points was inevitable, but a coach can only work with what the club gave him and what he has right now is pretty mediocre and mediocre teams sometimes lose to opponents like Augsburg.

How did this mediocre team finish 3rd last season then? We havent lost any key players or even starters in the summer. Obviously Schicks long-term injury and the fact Wirtz is only back since 3 games and understandably not at 100% yet doesnt help but even if you take those two out of it, we have a better team than Augsburg, in every regard. I dont understand how this is even a point of discussion.

 

[stefan92]

How did this mediocre team finish 3rd last season then? We havent lost any key players or even starters in the summer. Obviously Schicks long-term injury and the fact Wirtz is only back since 3 games and understandably not at 100% yet doesnt help but even if you take those two out of it, we have a better team than Augsburg, in every regard. I dont understand how this is even a point of discussion.

You sounded like Bayer's team is pretty shit. It obviously isn't, but that's how it appeared, so I get @do.ob's response and wouldn’t take it really serious.

 

[do.ob]

How did this mediocre team finish 3rd last season then? We havent lost any key players or even starters in the summer. Obviously Schicks long-term injury and the fact Wirtz is only back since 3 games and understandably not at 100% yet doesnt help but even if you take those two out of it, we have a better team than Augsburg, in every regard. I dont understand how this is even a point of discussion.

a) It wasn't the same team, they had Schick and Wirtz (for the most part) fit and in form, yesterday those were replaced by Adli and Hlozek, who have 2 goals 6 assists between them and don't even come close to filling the same tactical role.
b) I think they were a bit fortunate that season.
c) I'm not saying Augsburg has a better squad, but it was a lot closer yesterday than one would expect when looking at the club names and certainly not so far apart that you can expect an automatic win.

 

[GhastlyHun]

Koan Neuer mehr?

The bridge is certainly burning on both ends

 

[Blackwidow]

The bridge is certainly burning on both ends

And not just about that topic. Like usually - might be a symptom but not the cause...

 

[do.ob]

The bridge is certainly burning on both ends

The fact that his interview was also released in the Athletic certainly seems omnious.

 

[stefan92]

Freiburg in Dortmund, always tragic. Harsh but understandable to send Sildililla off, but when it happens this quickly (second yellow 73 seconds after his first) and this early in a match, it always feels too hard.

 

[do.ob]

I can't wait for Streich's post match interview.

 

[stefan92]

VAR madness in Frankfurt... never before heard "there will be a penalty, we just don't know yet for which team"

But everything correct as I see - Hasebe unintentionally got the ball to his hand (potential penalty for Hertha) which started Frankfurt's counter attack, which was stopped by a foul and resulted in a penalty for Frankfurt (goal Kolo Muani).

 

[do.ob]

Death.. Taxes.. and a Bundesliga purchase of Dortmund scoring against his former club.

 

[stefan92]

Is it just me, or does it feel like an insult to any talented professional player that Kevin f***ing Behrens is successfully playing in the Bundesliga? (Nothing against him, but he is the typical player who spends his career somewhere between 2nd/3rd tier, not fighting for winning titles )

 

[do.ob]

Is it just me, or does it feel like an insult to any talented professional player that Kevin f***ing Behrens is successfully playing in the Bundesliga? (Nothing against him, but he is the typical player who spends his career somewhere between 2nd/3rd tier, not fighting for winning titles )

I think it mainly proves how much bigger than the individual the collective is at Union.

 

[kaiser1]

If you missed the B&B partnership of Bramble and Boumsong then you shouldn't miss the N&N of Niklas and Nico. S&S Sule and Schloterbeck in Dortmund

 

[Blackwidow]

Freiburg in Dortmund, always tragic. Harsh but understandable to send Sildililla off, but when it happens this quickly (second yellow 73 seconds after his first) and this early in a match, it always feels too hard.

No. It is not understandable as it was not hard or the action anywhere decisive. Why don't the opponents of Dortmund get the same empathy by the referees that Dortmund gets?

 

[do.ob]

Not that it matters at this point, but how did Ginter escape without a card after wrestling down an opponent, who was about to pass into the box.

 

[do.ob]

Sounded like Streich suggested to the refs that they should put on yellow shirts..

 

[do.ob]

Tweet

— Twitter API (@user) date

Had a busy night, @stefan92?

 

[stefan92]

Tweet

— Twitter API (@user) date

Had a busy night, @stefan92?

Yes but not in Braunschweig

 

[do.ob]

Bayern's squad is getting a bit thin at the moment, isn't it?

 

[GhastlyHun]

Bayern's squad is getting a bit thin at the moment, isn't it?

Paperthin.

 

[GhastlyHun]

Lads, it's Wolfsburg

Another beauty from Cancelo and equally from Coman

 

[do.ob]

Losing to Bayern in 2023... how embarassing...

 

[GhastlyHun]

Our current shitness has been canceloed!

 

[Acrobat7]

Bayern's squad is getting a bit thin at the moment, isn't it?

Neuer, Mazraoui, Mané, Hernandez… But Upamecano, Choupo, Gravenberch should be back soon.

 

[GhastlyHun]

Wolfsburg have been much better after switching their backline. We have been a bit lucky they have not equalized by now.

Kimmich sent off... this will be a looooong 40 minutes

 

[strongwalker]

Not so half bad from Jamal

 

[Blackwidow]

MUSIALA!!!!