xG

[Thomas Bathurst]

I'd love to know what peoples theories are as to why we consistently always underperform xG.

I don't believe there has ever been a time in recent years when we have actually been above where xG suggests we should. I find that really odd. Even when doing really well in the Championship, xG suggested we should be higher still. We often joke around us at games when we are behind, that it's alright as we are in front with regard xG. You'd have thought though through the law of averages we'd have at least at some point been above our xG position. This is even more strange considering the quality of the forwards we've had in recent years. According to xG we should win pretty much 90% of all games we play in and that has continued into the Premier League.

Is it down to the "type" of chances we are choosing to create that result in a high xG chance, but in actual reality the chance is much more difficult than it may otherwise seem. I'm thinking perhaps about the back post overload (from open play or the similar set piece) that results in a header back into the 6 yard box and a chance from a few yards out. The resulting scramble etc and the reactions needed to put the ball in the back of the net often (for us anyway) see the ball saved, scrambled away, missed etc. Perhaps lower down the league with poorer keepers, with poorer reactions etc - that more of these end up as a goal.

Conversely, we rarely seem to score the type of goals we concede. I am guessing that shots from 25 yards out that we've conceded (Tielemans, Chillwell, Beundia) or someone waltzing through our defence (Normann) show up as low xG but again facing players of this quality the further up the league is actually more likely to result in a goal than the xG suggests. I guess (but don't know?) the xG for a given type of chance is the same irrespective of who is hitting it, where Tielemans is much more likely to stick it in the top corner than Harry Bunn at Scunthorpe.

We have players that can certainly hit the ball that seem reluctant to do so Norgaard and particularly Jensen (often finds himself in a decent position to hit one, and looks for another pass).

 

[Les Beeavinu]

The question is probably best answered by Christian Norgaard and Bryan Mbeumo.

 

[hanworthbee]

Perhaps the goalies are better

 

[Thomas Bathurst]

The question is probably best answered by Christian Norgaard and Bryan Mbeumo.

Of late, maybe. But this has actually been a consistent theme for a few seasons. I think it is being exacerbated further this season by the quality of opposition.

 

[Thomas Bathurst]

Perhaps the goalies are better

Again - I'm not talking about this season. If you go back to previous seasons when we were doing well in the Championship - xG always suggests we should be higher still.

 

[Scobee]

I think it shows we are on the right path, creating lots of chances, ultimately however, we are not scoring them

 

[Thomas Bathurst]

I think it shows we are on the right path, creating lots of chances, ultimately however, we are not scoring them

Just to underline the point. If this was just a short term thing this season then I might be inclined to agree. However this situation has been the same for a number of seasons. It mattered less when winning more games in the Championship and ultimately getting ourselves promoted (even though perhaps according to xG we should have won the league).

 

[Mick.C]

I read Thomas’ post match press conferences as

xG always gives an indication if the side is performing well. Which apart from three halves out of 22 they have.
Goals scored tells you if we are clinical enough, Liverpool Arsenal games yes, Leicester, Brighton Norwich No.
Games lost indicate fine margins only Burnley defeat in 5 of them by a two goal margin.
Games continually lost indicate fine margins, lack of luck and missed opportunity.

Summary Thomas believes we are fine.
We are understandably nervous being new boys and going belly up to the dead men of the division.
We beat Arsenal Wolves and West Ham and drew with Liverpool
Lost to Norwich Burnley and Brighton.
Football being football you could see a reverse result of all those fixtures in the return games.
12 points is fine at this point.

 

[Scobee]

Just to underline the point. If this was just a short term thing this season then I might be inclined to agree. However this situation has been the same for a number of seasons. It mattered less when winning more games in the Championship and ultimately getting ourselves promoted (even though perhaps according to xG we should have won the league).

I suppose we consistently under achieve, but perhaps not in a bad way , as we have still been a success over the last few seasons

 

[Thomas Bathurst]

I read Thomas’ post match press conferences as

xG always gives an indication if the side is performing well. Which apart from three halves out of 22 they have.
Goals scored tells you if we are clinical enough, Liverpool Arsenal games yes, Leicester, Brighton Norwich No.
Games lost indicate fine margins only Burnley defeat in 5 of them by a two goal margin.
Games continually lost indicate fine margins, lack of luck and missed opportunity.

Summary Thomas believes we are fine.
We are understandably nervous being new boys and going belly up to the dead men of the division.
We beat Arsenal Wolves and West Ham and drew with Liverpool
Lost to Norwich Burnley and Brighton.
Football being football you could see a reverse result of all those fixtures in the return games.
12 points is fine at this point.

I kind of agree with all of that and hope that we will be fine, don't panic and start doing what we did yesterday in the second half. We moved Bryan to right wing back - he created one half decent chance with a cross teased in and then barely utilised him thereafter. The goal came from an opening down the right hand side (and a wonderful cross). However rather than persevere down that side, we spent 20 minutes lumping continual long balls into the left channel for Ivan & Charlie. Anyway I digress slightly ...

The point I don't necessarily agree with and is the crux of my query:

"Games continually lost indicate fine margins, lack of luck and missed opportunity."

We nearly always win from an xG perspective - this season and other seasons. I think we should have been promoted a couple of times already by the time we did. Each and every time, is this down to luck? Why don't we ever get the lucky breaks (over an extended period of time)?

 

[Mick.C]

You are correct Thomas in your summing up - it does seem we don’t get the rub and it is down to luck somedays and you can’t predict it - the key to the model they tell us is trust it.
To Thomas xG equates to is the team playing to its potential or goal potential - It certainly is.
Earlier up the thread Norgaard and Mbeumo were named as an answer - so true, players that don’t convert golden opportunities - you can’t legislate for that.
If Mbeumo could finish better, quite frankly he wouldn’t be playing for us but a top 6 side and there’s still time.
Norgaard missed two great chances yesterday through poor finishing and Krul.
We as supporters are emotionally invested in the performance AND the result so XG is irrelevant to us it’s the number that hit the onion bag that matters.
Those who plan longer look at the overall strategy say 14th is fine at the moment but we could and should be 7-10 points better off with our play. That shows the standard we are at.

However irritating it sounds after a belly up home reverse Franks assessment we will be fine has a ring of truth.

We all muttered leaving the ground last night not good enough, Goode up front, Ivan too deep, But in reality and the frustration subsiding we know where we are short, on our day with a full squad we are good enough and XG is just the reminder that the sky is not falling in.

 

[Thomas Bathurst]

You are correct Thomas in your summing up - it does seem we don’t get the rub and it is down to luck somedays and you can’t predict it - the key to the model they tell us is trust it.
To Thomas xG equates to is the team playing to its potential or goal potential - It certainly is.
Earlier up the thread Norgaard and Mbeumo were named as an answer - so true, players that don’t convert golden opportunities - you can’t legislate for that.
If Mbeumo could finish better, quite frankly he wouldn’t be playing for us but a top 6 side and there’s still time.
Norgaard missed two great chances yesterday through poor finishing and Krul.
We as supporters are emotionally invested in the performance AND the result so XG is irrelevant to us it’s the number that hit the onion bag that matters.
Those who plan longer look at the overall strategy say 14th is fine at the moment but we could and should be 7-10 points better off with our play. That shows the standard we are at.

However irritating it sounds after a belly up home reverse Franks assessment we will be fine has a ring of truth.

We all muttered leaving the ground last night not good enough, Goode up front, Ivan too deep, But in reality and the frustration subsiding we know where we are short, on our day with a full squad we are good enough and XG is just the reminder that the sky is not falling in.

Great summary as ever Mick, and I suppose a reminder to trust the process as it has got us this far ... part of me (perversely) though wants to be absolutely hammered one week from an xG perspective but sneak away with a 1.0. Just to show we do get the breaks sometimes. Maybe Man City away

 

[Hedley Lamarr]

Interesting reading at https://understat.com/league/EPL/2021

We have the most expected points to actual difference (7.66) in the PL and in the expected points table we are 4th.

We also have the most expected goals to actual difference (4.45): we have to be more clinical, the stats say Norgaard, Mbuemo and Toney have been the worst finishers.

 

[BFCDi]

Because xG isn't real and doesn't work. This whole stats delusion that has festered in this fanbase is ridiculous and it's propagated mostly by our manager who can't seem to be able to admit when we have played badly. Finishing is part of the game as much as anything else so no we didn't deserve three points against Norwich or against Leicester or Brighton or Chelsea. Less talk more scoring goals please

 

[Thomas Bathurst]

Because xG isn't real and doesn't work. This whole stats delusion that has festered in this fanbase is ridiculous and it's propagated mostly by our manager who can't seem to be able to admit when we have played badly. Finishing is part of the game as much as anything else so no we didn't deserve three points against Norwich or against Leicester or Brighton or Chelsea. Less talk more scoring goals please

Whilst I don't particularly agree with you or certainly to the extreme levels you have stated, I suppose this is an example to what I myself was pondering on some level.

Clearly there is some value in xG, but is it a bit of a blunt tool, especially the further up the leagues you go due to the disparity in players ability to defend/attack?

Are we placing too much emphasis on xG and ultimately try to create a high xG during games which don't truly convert into as "big" a chance as being suggested. Then on the other hand we perhaps fail to hit a few from further out, etc (because they don't have a high xG), but may fly in or result in the keeper parrying into the path of someone etc etc).

 

[BFC1997]

Because xG isn't real and doesn't work. This whole stats delusion that has festered in this fanbase is ridiculous and it's propagated mostly by our manager who can't seem to be able to admit when we have played badly. Finishing is part of the game as much as anything else so no we didn't deserve three points against Norwich or against Leicester or Brighton or Chelsea. Less talk more scoring goals please

xG is just a measure of the combined frequency and quality of chances created.

Why would you not be interested in maximising this for us and minimising this for the opposition?

Our finishing has to improve but that doesn’t mean you abandon those principles.

 

[Frothybaby]

Norgaard hasn't been getting in goal scoring positions in previous seasons that I remember. This season he has had a few opportunities. Maybe this is a new tactic that he is getting used to and will improve over time. Maybe why the xg is reading it differently. Bryan has hit woodwork 7 times. If those had gone in, maybe his xg would be under his goals.

 

[grepo]

I kind of agree with all of that and hope that we will be fine, don't panic and start doing what we did yesterday in the second half. We moved Bryan to right wing back - he created one half decent chance with a cross teased in and then barely utilised him thereafter. The goal came from an opening down the right hand side (and a wonderful cross). However rather than persevere down that side, we spent 20 minutes lumping continual long balls into the left channel for Ivan & Charlie. Anyway I digress slightly ...

The point I don't necessarily agree with and is the crux of my query:

"Games continually lost indicate fine margins, lack of luck and missed opportunity."

We nearly always win from an xG perspective - this season and other seasons. I think we should have been promoted a couple of times already by the time we did. Each and every time, is this down to luck? Why don't we ever get the lucky breaks (over an extended period of time)?

Didn't someone once put a curse on our luck?

 

[grepo]

Interesting reading at https://understat.com/league/EPL/2021

We have the most expected points to actual difference (7.66) in the PL and in the expected points table we are 4th.

We also have the most expected goals to actual difference (4.45): we have to be more clinical, the stats say Norgaard, Mbuemo and Toney have been the worst finishers.

Point blank reflex and lucky saves and Mbeumo hitting the woodwork 7 times. We've more than our fair share of bad luck and hopefully, this will even out over the season.

 

[grepo]

Because xG isn't real and doesn't work. This whole stats delusion that has festered in this fanbase is ridiculous and it's propagated mostly by our manager who can't seem to be able to admit when we have played badly. Finishing is part of the game as much as anything else so no we didn't deserve three points against Norwich or against Leicester or Brighton or Chelsea. Less talk more scoring goals please

....but you miss the point. The data does suggest we are playing well and better than our position would suggest and that gives optimism and suggests we are on the right path to do OK this season.

 

[AB]

Because xG isn't real and doesn't work. This whole stats delusion that has festered in this fanbase is ridiculous and it's propagated mostly by our manager who can't seem to be able to admit when we have played badly. Finishing is part of the game as much as anything else so no we didn't deserve three points against Norwich or against Leicester or Brighton or Chelsea. Less talk more scoring goals please

It’s so unreal and ineffective we’re actually hallucinating and are in our perennial 14th place in L1 worrying about having to sell the only vaguely good player we have in January.

 

[grepo]

At this point of the season would you rather be 13th with inferior xG or 14th with robust xG?

 

[Banana]

Because xG isn't real and doesn't work. This whole stats delusion that has festered in this fanbase is ridiculous and it's propagated mostly by our manager who can't seem to be able to admit when we have played badly. Finishing is part of the game as much as anything else so no we didn't deserve three points against Norwich or against Leicester or Brighton or Chelsea. Less talk more scoring goals please

?

We are all currently deluded sitting in the PL because of, er, um, (checks notes) the edge we have in the analysis of various metrics.

 

[rebus]

Because xG isn't real and doesn't work. This whole stats delusion that has festered in this fanbase is ridiculous and it's propagated mostly by our manager who can't seem to be able to admit when we have played badly. Finishing is part of the game as much as anything else so no we didn't deserve three points against Norwich or against Leicester or Brighton or Chelsea. Less talk more scoring goals please

Your recent posts have been absolute whoppers and this one is no different. Suck on a fisherman’s friend and stay away from the keyboard.

 

[AB]

Your recent posts have been absolute whoppers and this one is no different. Suck on a fisherman’s friend and stay away from the keyboard.

We should be more welcoming, particularly when former managers continue to take enough interest to post.

 

[BFCDi]

Your recent posts have been absolute whoppers and this one is no different. Suck on a fisherman’s friend and stay away from the keyboard.

I'm on my phone tbf but thanks for the advice I just get frustrated when we lose and then hearing people say how we should have won and it was all bad luck and all that makes it worse for some reason

 

[AB]

I'm on my phone tbf but thanks for the advice I just get frustrated when we lose and then hearing people say how we should have won and it was all bad luck and all that makes it worse for some reason

I mean if you feel happier with us being deserved losers knock yourself out.

 

[Braemarroad_Bee]

I'd love to know what peoples theories are as to why we consistently always underperform xG.

I don't believe there has ever been a time in recent years when we have actually been above where xG suggests we should. I find that really odd. Even when doing really well in the Championship, xG suggested we should be higher still. We often joke around us at games when we are behind, that it's alright as we are in front with regard xG. You'd have thought though through the law of averages we'd have at least at some point been above our xG position. This is even more strange considering the quality of the forwards we've had in recent years. According to xG we should win pretty much 90% of all games we play in and that has continued into the Premier League.

Is it down to the "type" of chances we are choosing to create that result in a high xG chance, but in actual reality the chance is much more difficult than it may otherwise seem. I'm thinking perhaps about the back post overload (from open play or the similar set piece) that results in a header back into the 6 yard box and a chance from a few yards out. The resulting scramble etc and the reactions needed to put the ball in the back of the net often (for us anyway) see the ball saved, scrambled away, missed etc. Perhaps lower down the league with poorer keepers, with poorer reactions etc - that more of these end up as a goal.

Conversely, we rarely seem to score the type of goals we concede. I am guessing that shots from 25 yards out that we've conceded (Tielemans, Chillwell, Beundia) or someone waltzing through our defence (Normann) show up as low xG but again facing players of this quality the further up the league is actually more likely to result in a goal than the xG suggests. I guess (but don't know?) the xG for a given type of chance is the same irrespective of who is hitting it, where Tielemans is much more likely to stick it in the top corner than Harry Bunn at Scunthorpe.

We have players that can certainly hit the ball that seem reluctant to do so Norgaard and particularly Jensen (often finds himself in a decent position to hit one, and looks for another pass).

Yes I've noticed this too over the last few seasons. We're consistently near the top of the xg standings but to back that up we did finish 3rd in consecutive seasons in the Championship.

This season we do seem to have conceded several goals from distance, villa away, Brighton home, Norwich home, chelsea home, Leicester home, Burnley away. The first three of those we didn't seem to close down their goalscorer sufficiently but I might be being too harsh. I'm guessing quite a few of the goals against were very low in the xg.

I'm a massive Bryan fan, but he has failed to hit the target a few times, seems to be aiming for the top corner a lot. But some of those efforts that hit the woodwork, Palace away, west ham away were not great chances. So on xg the quality of chances he's had haven't been that easy. Plus he's creating quite a few of those chances himself through his own intelligent play.

 

[el57]

Not putting away your early chances, and maintaining a similarly attacking approach throughout matches will make you more likely to underperform xG compared to the reverse. Why?

Look at Chelsea this season: they have allowed very few goals, so they have a great defense, sure, but it also means when they get their first goal of the game it gives them the lead. If they score some medium chance 0.15 xG goal in, say, the first half, they will shut up shop and hold the lead, and if they score again it will likely come from singular moments of exploiting mistake by an opposition that is having to push up to level the match. Compare to Brentford after we fail to put away an early goal after a blazing start: we're underperforming xG and have to keep attacking and keeping to rack up xG until the breakthrough comes. So that means:

Calmly holding onto your lead = lower xG because you aren't shooting a ton = (probably) overperform xG because you've already scored
Chasing matches to get a better result = higher xG because you have to shoot and attack more = (probably) underperform xG because you are already lacking the goals for the result you want

This extends a bit to style too as I mentioned at the start. If you calmly hold your lead like Chelsea with disciplined defending then you won't make a ton of desperate shots or heap endless pressure in attack. Brentford like to keep attacking, which teams know, so they plant to defend more and can absorb more of that pressure which -> suggests you will rack up more xG than you score.

It's difficult to know whether to see this as good or bad, to be honest. Scoring your early chances is definitely a straightforward good, so we need Mbeumo and Toney to get a bit more clinical in the first half.

EDIT: One other big factor I'd say is that every season Brentford have had to bed in new attackers since we sold the best ones form the year before. Those periods of adjustment and settling in for new players means they will probably be underperforming until things gel.

 

[Edmundo]

Not putting away your early chances, and maintaining a similarly attacking approach throughout matches will make you more likely to underperform xG compared to the reverse. Why?

Look at Chelsea this season: they have allowed very few goals, so they have a great defense, sure, but it also means when they get their first goal of the game it gives them the lead. If they score some medium chance 0.15 xG goal in, say, the first half, they will shut up shop and hold the lead, and if they score again it will likely come from singular moments of exploiting mistake by an opposition that is having to push up to level the match. Compare to Brentford after we fail to put away an early goal after a blazing start: we're underperforming xG and have to keep attacking and keeping to rack up xG until the breakthrough comes. So that means:

Calmly holding onto your lead = lower xG because you aren't shooting a ton = (probably) overperform xG because you've already scored
Chasing matches to get a better result = higher xG because you have to shoot and attack more = (probably) underperform xG because you are already lacking the goals for the result you want

This extends a bit to style too as I mentioned at the start. If you calmly hold your lead like Chelsea with disciplined defending then you won't make a ton of desperate shots or heap endless pressure in attack. Brentford like to keep attacking, which teams know, so they plant to defend more and can absorb more of that pressure which -> suggests you will rack up more xG than you score.

It's difficult to know whether to see this as good or bad, to be honest. Scoring your early chances is definitely a straightforward good, so we need Mbeumo and Toney to get a bit more clinical in the first half.

EDIT: One other big factor I'd say is that every season Brentford have had to bed in new attackers since we sold the best ones form the year before. Those periods of adjustment and settling in for new players means they will probably be underperforming until things gel.

Whilst I get what you're trying to say here, I think it boils down to you thinking that the xG metric is wrong. Which it may be, of course - any such tool of measurement brings its own variabilities.

I think the only match we've lost on xG was against West Ham. Go figure!

 

[el57]

Whilst I get what you're trying to say here, I think it boils down to you thinking that the xG metric is wrong. Which it may be, of course - any such tool of measurement brings its own variabilities.

I think the only match we've lost on xG was against West Ham. Go figure!

Huh? I don't think it's wrong at all, but I do think it is defined a certain way and there are ways in which style, period of match, and squad state come into play within that definition beyond it being just a measure of luck and clinical finishing. I suppose xG might be "wrong" if you expect every deviation to be random or from low sample size, but it's not wrong in the sense that having more raw xG is always better. The OP was not about the actual raw xG, with but *underperforming* xG which is a bit more vague on being good or bad.

 

[Edmundo]

If it doesn't take into account something that you think is important and actually changes whether or not a goal is to be expected, then by definition, you think it's wrong, or at least not a perfect measure. We all know it isn't perfect, but if there is a factor that we can identify it doesn't take into account, then it should be modified by taking it into account.

 

[el57]

If it doesn't take into account something that you think is important and actually changes whether or not a goal is to be expected, then by definition, you think it's wrong, or at least not a perfect measure. We all know it isn't perfect, but if there is a factor that we can identify it doesn't take into account, then it should be modified by taking it into account.

This would imply there should be one perfect measure. Goals scored don't take into account difficulty of shot, but that doesn't mean goals scored is a bad metric to look at. Metrics are meant to give you different ways to group together the raw numbers to get a different view, so you are supposed to look at many to get a well rounded idea. Can't just look at xG that you've accumulated without looking at xG accumulated against you as well, for example.

You can't account for everything in one metric, otherwise you wont have any interesting groups or patterns; it will just be a unique but boring history of touches by each player. Plus, xG isn't even one statistic, it differs from company to company, so for all I know the Brentford in-house version does "adjust" for the stuff I mentioned. Or maybe not, and they instead account for it through a different view from another metric. Either way, I'm sure the club have a sophisticated and smart way of breaking it down.

 

[el57]

actually changes whether or not a goal is to be expected

Actually, I think I see the area of disagreement here, you're specifically targeting the idea about whether a goal is expected. I'm saying that what you will expect depends on what baseline perspective you want to take and a statistic can only take one. You can make a perfect one for Brentford, but then it won't give perfect expectations from Chelsea's perspective and their expectations.

Understat probably has their model calibrated to all the top European leagues over some # of years so that its "expected for random player on arbitrary club in top European league". But you could calibrate it instead just to Premier League data and get different expectations; now its "expected for random player on arbitrary Premier League club". You can refine it even more and calibrate the same data to a baseline made of only the clubs who get relegated (or alternatively who come top 4) and then its "expected for random player on arbitrary relegated (top 4) team". Brentford might account for having to bed in new attackers that need time to gel by including a control for players joining new clubs. That would still be xG, but it would be an custom in-house version, not one of the public versions that we as fans have access to.

 

[Muzo]

First of all, I believe that Xg is a significant indicator of the success of the club. But I do have a thesis about why the reality "underperforms" the common Xg statistics.

Let us say that we are applying for an IVY league Business School.
The application consists of 1) GMAT-test 2) resume 3) application.
We have identified that the GMAT test is the most significant variable to ensure our success. So we hire a tutor and plan our schedule around studying, which naturally provide a terrific score.

The competitors (the other PL teams) have a diverse set of strategies. Some are data-driven and will use the same strategy as us. They spend the time studying for the test, while others will do volunteer work or use their time on other valuable endeavours for the application.

The different strategies create a gap in the Xg-statistics it self. Some teams (like us) optimize their score, while others are measured but only show their natural ability. They have spent their time training set-pieces (the old Stoke strategy) or investing in players with a lot of speed. If Stoke bought five players over 190 cm for set-pieces, that is probably not naturally factored into the Xg percentages in the standard analytical models. It might be in our clubs model, though.

There will always be a bias giving us a generous Xg-score. Not because the club cares about the Xg-score, but because a high Xg-score is a natural biproduct of our data-driven strategy.

The interesting question is then: Considered this bias - how many % do we consistently have to outperform the Xg-index to ensure survial in Premier League?

 

[W13 Bee]

There will always be a bias giving us a generous Xg-score. Not because the club cares about the Xg-score, but because a high Xg-score is a natural biproduct of our data-driven strategy.

Surely a chance has a certain probability attached to it, based on 1000's of games. Nothing more to it than that?


We actually scored 8 goals more than we 'should' have done last season, but we were 13 points worse off (on one model).
So it was the distribution of goals that wasn't favourable to us. Perhaps - simplistically - where we won heavily, we 'shouldn't' have won with such a big score and where we lost narrowly, we 'should' have either drawn or won.

It is really frustrating that the unfavourable distribution of points is happening yet again this season, with the xg tables saying we are around 5th with 19 or so points.

 

[Eieieio]

Because xG isn't real and doesn't work. This whole stats delusion that has festered in this fanbase is ridiculous and it's propagated mostly by our manager who can't seem to be able to admit when we have played badly. Finishing is part of the game as much as anything else so no we didn't deserve three points against Norwich or against Leicester or Brighton or Chelsea. Less talk more scoring goals please

Can only presume this guy has been in a coma for the last few years and isn't aware we're in the EPL

 

[Thomas Bathurst]

Not putting away your early chances, and maintaining a similarly attacking approach throughout matches will make you more likely to underperform xG compared to the reverse. Why?

...

Calmly holding onto your lead = lower xG because you aren't shooting a ton = (probably) overperform xG because you've already scored
Chasing matches to get a better result = higher xG because you have to shoot and attack more = (probably) underperform xG because you are already lacking the goals for the result you want

I reckon this probably comes closest to explaining our ongoing xG anomalies. It's amazing how often we miss a big chance in the first 5 minutes of a game (this season at least). Score that (or certainly don't concede 1st) and we're in a different game - teams have to come out and attack us more and we can pick them off / open them up with greater ease as opposed to trying to break down a dogged defence or creating the sort of chances where everyone is back in the box and we're piling up the xG score simply because the other team chooses to sit back themselves.

 

[AB]

Actually, I think I see the area of disagreement here, you're specifically targeting the idea about whether a goal is expected. I'm saying that what you will expect depends on what baseline perspective you want to take and a statistic can only take one. You can make a perfect one for Brentford, but then it won't give perfect expectations from Chelsea's perspective and their expectations.

Understat probably has their model calibrated to all the top European leagues over some # of years so that its "expected for random player on arbitrary club in top European league". But you could calibrate it instead just to Premier League data and get different expectations; now its "expected for random player on arbitrary Premier League club". You can refine it even more and calibrate the same data to a baseline made of only the clubs who get relegated (or alternatively who come top 4) and then its "expected for random player on arbitrary relegated (top 4) team". Brentford might account for having to bed in new attackers that need time to gel by including a control for players joining new clubs. That would still be xG, but it would be an custom in-house version, not one of the public versions that we as fans have access to.

I think there’s something in the contextual xG you described that relates to game state and whether a team is chasing or holding. So it is misleading to look at the 90 min total v Chelsea and say we should have won 2-1 rather than lose 0-1. Had we scored early or equalised in our final 25 minute onslaught Chelsea may have played differently. That probably can be modelled but is complex. What the bare number does say though is that we made a number of good opportunities so were still in the game rather than outclassed. Pinpointing the errors is easy. Looking to see how we can make more of our shots result in goals and Fixing the errors in game or between games is harder and where coaching really comes in.

Over the last 3 seasons we’ve increased our xG by moving our average attacking position from being edge of the penalty area to inside the box. Almost all our conceded goals have been from further out. That looks like an xG driven approach. But maybe we are finding that in the PL we face players of high enough quality both to convert longer range shots more regularly (eg in the Championship there was only one Buendia, now most teams have one, and we’re shooting against top international goalies). So there might be an argument for mixing it up a bit more now as our opponents have done and being ready to find and take chances from further out rather than focus on close range overloads.

 

[Nottsbee]

Thomas Bathurst asks a great question and one that I have been pondering on for quite some time.

In past seasons, I put it down to various factors. Gray, Maupay, Vibe, even Ollie Watkins were all prolific without being clinical in their finishing. I still have nightmares about Neal's miss at Cardiff. That they managed to score so many was down to the extraordinary number of chances that we manage to create. There was the season when we succumbed to about 9 or 10 sh*thoused 1-0 away defeats. And remember the spell during which Daniel Bentley was throwing them in for fun - a sure way of confounding the xG figures.

This season, the poor finishing and even more devious sh*thousery from opponents are still factors. But the of finishing in the chances converted against us is on a completely different level. That doesn't mean there's nothing we can do about it. There were great strikes from Jones, Bowen, Chilwell and Tielemanns - but they were all in acres of space. There were brilliant solo goals from Buendia, Trossard and Normann but they were allowed to waltz through a number of players unchallenged. On the positive side, we have conceded very few high value xG chances from inside the box.

At the other end, we have underperformed xG due to a combination of bad luck, outstanding goalkeeping and poor finishing. Can we improve Bryan's finishing ability? Would it help if Ivan was further up the pitch to be on the end of those chances? Will Pontus ever score the goal he so richly deserves?

So though the superstitious pessimist in me fears that xG underperformance is a permanent condition and built into Brentford's DNA, my more rational side still trusts in Thomas Frank and the data.

 

[Shamrock_Bee]

Can only presume this guy has been in a coma for the last few years and isn't aware we're in the EPL

He certainly doesn't agree with the stats and analysis side of the game lol. In saying that he has a certain point, you can have all the stats in the world, but if you don't score goals I'm in the camp you don't deserve to win the game of football despite what xg etc tells you should have happened.

Although I do agree stats and analysis can be used to find out why we aren't scoring goals or why we're conceding goals for future games.

 

[AB]

He certainly doesn't agree with the stats and analysis side of the game lol. In saying that he has a certain point, you can have all the stats in the world, but if you don't score goals I'm in the camp you don't deserve to win the game of football despite what xg etc tells you should have happened.

Although I do agree stats and analysis can be used to find out why we aren't scoring goals or why we're conceding goals for future games.

We’ve scored 13 goals despite Ivan and Bryan not getting as many as we’d want. That’s more than 9 other teams. Ten teams have conceded more than our 14. In pure goal terms we’re nearly identical to Arsenal in fifth.

 

[olliebfc]

Some interesting points on this thread about game context/stage of the 90, missing close range chances vs having people who take risks and shoot well from far, which have always been in my mind since xG got trendy. I’m sure Brentford use a much more complicated measure then xG and I think some internet fans (Twitter) like to forget you need to watch a game with your eyes combined with the stats to tell a whole story.
I’m an analyst and if an area in the organisation fail a KPI they generally need to write an exception report to give context and a plan to improve going forward as I’m not the person on the ground so the RAG rated percentage isn’t enough to purely tell a story - it’s a flagging system so things can’t be ignored. You’d hope in a football club they’d use the stats (kpis) and watch the tapes and explain what went wrong ‘exception report’ amongst themselves in training to improve (finishing for me) even if to us emotional fans it looks like they are making the same mistakes repeatedly and saying to the media everything is ok. I don’t really believe what Thomas tells the media is more then 50% what he truly believes himself, it’s all a game, the same as organisations won’t tell outsiders everything unless legally required. Things take time and as has been proven the club hierarchy know what they’re doing. The minute the stats and the eyes are no longer tallying up or showing signs of improving I’m sure big changes are made ala Marinus, or bringing extra coaches in under our various bosses. I was so pissed off at the Norwich result but I’ve just got to trust the whole club that got us here that they can correct things eventually - as they have got us here in a sustainable fashion with lots of changes already. Plus it helps I don’t think we’re going down at all.

 

[Shamrock_Bee]

We’ve scored 13 goals despite Ivan and Bryan not getting as many as we’d want. That’s more than 9 other teams. Ten teams have conceded more than our 14. In pure goal terms we’re nearly identical to Arsenal in fifth.

Yes those stats are true but I think the poster who didn't agree with the benefits of stats was more making a point about xg etc telling us on chances created we should have won or got a result in the Norwich, Brighton, Chelsea and Leicester games and didn't deserve to lose.

The only stat that matters is we didn't convert the chances or get the goals required so from that point of view I'm a bit old fashioned and would agree with him, we didn't deserve the result. Those four might of had less chances but they converted them and particularly in the PL it comes down to performance on the field as all clubs at this level spend significant resources on stats and analysis systems that suits their needs.

In saying that I 100% agree with using stats and our system is one of the best and gone a long way to bringing us to where we are. They can certainly be used to help us convert the chances we're making or stopping the long range goals we're conceding a lot of in the PL.

However I wouldn't move away from scoreline is the only stat that really matters on 95 plus minutes, yes you can be unlucky but even with that scoreline usually tells the truth and our league position currently is probably right on performances over all the games.

 

[AB]

Yes those stats are true but I think the poster who didn't agree with the benefits of stats was more making a point about xg etc telling us on chances created we should have won or got a result in the Norwich, Brighton, Chelsea and Leicester games and didn't deserve to lose.

The only stat that matters is we didn't convert the chances or get the goals required so from that point of view I'm a bit old fashioned and would agree with him, we didn't deserve the result. Those four might of had less chances but they converted them and particularly in the PL it comes down to performance on the field as all clubs at this level spend significant resources on stats and analysis systems that suits their needs.

In saying that I 100% agree with using stats and our system is one of the best and gone a long way to bringing us to where we are. They can certainly be used to help us convert the chances we're making or stopping the long range goals we're conceding a lot of in the PL.

Yes, I’m not seriously going to tot up the points we “ought” to have won to say we “ought” to be 4th and therefore lalalala nothing to see here! There’s a big difference between what is and what ought to have been. But the club and anyone looking at stats sensibly uses them as long term measures knowing that there’s also a lot of luck involved. The stats suggest we have it in us to be doing better and haven’t simply been outclassed. If we react right in making the tweaks to get the results we should be safe and be ready to improve next season (cf Brighton who underperformed their xG by 20 points last season but have calmly kept faith with Potter, recruited well and are now getting the results, including shading it to beat us at the death).

We may well lose 15+ games this season and those can and will come against teams at either end of the table. But we can play positively enough to have a chance of a result against anyone as the liverpool game showed. Losing on Saturday aside, that’s the difference with Norwich under Farke who basically said they had no chance against the top teams. Like Burnley under Dyche, we believe we have the chance to get something against anyone and are realistic that also means we can lose to anyone. It’s quite different from the Championship where on quality for the last two seasons we were obviously one of the best 3-4 teams, maybe 1st or 2nd in squad quality, and were a little disappointing in not quite being remorseless enough through 46 games to get automatic promotion. Now we’re one of the weakest and need every little edge, decision and bobble we can get.

 

[Shamrock_Bee]

Yes, I’m not seriously going to tot up the points we “ought” to have won to say we “ought” to be 4th and therefore lalalala nothing to see here! There’s a big difference between what is and what ought to have been. But the club and anyone looking at stats sensibly uses them as long term measures knowing that there’s also a lot of luck involved. The stats suggest we have it in us to be doing better and haven’t simply been outclassed. If we react right in making the tweaks to get the results we should be safe and be ready to improve next season (cf Brighton who underperformed their xG by 20 points last season but have calmly kept faith with Potter, recruited well and are now getting the results, including shading it to beat us at the death).

We may well lose 15+ games this season and those can and will come against teams at either end of the table. But we can play positively enough to have a chance of a result against anyone as the liverpool game showed. Losing on Saturday aside, that’s the difference with Norwich under Farke who basically said they had no chance against the top teams. Like Burnley under Dyche, we believe we have the chance to get something against anyone and are realistic that also means we can lose to anyone. It’s quite different from the Championship where on quality for the last two seasons we were obviously one of the best 3-4 teams, maybe 1st or 2nd in squad quality, and were a little disappointing in not quite being remorseless enough through 46 games to get automatic promotion. Now we’re one of the weakest and need every little edge, decision and bobble we can get.

Good post and I agree! The big difference between the two leagues as well though is we could genuinely compete in every game but lose over 20 of them by a single goal despite having the bulk of possession and chances.

Stats; Xg are great and can be a huge benefit to us but competing against other world class coaches and good statistical models will reduce it's effectiveness, but I agree it gives us a good pathform to make sensible decisions and learn what we need to do in the future.

Luck plays a part but I don't think it plays too big a part. The Brighton example is a perfect one as I do think models can improve your performance over a period of time but balancing that I also think you can't remain static as what you have improved will be found out very quickly by these other PL clubs. It'll be interesting to see if Brighton do as well when they play teams second time around this season.

 

[SH]

Here's the data... on (infogol) xG we "should" be 5th not 14th, because Christian "should" have three goals rather than one, Bryan three or four rather than two and Ivan three rather than two... and that's the four goal "shortfall" in goals for. Our goals against is bang on. Which is pretty much what we all know with our own eyes: Christian's been unlucky, Bryan's finishing has been poor by PL standard and Ivan's not quite at his usual high standard.

Last year Brighton were the great xG underperformers, which is why they were heavily touted to do better this year. This year they're five places higher than they "ought" to be (as we know from the game at BCS). If you read one of Rasmus's books, he gives a great example of how Newcastle dramatically outperformed their xG a few years back. In the first half of one season (2012, I think), one of their centre forwards basically took every chance going going to a ridiculous level, and then in the second exactly the same happened with forward #2. As a result, they finished 5th when they "should' have been 5th from bottom or some such. The next year they duly finished 16th.

Eventually the law of averages catches up... so keep backing Bryan to score and Bees to win

 

[Banana]

Very quick answer, I reckon it is because we are using xG (or similar) as a measure of how well we are doing. So we put emphasis on creating chances, which we do well.

The old adage of "if you measure something then you change it" applies here.

 

[el57]

Very quick answer, I reckon it is because we are using xG (or similar) as a measure of how well we are doing. So we put emphasis on creating chances, which we do well.

The old adage of "if you measure something then you change it" applies here.

In the Championship I think we may have stuck out a bit on the stats approach, but in the Prem my sense is that it's all standard now.

But there's an interesting wrinkle in what you're saying that maybe my earlier way of framing it is missing, which is that you can ignore the between-club comparisons and ignore the under/over performance stuff and just take your starting lineup+bench+formation as a fixed thing, and ask "How do I maximize the xG over season with these players?" That's probably closer to what TF's remit is, and I do see a risk of what you say, i.e. "when your measurement becomes the target, it's no longer measuring"

Would explain why we put *so much work* into innovating on set piece routines, long throws, cutbacks, committing more men forward etc. We can raise xG here since we actually control these parts of the game through coaching and strategic data analysis, whereas we probably can't expect to raise xG just by getting more open-play shots from better locations without player upgrades. I've heard Klopp say something to that effect so I'm sure it's well-known but we do seem to be quite good at it even in PL level.

 

[Bee(h)Iver]

I think there is a lot to digest here, and I'm guessing the club uses more stats than just simple xG to assess our performance.

We have had more than our fair share of bad luck with injuries, had keepers pull off some worldies and been punished by a few of those '1 in a 100 go in' strikes. There are a lot of things we are doing well, and a few things that we need to improve.

At the start of the season if someone had said after 11 games we'd have 12 points, be 4 points clear of the bottom 3 and have a -1 goal difference I'd have taken that. It's just the way the results have skewed that has us worrying. If we can 12 points from our next 11 we're well ahead of schedule to stay up and the fact that we have been capable of taking 11 of those from teams in the top half of the table gives me hope we can achieve that.

The only thing I can say with much certainty is that if our xG was considerably lower than our opponents I'd be more worried.