[Brighton] XPG - A Fourth Place Finish

[Jim in the West]

Wrong.

The xG table is putting 7 out of the 20 teams in exactly the right position. A further 8 within +/- 2 places of the correct position. As an "after the fact" prediction tool, that's actually pretty good. You've then got 5 teams who diverge a bit further. 2 of them are only 3 places incorrect (Everton, Palace), but in both cases it's looking like the divergence is a product of fine margins in a handful of games. Palace ... just got to look at the robbery they pulled off against us, for a prime example.

So: three teams then that are genuinely sticking out as outliers. Tottenham (explained by how clinical Kane and Son were early season. Once those 2 came off the boil a bit, they began sliding out of CL contention). Man U (are they in for a difficult season ahead?). And Brighton. We all know why.

Professional gamblers will be looking at this (along with other stats and key signings over the summer). It'll be affecting their pre-season bets. I think, pre-season, if we can keep Potter, reinforce in attack, and keep most of our key players, you'll see plenty of other teams being more favoured for the drop than us. I certainly don't see us pushing for Europe places next season, but a solid mid-table finish is within reach. You'll also likely see Man U behind a few other teams in the betting for Champions despite their solid 2nd this season.

If you look at last season's xG/xPts stats, it effectively predicted that Sheff Utd, Brighton, Newcastle, Villa, West Ham, Burnley and Palace would struggle (despite Sheff Utd and Burnley finishing 9th and 10th respectively). The outcome was pretty close....obviously Villa spent big (in particular on a decent keeper and a decent goalscorer). West Ham picked up Coufal and Soucek, who had amazing seasons. Logic says we are VERY close to being a top 10 side....we just need to follow your advice!

 

[blue-shifted]

Wrong.

The xG table is putting 7 out of the 20 teams in exactly the right position. A further 8 within +/- 2 places of the correct position. As an "after the fact" prediction tool, that's actually pretty good. You've then got 5 teams who diverge a bit further. 2 of them are only 3 places incorrect (Everton, Palace), but in both cases it's looking like the divergence is a product of fine margins in a handful of games. Palace ... just got to look at the robbery they pulled off against us, for a prime example.

So: three teams then that are genuinely sticking out as outliers. Tottenham (explained by how clinical Kane and Son were early season. Once those 2 came off the boil a bit, they began sliding out of CL contention). Man U (are they in for a difficult season ahead?). And Brighton. We all know why.

Professional gamblers will be looking at this (along with other stats and key signings over the summer). It'll be affecting their pre-season bets. I think, pre-season, if we can keep Potter, reinforce in attack, and keep most of our key players, you'll see plenty of other teams being more favoured for the drop than us. I certainly don't see us pushing for Europe places next season, but a solid mid-table finish is within reach. You'll also likely see Man U behind a few other teams in the betting for Champions despite their solid 2nd this season.

Don't really agree

Not sure that XG is an accurate predictor of the future. So it's not luck, which will even out in the long term, that we're underperforming it so badly. It's because our shooting is about as threatening as a eunuch's crotch. There's nothing to say this is going to improve next season. If we go out and sign Toney or Wood or some other obviously good striker, then we'll be a good bet to finish higher in the table. but you wouldn't need XG to tell you this.

 

[dazzer6666]

It's a start but the analysis will need to be much deeper than that (and will be - POtter's team will be all over it obviously) - the table doesn't, for example, make any distinction re the number/quality of chances - are we creating dozens of fairly poor chances, or a missing a small number of very good chances ? In several games I seem to recall having a very large number of attempts (so XG will be high on a cumulative basis even if every chance was crap) but not scoring many. Five really good chances could give a much better game XG than 25 'lesser' chances (Bissouma shots, for example LOL)..............

 

[BadFish]

XPG is clearly not as important as actually bonafide points. This much is obvious.

They say over here in Aus that bad goal kicking is bad footy and this also stands to reason.

However . . .

There was an old adage that you build a new team from the back. As we continue to improve, finishing may be the final piece of the jigsaw.

What the XPG tells us is that if we improve in our finishing we can be a really good team. to mean this is what it means . . . and fair enough we don't need it to tell us that because we can see it with our own eyes. It is nice to have some stats to back up what we are seeing though.

 

[zefarelly]

XG, wasn't it the B side to this

https://www.youtube.com/watch?v=xvqeSJlgaNk

 

[Swansman]

Looking forward to a lengthy discussions between three sides:

- Impressive, shows how good we are
- Depressing, shows our weaknesses in getting those points
- In 1953 there was nothing like this so its not interesting. I also throw my garbage in the river.

Who is right or wrong? Maybe no one or everyone.

 

[BadFish]

Looking forward to a lengthy discussions between three sides:

- Impressive, shows how good we are
- Depressing, shows our weaknesses in getting those points
- In 1953 there was nothing like this so its not interesting. I also throw my garbage in the river.

Who is right or wrong? Maybe no one or everyone.

Everyone is right - apart from throwing rubbish in the river, this is wrong.

 

[dsr-burnley]

Wrong.

The xG table is putting 7 out of the 20 teams in exactly the right position. A further 8 within +/- 2 places of the correct position. As an "after the fact" prediction tool, that's actually pretty good. You've then got 5 teams who diverge a bit further. 2 of them are only 3 places incorrect (Everton, Palace), but in both cases it's looking like the divergence is a product of fine margins in a handful of games. Palace ... just got to look at the robbery they pulled off against us, for a prime example.

So: three teams then that are genuinely sticking out as outliers. Tottenham (explained by how clinical Kane and Son were early season. Once those 2 came off the boil a bit, they began sliding out of CL contention). Man U (are they in for a difficult season ahead?). And Brighton. We all know why.

Professional gamblers will be looking at this (along with other stats and key signings over the summer). It'll be affecting their pre-season bets. I think, pre-season, if we can keep Potter, reinforce in attack, and keep most of our key players, you'll see plenty of other teams being more favoured for the drop than us. I certainly don't see us pushing for Europe places next season, but a solid mid-table finish is within reach. You'll also likely see Man U behind a few other teams in the betting for Champions despite their solid 2nd this season.

I think you're overgenerous in your marking there.

For example, my club Burnley - actually 11 points clear of relegation, that table has us relegated by 5 clear points. Certainly within a couple of places of the correct position, but also significantly wrong.

I count it this way. Suppose this table is asked to predict every club's total points within a 21 point range. They can have the target score and up to 10 points above or below, so Fulham, for example, prediction 39, expected range 29-49. That's a huge range. And yet they got it wrong, it was only 28. Burnley are 17-37, Brighton 58-78, Chelsea 79-99, Man U 53-73. These are big, big ranges, but they still got 8 of the 20 wrong.

You could do as well by random guesswork at the start of the season. You don't need to wait till after the event to get it as wrong as that!

 

[Lethargic]

Like all stats it should not be taken too seriously and I avoid diving into specifics but always useful for trends.

On that note if you take a look at the table it generally maps onto the finishing positions of the league and resonates with what many say (eg Palace are very efficient with their few shoots) tops 3 obvious etc, so there must be something in it to show we re definitely making progress and the anomaly, I take this as a positive sign.

 

[eaglesdan]

All the time points are given out based on actual goals scored I couldnt really care less about XPG. Its not like ice skating or gymnastics where you get points for artistic merit. As has been said on other threads, we were battered by BHA in both games this season, with a sum total of 2 shots on target across both games, excluding the penalty, and yet we gained 4 points. Thats the beauty of football in my opinion, and I would say the same if the roles were reversed. Just my opinion though

 

[Guinness Boy]

Looking forward to a lengthy discussions between three sides:

- Impressive, shows how good we are
- Depressing, shows our weaknesses in getting those points
- In 1953 there was nothing like this so its not interesting. I also throw my garbage in the river.

Who is right or wrong? Maybe no one or everyone.

In 1953 gas guzzling SUVs in town centres, Pret a Manger and David Guetta also didn't exist. All are interesting to a degree - particularly if you are a invested in them as a business - but none are necessarily good or helpful

 

[Audax]

Don't really agree

Not sure that XG is an accurate predictor of the future. So it's not luck, which will even out in the long term, that we're underperforming it so badly. It's because our shooting is about as threatening as a eunuch's crotch. There's nothing to say this is going to improve next season. If we go out and sign Toney or Wood or some other obviously good striker, then we'll be a good bet to finish higher in the table. but you wouldn't need XG to tell you this.

You might not ... but they do. Not *just* xG, but xG plays a role. They also look at goal diff and a range of other metrics to get a feel for whether a team has over-performed, done about right, or under-performed over a season. They'll then look at summer transfers, and make a call on what they think will happen in the following season. There's a famous example with Newcastle (video in the post above mine that you quoted, actually) - They finished 5th in 11/12, but for the following season they were not fancied by pro gamblers and delivered a poor season. Why didn't the gamblers like them? Because the underlying stats of their 5th place finish clearly showed that they had massively over-performed, and that performance level probably wasn't sustainable unless they spent on transfers. They didn't spend, the following season they sank down the table. The pro-gamblers got it right, the so-called expert pundits who love to poo-poo xG etc and trot out the "the table doesn't lie" line got it wrong.

There's no guarantees in Football, though. It can be a game of very fine margins (and far too many times this season, we found ourselves on the wrong side of the margin). Next season, our striking could continue to stink up the league and see us once again in the bottom quarter and fighting relegation for most of the season. Or Maupay and Connolly might come good, Trossard find some consistency, we sell Bissouma but Caicedo comes in and contributes a few more goals and we shoot up the table.

Fact is - stats like these tell us what our eyes told us: we under-performed this season, by a lot. We have the potential to shoot up the table next season. That will be reflected in the odds for the season when they come out. I would hazard a guess there will be at least 7 clubs more fancied for relegation than us. From the current season, I'd suggest Burnley, Saints, Palace, Newcastle. And then throw in the 3 promoted sides. Maybe Villa into that mix if they sell Grealish.

 

[Seasidesage]

xG is an indicator nothing more and its a fairly unsophisticated one at that. It's a bit like not understanding why the car isn't going very fast even though the rev counter is high without considering the 3 tons of scrap you are towing and the gear you are in! But it does show that we create a lot of shooting opportunities that don't become goals (who knew?) what it doesn't show other than as a rough measure is the quality of chances created. While my eyes clearly tell me we are a better team than our league position says we are I don't think we are 27 points behind where we should be?

The data analysts the club employ will have been constantly looking at a much broader range of metrics both at team level and for the individual players and I would guess as well as a target list of replacements there is probably a valuation of existing players both current and potential and a list of those we could replace if offers came in and for how much? This is very much a science and data is already incredibly important and only going to get more so over the next few years. It is not bollocks but xG alone is not the holy grail either...

Data and its manipulation is at the very core of how TB thinks he can buck the market and if as a football club you are not making decisions using it you are going to fail. The only question is who understands that data the best and who can utilize it effectively where it matters on the pitch..

 

[blue-shifted]

You might not ... but they do. Not *just* xG, but xG plays a role. They also look at goal diff and a range of other metrics to get a feel for whether a team has over-performed, done about right, or under-performed over a season. They'll then look at summer transfers, and make a call on what they think will happen in the following season. There's a famous example with Newcastle (video in the post above mine that you quoted, actually) - They finished 5th in 11/12, but for the following season they were not fancied by pro gamblers and delivered a poor season. Why didn't the gamblers like them? Because the underlying stats of their 5th place finish clearly showed that they had massively over-performed, and that performance level probably wasn't sustainable unless they spent on transfers. They didn't spend, the following season they sank down the table. The pro-gamblers got it right, the so-called expert pundits who love to poo-poo xG etc and trot out the "the table doesn't lie" line got it wrong.

There's no guarantees in Football, though. It can be a game of very fine margins (and far too many times this season, we found ourselves on the wrong side of the margin). Next season, our striking could continue to stink up the league and see us once again in the bottom quarter and fighting relegation for most of the season. Or Maupay and Connolly might come good, Trossard find some consistency, we sell Bissouma but Caicedo comes in and contributes a few more goals and we shoot up the table.

Fact is - stats like these tell us what our eyes told us: we under-performed this season, by a lot. We have the potential to shoot up the table next season. That will be reflected in the odds for the season when they come out. I would hazard a guess there will be at least 7 clubs more fancied for relegation than us. From the current season, I'd suggest Burnley, Saints, Palace, Newcastle. And then throw in the 3 promoted sides. Maybe Villa into that mix if they sell Grealish.

I don't think that's what our eyes say or the stats say. We didn't underperform. We were brilliant in the middle third and defence but woeful in front of goal. 38 games is enough for any peaks and troughs of luck to iron themselves out. If we keep the same attacking players there's nothing in XG or XP to suggest that the players who failed this time are going to do any better next time.

I still think XG can tell you a story about a team. I just don't see how it can predict anything which happens in the future.

 

[Pevenseagull]

Can you bet on xG yet?

One of the bookies did a promotion awhile back where you got your stake back as a free bet it your backed team lost but had higher xG.

 

[Kalimantan Gull]

Expected goals is a useful stat for individual players, or moments, for the analysis of a squad and tactics and so on. But a points table based on xG is laughably pointless, for the simple reason that each part of a match is not created equal.

Consider the team losing 1-0 against a weaker team. They will batter and batter them, racking up 'expected goals' until the moment they get an actual goal. Then, once the scores are level, the pattern of the game will inevitably shift. Now that team has something to defend and will apply different tactics accordingly.

So the total value of the xG during that period they were behind depends entirely on when they actually score. Put the first chance away, the xG will be low. Take ten goes to score, the xG will be much higher. And yet the match will likely play out the same way in the end. All this stat tells us is that we were rubbish at scoring chances. It does NOT tell us that only bad luck stopped us finishing high up the table.

 

[Stato]

A spotty 16 year olds website says we should be in Europe. Forgive me if I don't dig out my Passport...

Your spotty 16 year old is actually Ben Mayhew, head of data analysis at PA Media (formerly the press association). His work is credited in articles in the Mirror, Independent, Mail, Standard, as well as several Irish and Scottish and online publications.

Xg is just an interesting and sometimes useful analytical tool. Anybody arguing that it should be able to get everything right is misunderstanding its intent. Of course real world results are what matter and analytical tools cannot take into account every possibility in a dynamic game like football. I've been reading Experimental 361 for years. It is very well put together and, if you have an analytical bent. its fascinating. Here's it's end of season scatter graphic which hints at the way teams are set up to play and how well they have achieved their plan: https://experimental361.com/2021/05/24/scatter-graphics-premier-league-2020-21/

The mistake is to think that it predicts the future. A wide disparity between Xp & Xg won't always revert to the mean. It could be a suggestion that a team is set up to dominate, to create chances, but that it is giving its opponents some significant advantages by doing so, or vice versa. Us and Palace are the classic examples of this for the 2020/21 season. Their game plan has been based on facing a lot of shots, but putting away the chances that they get. Ours has been to involve the whole side in creating chances. As we can see from our games against them, this can mean that the few chances that they get happen in spaces and against backtracking defenders. Ours tend to occur in front of a packed area. They haven't been particularly successful in defending against heavy pressure, but they've been very good at putting away the few chances that they create. We have conceded few shots, but struggled to score our chances. I argue that this is because our playing style leans towards creating a lot of chances and needs to instead find ways of creating space for the attacking players to manoeuvre in.

I think that Xg would be more generally accepted if it were named something else. It's about run of play. We are all familiar with the concept of goals being scored against the run of play. When Hyypia was our manager we were all to familiar and it wasn't about bad luck, it was about over committing in some areas and risking too much in others. Talking about it in these terms rather than in terms of goals that 'could have happened' would put distance between the findings and consumer expectations. It's never going to be perfect because football has far too many measurables to be easily predictable. That, as eaglesdan has posted, is why we all love it. However, the analysis is not meaningless either. If you take a look at the performances of teams that have under-performed their Xp over each full season (this one goes back to 2014/15 https://understat.com/league/EPL/2020), you'll generally see that those who under-performed are roughly two thirds more likely to be found in the bottom half of the table. Only 2 from 24 teams relegated in the eight seasons shown have over-performed and still been relegated.

In our case you can see a distinct shift between being very close to our Xp in the first two seasons where we all knew that we needed to be, and were, performing to our maximums in order to survive, to first a significant and then an unprecedented under-performance in our next two seasons. This just states a measurement. The underlying potential meanings are what make it interesting. Does it mean that we are more competitive, but not yet able to put teams away? Does it show the sea change between Hughton and Potter? What is the significance of Murray's absence? A lot of people argue that Xg is meaningless because it fails to give or provides the wrong answers. I view the tool as more useful for considering the right questions. Others may argue that they can see problems without having underlying data, but the point made in Moneyball is that this can lead to group think and leave gaps for exploitation available to those who don't just rely on their perception alone.

 

[jakarta]

Your spotty 16 year old is actually Ben Mayhew, head of data analysis at PA Media (formerly the press association). His work is credited in articles in the Mirror, Independent, Mail, Standard, as well as several Irish and Scottish and online publications.

Good for him for conning a living out of people who think anything except results actually means anything.

 

[dazzer6666]

Your spotty 16 year old is actually Ben Mayhew, head of data analysis at PA Media (formerly the press association). His work is credited in articles in the Mirror, Independent, Mail, Standard, as well as several Irish and Scottish and online publications.

Xg is just an interesting and sometimes useful analytical tool. Anybody arguing that it should be able to get everything right is misunderstanding its intent. Of course real world results are what matter and analytical tools cannot take into account every possibility in a dynamic game like football. I've been reading Experimental 361 for years. It is very well put together and, if you have an analytical bent. its fascinating. Here's it's end of season scatter graphic which hints at the way teams are set up to play and how well they have achieved their plan: https://experimental361.com/2021/05/24/scatter-graphics-premier-league-2020-21/

The mistake is to think that it predicts the future. A wide disparity between Xp & Xg won't always revert to the mean. It could be a suggestion that a team is set up to dominate, to create chances, but that it is giving its opponents some significant advantages by doing so, or vice versa. Us and Palace are the classic examples of this for the 2020/21 season. Their game plan has been based on facing a lot of shots, but putting away the chances that they get. Ours has been to involve the whole side in creating chances. As we can see from our games against them, this can mean that the few chances that they get happen in spaces and against backtracking defenders. Ours tend to occur in front of a packed area. They haven't been particularly successful in defending against heavy pressure, but they've been very good at putting away the few chances that they create. We have conceded few shots, but struggled to score our chances. I argue that this is because our playing style leans towards creating a lot of chances and needs to instead find ways of creating space for the attacking players to manoeuvre in.

I think that Xg would be more generally accepted if it were named something else. It's about run of play. We are all familiar with the concept of goals being scored against the run of play. When Hyypia was our manager we were all to familiar and it wasn't about bad luck, it was about over committing in some areas and risking too much in others. Talking about it in these terms rather than in terms of goals that 'could have happened' would put distance between the findings and consumer expectations. It's never going to be perfect because football has far too many measurables to be easily predictable. That, as eaglesdan has posted, is why we all love it. However, the analysis is not meaningless either. If you take a look at the performances of teams that have under-performed their Xp over each full season (this one goes back to 2014/15 https://understat.com/league/EPL/2020), you'll generally see that those who under-performed are roughly two thirds more likely to be found in the bottom half of the table. Only 2 from 24 teams relegated in the eight seasons shown have over-performed and still been relegated.

In our case you can see a distinct shift between being very close to our Xp in the first two seasons where we all knew that we needed to be, and were, performing to our maximums in order to survive, to first a significant and then an unprecedented under-performance in our next two seasons. This just states a measurement. The underlying potential meanings are what make it interesting. Does it mean that we are more competitive, but not yet able to put teams away? Does it show the sea change between Hughton and Potter? What is the significance of Murray's absence? A lot of people argue that Xg is meaningless because it fails to give or provides the wrong answers. I view the tool as more useful for considering the right questions. Others may argue that they can see problems without having underlying data, but the point made in Moneyball is that this can lead to group think and leave gaps for exploitation available to those who don't just rely on their perception alone.

Nailed it......

It's one of a number of measures and indicators - it shouldn't be lauded as the be-all and end-all or anywhere near it, but it's equally daft to dismiss it. Every professional team (and a lot of semi/non professional teams) is using data like this.

 

[Icy Gull]

- In 1953 there was nothing like this so its not interesting. I also throw my garbage in the river..

I think throwing your garbage in the river was a thing in 1853 not 1953

Now throwing supermarket trolleys in the river is a very current thing