Silvio Dante Posted October 9, 2019 Report Share Posted October 9, 2019 Quite often recently both on here and in the wider world, the term xG (expected goals) has been used to show how many goals a team were expected to score in a game, and therefore if they’re lucky/over performing Cards on the table, I find the xG thing is given more weight than it should be. You may create a clear cut chance, but the dependency on it finishing is due to the skill of the attacker and goalkeeper respectively. In short, there’s no point in Bas Savage having 8 xG because in reality I expect him to score none of those. On the other side, I expect Afobe to score most of his chances and Bentley to save a reasonable proportion of the opposition xG (some oversimplification here). In short, I think it’s a decent datapoint but currently overweighted. The article in the BEP today really highlights that. Based on the expected points from xG, you’d have the following: - We should be 23rd vs 6th - Charlton should be 20th vs 10th - Forest should be 17th vs 2nd - Reading should be 11th vs 22nd - Stoke should be 7th vs 24th - Brentford should be 3rd vs 17th So, roughly 25% of the way into the season (so a very decent sized data sample), 25% of teams have a deviation of +10 places based on xG. On a statistical front, that’s enough to make it unreliable so, question is - should we even look at it? Link to comment Share on other sites More sharing options...
pongo88 Posted October 9, 2019 Report Share Posted October 9, 2019 Got to agree with you @Silvio Dante If the “experts” who produce the various statistics were as good as they say they are they wouldn’t be working. They would have retired long ago on the money they’d made betting on football results. Link to comment Share on other sites More sharing options...
GeoTheCiderHead Posted October 9, 2019 Report Share Posted October 9, 2019 18 minutes ago, Silvio Dante said: So, roughly 25% of the way into the season (so a very decent sized data sample), 25% of teams have a deviation of +10 places based on xG. On a statistical front, that’s enough to make it unreliable so, question is - should we even look at it? Well I don't think you should Link to comment Share on other sites More sharing options...
Mr Popodopolous Posted October 9, 2019 Report Share Posted October 9, 2019 The other side of the coin though is that if things regress to the mean a bit, how likely is it that Player X will keep scoring at that rate, or Player Y won't keep saving at that rate, conceding of chances will eventually lead to worse defence etc. Different League, but will Tammy for example keep up a goal a game between now and May in the PL? Should also add, Leeds would surely be top by a few points if XG was decisive. I can't say I'm fully sold on it but it has potential to be an indicator of future trends. Link to comment Share on other sites More sharing options...
Cowshed Posted October 9, 2019 Report Share Posted October 9, 2019 That is not how xG stats really work. They are not calculated over 25% of the season. In reality xG stats are supportive. Interesting as a topic here for discussion but used differently elsewhere. Stats are used to support e.g. does this player really cover enough ground, really make enough key passes, really pass the ball consistently well, score enough … Support to eyes of the Manager coach and to the player. As humans we all have bias in our views. Its natural. We focus in extremes, and form views via our experiences and opportunities which are not necessarily fact. Players and Coaches and Managers can have differing bias due to cultural drivers. Stats merely help to level our bias. Link to comment Share on other sites More sharing options...
Silvio Dante Posted October 9, 2019 Author Report Share Posted October 9, 2019 7 minutes ago, Mr Popodopolous said: The other side of the coin though is that if things regress to the mean a bit, how likely is it that Player X will keep scoring at that rate, or Player Y won't keep saving at that rate, conceding of chances will eventually lead to worse defence etc. Different League, but will Tammy for example keep up a goal a game between now and May in the PL? Should also add, Leeds would surely be top by a few points if XG was decisive. I can't say I'm fully sold on it but it has potential to be an indicator of future trends. Get that point, and indeed Leeds would be top (maybe they should play Eddie more...) My point here is that 25% is a very decent sample. The argument may be that it will level over the season (and that’s an estimate again not a definite) but isn’t it reasonable to state that it looks kind of flawed when you see the deviations it’s currently produced? Like I say, I see the benefit but I think it’s been overegged in its value thus far Link to comment Share on other sites More sharing options...
Northern Red Posted October 9, 2019 Report Share Posted October 9, 2019 xG had Spurs beating Bayern Munich last week, as opposed to losing 7-2. Link to comment Share on other sites More sharing options...
mozo Posted October 9, 2019 Report Share Posted October 9, 2019 13 minutes ago, Northern Red said: xG had Spurs beating Bayern Munich last week, as opposed to losing 7-2. That's a great example because xg can't account for the confidence levels of players or the momentum/pressure in a game. If footballers were robots of equal physical and emotional ability then xG would be a great guide. Link to comment Share on other sites More sharing options...
cidercity1987 Posted October 9, 2019 Report Share Posted October 9, 2019 Sounds about right from what I've seen this season. We aren't playing well but grinding out results with team spirit and pose in front of goal that is immeasurable. Forest are the same and yet their fans are confident it's their year. Charlton have the after burners on and are overachieving. Brentford play very well but don't take chances and always give away goals. Stoke are totally in a false position stats wise but don't have the confidence to finish teams off. Reading likewise, take the game on the weekend as a prime example, they were 'better' than us for a large part of the game. Link to comment Share on other sites More sharing options...
spudski Posted October 9, 2019 Report Share Posted October 9, 2019 They are worth watching as the season progresses, as are the scatter graphs given by Experimental 361. Check out last season's final Championship XG table compared to the actual one. The whole article as you scroll down is enlightening https://www.infogol.net/en/blog/reviews/championship-18-19-season-review-290519 Link to comment Share on other sites More sharing options...
Chivs Posted October 9, 2019 Report Share Posted October 9, 2019 1 hour ago, Cowshed said: That is not how xG stats really work. They are not calculated over 25% of the season. In reality xG stats are supportive. Interesting as a topic here for discussion but used differently elsewhere. Stats are used to support e.g. does this player really cover enough ground, really make enough key passes, really pass the ball consistently well, score enough … Support to eyes of the Manager coach and to the player. As humans we all have bias in our views. Its natural. We focus in extremes, and form views via our experiences and opportunities which are not necessarily fact. Players and Coaches and Managers can have differing bias due to cultural drivers. Stats merely help to level our bias. Quite the opposite. As you say, we all have biases. Stats merely help reinforce those biases. Look at the post title. I bet, for example, that you think stats are useful as an analysis tool, and when you use stats as an analysis tool, you find them useful... Link to comment Share on other sites More sharing options...
Cowshed Posted October 9, 2019 Report Share Posted October 9, 2019 3 minutes ago, Chivs said: Quite the opposite. As you say, we all have biases. Stats merely help reinforce those biases. Look at the post title. I bet, for example, that you think stats are useful as an analysis tool, and when you use stats as an analysis tool, you find them useful... I am not sure if I understand your post … Technology is now part of FA coaching badges. Coaching tutors do units on communication. Todays players are different and they use means of communication that did not exist decades years ago. Coaches and Managers are expected to know what they know - Be able to evidence base their knowledge. A player will have taken x amount of shots, ran x amount of km/s over a sample period or they have not. The data is a fact. The data is not biased. That evidence base can be used/is used to confront bias not reinforce it. We as humans communicate differently and have differing learning styles. And there stats are a supportive analysis tool. Yes. Link to comment Share on other sites More sharing options...
cidered abroad Posted October 9, 2019 Report Share Posted October 9, 2019 2 hours ago, Silvio Dante said: Quite often recently both on here and in the wider world, the term xG (expected goals) has been used to show how many goals a team were expected to score in a game, and therefore if they’re lucky/over performing Cards on the table, I find the xG thing is given more weight than it should be. You may create a clear cut chance, but the dependency on it finishing is due to the skill of the attacker and goalkeeper respectively. In short, there’s no point in Bas Savage having 8 xG because in reality I expect him to score none of those. On the other side, I expect Afobe to score most of his chances and Bentley to save a reasonable proportion of the opposition xG (some oversimplification here). In short, I think it’s a decent datapoint but currently overweighted. The article in the BEP today really highlights that. Based on the expected points from xG, you’d have the following: - We should be 23rd vs 6th - Charlton should be 20th vs 10th - Forest should be 17th vs 2nd - Reading should be 11th vs 22nd - Stoke should be 7th vs 24th - Brentford should be 3rd vs 17th So, roughly 25% of the way into the season (so a very decent sized data sample), 25% of teams have a deviation of +10 places based on xG. On a statistical front, that’s enough to make it unreliable so, question is - should we even look at it? I don't look at it because I do not have a clue where to find it! Link to comment Share on other sites More sharing options...
Guest Posted October 9, 2019 Report Share Posted October 9, 2019 So basically it is nonsense. I would just rather watch a game and have my own opinion. I think with statistics in general it is important to not just look at them and come to a conclusion. High pass success rate does not mean that player is a great passer of the ball for example. Link to comment Share on other sites More sharing options...
And Its Smith Posted October 9, 2019 Report Share Posted October 9, 2019 3 hours ago, Silvio Dante said: Quite often recently both on here and in the wider world, the term xG (expected goals) has been used to show how many goals a team were expected to score in a game, and therefore if they’re lucky/over performing Cards on the table, I find the xG thing is given more weight than it should be. You may create a clear cut chance, but the dependency on it finishing is due to the skill of the attacker and goalkeeper respectively. In short, there’s no point in Bas Savage having 8 xG because in reality I expect him to score none of those. On the other side, I expect Afobe to score most of his chances and Bentley to save a reasonable proportion of the opposition xG (some oversimplification here). In short, I think it’s a decent datapoint but currently overweighted. The article in the BEP today really highlights that. Based on the expected points from xG, you’d have the following: - We should be 23rd vs 6th - Charlton should be 20th vs 10th - Forest should be 17th vs 2nd - Reading should be 11th vs 22nd - Stoke should be 7th vs 24th - Brentford should be 3rd vs 17th So, roughly 25% of the way into the season (so a very decent sized data sample), 25% of teams have a deviation of +10 places based on xG. On a statistical front, that’s enough to make it unreliable so, question is - should we even look at it? Couldn’t agree more that it’s given more time than it should I started this topic about it. Basically I think a lot of people that use it want to have an opinion about a game they haven’t seen Link to comment Share on other sites More sharing options...
And Its Smith Posted October 9, 2019 Report Share Posted October 9, 2019 2 hours ago, Cowshed said: I am not sure if I understand your post … Technology is now part of FA coaching badges. Coaching tutors do units on communication. Todays players are different and they use means of communication that did not exist decades years ago. Coaches and Managers are expected to know what they know - Be able to evidence base their knowledge. A player will have taken x amount of shots, ran x amount of km/s over a sample period or they have not. The data is a fact. The data is not biased. That evidence base can be used/is used to confront bias not reinforce it. We as humans communicate differently and have differing learning styles. And there stats are a supportive analysis tool. Yes. A lot of stats are useful. xG isn’t one of them. Different companies report different xG for a start. People quote it in incorrect context as well. If team A beat Team B 2-1 but xG said it was Team A 1.1 v 2.3 Team B, people report that this means Team B deserved to win. Link to comment Share on other sites More sharing options...
Cowshed Posted October 9, 2019 Report Share Posted October 9, 2019 11 minutes ago, RedDave said: A lot of stats are useful. xG isn’t one of them. Different companies report different xG for a start. People quote it in incorrect context as well. If team A beat Team B 2-1 but xG said it was Team A 1.1 v 2.3 Team B, people report that this means Team B deserved to win. xG is useful. It can provide a historical basis to review a players performance be that over or underperforming. If a player is underperforming, why? Data can assist assessing if the decline is technical and tactical. Yes different companies and clubs use differing xG's. xG can omit penalties, free kicks, thirty yarders as irrelevant. It is not what is being measured. What is relevant is the patterns of play the team use x an individuals ability. A teams xG can be different based upon its crossing and to who. Teams again may consider this to be irrelevant as they go to feet rather than heads v other placing great emphasis on their expected xG from crosses and set plays. Link to comment Share on other sites More sharing options...
Davefevs Posted October 9, 2019 Report Share Posted October 9, 2019 6 hours ago, Silvio Dante said: Get that point, and indeed Leeds would be top (maybe they should play Eddie more...) My point here is that 25% is a very decent sample. The argument may be that it will level over the season (and that’s an estimate again not a definite) but isn’t it reasonable to state that it looks kind of flawed when you see the deviations it’s currently produced? Like I say, I see the benefit but I think it’s been overegged in its value thus far 25% of games might usually be a good sample size, it is in a context of games played, but xG is calculated from thousands and thousands of events. An MLS season has 10k shots. What people are incorrectly doing imho is taking say x number shots in a particular game (e.g. 10 per side per game) and calculating one result....and then repeating. It’s gonna throw up weird results. Basically a single game sample is irrelevant! xG is useful but it ain’t the be-all and end-all and should carry a big footnote!!! Link to comment Share on other sites More sharing options...
Guest Posted October 9, 2019 Report Share Posted October 9, 2019 It seems XG leaves out many different reasons why shots can be easier or harder. Positioning of players and position of ball when struck are just 2 of many things that should be taken into account to truly judge how easy or difficult a chance is. Unless I'm wrong, but that's what I have just read. Link to comment Share on other sites More sharing options...
spudski Posted October 9, 2019 Report Share Posted October 9, 2019 1 hour ago, JonDolman said: It seems XG leaves out many different reasons why shots can be easier or harder. Positioning of players and position of ball when struck are just 2 of many things that should be taken into account to truly judge how easy or difficult a chance is. Unless I'm wrong, but that's what I have just read. You might want to read through this for a fuller explanation mate.. https://www.theguardian.com/football/2017/mar/30/expected-goals-big-football-data-leicester-city-norwich Link to comment Share on other sites More sharing options...
Guest Posted October 9, 2019 Report Share Posted October 9, 2019 24 minutes ago, spudski said: You might want to read through this for a fuller explanation mate.. https://www.theguardian.com/football/2017/mar/30/expected-goals-big-football-data-leicester-city-norwich Ok fair enough. Still don't really see the benefit of it? Makes no real sense to me. How is Maupay header from 1 yard against Spurs with open goal got an Xg of 0.61? It's almost impossible to miss. Link to comment Share on other sites More sharing options...
Davefevs Posted October 9, 2019 Report Share Posted October 9, 2019 27 minutes ago, JonDolman said: Ok fair enough. Still don't really see the benefit of it? Makes no real sense to me. How is Maupay header from 1 yard against Spurs with open goal got an Xg of 0.61? It's almost impossible to miss. Because the sample of similar chances shows 0.39 (39 in a 100) were missed. Link to comment Share on other sites More sharing options...
cidered abroad Posted October 9, 2019 Report Share Posted October 9, 2019 49 minutes ago, JonDolman said: Ok fair enough. Still don't really see the benefit of it? Makes no real sense to me. How is Maupay header from 1 yard against Spurs with open goal got an Xg of 0.61? It's almost impossible to miss. Remember Rosenthal! Link to comment Share on other sites More sharing options...
And Its Smith Posted October 9, 2019 Report Share Posted October 9, 2019 1 hour ago, spudski said: You might want to read through this for a fuller explanation mate.. https://www.theguardian.com/football/2017/mar/30/expected-goals-big-football-data-leicester-city-norwich I can see what they want it to tell us but it really doesn’t work. xG is hugely flawed. Link to comment Share on other sites More sharing options...
Guest Posted October 10, 2019 Report Share Posted October 10, 2019 7 hours ago, Davefevs said: Because the sample of similar chances shows 0.39 (39 in a 100) were missed. There's surely no way that many were missed of that chance! Link to comment Share on other sites More sharing options...
reddevil Posted October 10, 2019 Report Share Posted October 10, 2019 There are a couple things about xG I don’t quite get and haven’t been able to google an answer. 1) how does xG recognise own goals, for example in the Hull away game earlier this season the Hunt created og and Afobe’s second came from broadly similar build up play but the own goal would have xG of zero. 2) how does xG factor in a pattern of play, for example striker has a shot, it is saved, follow up shot hits the bar then a third shot is tapped in from a yard. Do the models count xG for all three shots, the first or the last? For me the issue with xG is people using it for to prove arguements when it is mostly a subjective measure and is too flawed for that. It’s interesting but doesn’t necessarily mean that much 6 hours ago, RedDave said: I can see what they want it to tell us but it really doesn’t work. xG is hugely flawed. I agree with this, as another example the goal Diedhiou scored on Saturday probably was quite low for xG as it was a diving header. However it was a chance that suited our style of play because Diedhiou is a really good header of the ball, in effect we would expect Fam to our perform his xG on that type of chance. I can understand it working as a measure better in say basketball where a shot from distance is broadly similar whoever is taking it but in football there are too many variables to assess every chance evenly and there is no such thing as an average player Link to comment Share on other sites More sharing options...
Port Said Red Posted October 10, 2019 Report Share Posted October 10, 2019 I don’t know anything about xG, if anyone bothers reading any of my posts, they probably have worked out that I am more interested in the human factors than the tactical side of the game. It appears from the descriptions that people have put on here, that xG doesn’t seem to factor in the opposition much, or more importantly the “run of the ball”. I think we all want to see beautifully crafted goals, but the majority have an element in them (I am loathe to call it luck) that just can’t be factored into equations. I do have one question, are the players given these stats as a benchmark to work with? If so then it worries me that they may become obsessed with them. @Cowshed suggests that 30 yard blockbusters don’t get factored for example, will players start thinking that shooting from range is pointless as it won’t improve their xG? I have been watching some of the rugby, and although all the teams have improved dramatically from a tactical standpoint, it’s still the teams that are newest to the big stage, that will more often do something off the cuff and raise the excitement level. Is the next Messi or Ronaldo currently in an academy somewhere having his unique ability coached out of him? Link to comment Share on other sites More sharing options...
Cowshed Posted October 10, 2019 Report Share Posted October 10, 2019 5 hours ago, Port Said Red said: theirdon’t know anything about xG, if anyone bothers reading any of my posts, they probably have worked out that I am more interested in the human factors than the tactical side of the game. It appears from the descriptions that people have put on here, that xG doesn’t seem to factor in the opposition much, or more importantly the “run of the ball”. I think we all want to see beautifully crafted goals, but the majority have an element in them (I am loathe to call it luck) that just can’t be factored into equations. I do have one question, are the players given these stats as a benchmark to work with? If so then it worries me that they may become obsessed with them. @Cowshed suggests that 30 yard blockbusters don’t get factored for example, will players start thinking that shooting from range is pointless as it won’t improve their xG? I have been watching some of the rugby, and although all the teams have improved dramatically from a tactical standpoint, it’s still the teams that are newest to the big stage, that will more often do something off the cuff and raise the excitement level. Is the next Messi or Ronaldo currently in an academy somewhere having his unique ability coached out of him? Yes stats are used to provide benchmarks to work with, but this is a positive. It drives performance, target setting is healthy. The psychologist Bill Beswick put great emphasis on goal setting, information, standards and how these should be exciting positive challenges to meet and improve. Stats are just one means along side other forms of communication to drive improvement. Think of this as like being a runner improving by fractions of second but with footballers it will be working on differing elements of their game - marginal gains. In regards to 30 yarders you may have mistook what I meant. Maradonas goal v England is so untypical its validity to be measured is questionable. 30 yard volleyed top bins again its questionable. Measurements tend to be created v average, average teams, average players, average performances norms not outliers. This then identifies if players are performing above or below expected standards. If your pressing teams CM can do six miles plus get rid, or he improves, if your CF can't hit he target over a significant period from where he should, why? Its not witch craft, its very logical. Messi ability was coached into him via academies and La Masia. His unique ability is the result of thousands of hours, year upon year of intense, integrated training supported by coaches and science which most individuals cannot adapt to - That could be his unique. Link to comment Share on other sites More sharing options...
Numero Uno Posted October 10, 2019 Report Share Posted October 10, 2019 The fact that the league table that counts varies significantly from the xG league table only reinforces to me that these things should only be used to sift out relevant data on individual players. Over reliance on this stuff is pointless in my opinion, similarly to Prozone and other “tools” that are used. I still like to think that coaches still put a lot of emphasis on what they actually SEE not what a piece of software is telling them. Football is not and never has been an exact science. Too much of the game is subjective and down to human error/luck at times not to mention mental factors that simply cannot be measured or predicted. To me, relying on xG to predict where you SHOULD be in the league table or what the result of a game SHOULD have been is a huge misinterpretation. Link to comment Share on other sites More sharing options...
054123 Posted October 10, 2019 Report Share Posted October 10, 2019 I listen to the Not The Top 20 Podcast and they live by XG to the point where it clouds their judgment IMO. Apparently our low XG means we’re about to slide away. Link to comment Share on other sites More sharing options...
Recommended Posts
Archived
This topic is now archived and is closed to further replies.