Attention: Please take a moment to consider our terms and conditions before posting.

Mail On-line Championship Preview - TEAM-BY-TEAM GUIDE

124»

Comments

  • When people keep going on about squad value you're not going by Transfermarkt are you?
  • seth plum said:

    Could seriously red be right, but for the wrong reasons?

    I could - correlation does not necessarily mean causation. Any football fan knows that adding better players improves club chances but guarantees nothing.
    Transfermarkt is not about transfer fees - it is about player valuation based on an unknown algorithm. Given that many arrivals this summer are on free transfers and are unknown to these shores I chose to look for their rough value and compare it to players who we know like Ajdarevic etc playing in both Belgium and SE7.
    I ran the squad valuations against the bookies odds for relegation a few days back and there is an even closer relationship!
    We all know bookies are wrong sometimes but they'd be out of business if they were wrong all the time.
    This is not all or nothing analysis...just a bit of fun to suggest that overall squad valuations give an indication that we will move up to mid table.
    I want to believe that our new look front six will play like a £9m(!) set of players but let's watch some football first. What I will say is that it's highly unlikely they will play like a bottom three club.
  • I'm going through transfer markt and it's ridiculous. Church valued at 660k yet a player I'd love to see at Charlton (Fabian Castillo) only 220k?

    If we could do transfer business according to this site we'd be laughing. I actually think this site is on the same level as Fifa for rating players.
  • Been somewhat reluctant to comment on the Transfermarkt debate, but for better or for worse I've succumbed to temptation.

    We all know that, all other things being equal, the better players a team has, relative to other teams in the league, the higher it's likely position in the table. Of course, other things will matter too, including the blend and balance of the squad, how well it is coached, motivated and managed etc. and, to a greater extent than we generally acknowledge, random chance.

    The use of Transfermarkt values as a proxy for how good players are is, potentially, an elegant way of calibrating how good each squad in the league is and then, in turn, predicting final league table positioning.

    I say "potentially" because the question is whether the methodology described works. I would suggest that a correlation (coefficient) between squad values and final league table position of somewhere in the 0.6 - 0.7 range, which I think is what @seriously_red‌ is suggesting, is highly significant. If you had a model for predicting stock prices or exchange rates, for example, with anywhere near that degree of explanatory power you'd not only be very happy you'd be very wealthy too.

    Of course, @seriously_red‌ 's model is not perfect, if it were the correlation coefficient would be 1.0. It's not 1.0 because of the factors I noted above and, of course, because there are errors in the underlying Transfermarkt data, i.e. a lot of values will be wrong, and this will create additional noise in the relationship.

    There is a very important health warning that needs to go with statistical analysis of this type though. The errors or unexplained effects I've talked about don't matter if we are dealing with a very large data set; those errors cancel each other out. If we were forecasting the league table position of twenty teams in twenty leagues in each of five years, say, then @seriously_red‌ 's model would give us a real edge, at least on the basis of the evidence presented.

    However, over small data sets the model is likely to be much less reliable, as idiosyncratic factors become more important. Put another way, there is less opportunity for errors to cancel out. Of course, this is especially true if we are focusing on one team in one season. In Charlton's case, errors in the Transfermarkt valuation of players might prove to be very significant as may be the blend of the squad assembled and Bob Peeters ability to get the team to perform.

    My own assessment, for what little it's worth, is that this summer's activity in the transfer market has been very encouraging and @seriously_red‌ 's analysis is one way of calibrating the potential improvement which has taken place. However, at this stage, I remain no more than cautiously optimistic about the prospects for next season. There are many unknowns, perhaps more so than for most other teams in the league, and it's simply too early to judge how we might fare even though it's probably reasonable to hope for a better outcome than last season. We'll know more 5 - 10 games in. Fingers crossed.

    PS My comment about predictability is deliberately oversimplified to ease explanation. To be of value any "model" needs to do better than the current market odds.
  • Been somewhat reluctant to comment on the Transfermarkt debate, but for better or for worse I've succumbed to temptation.

    We all know that, all other things being equal, the better players a team has, relative to other teams in the league, the higher it's likely position in the table. Of course, other things will matter too, including the blend and balance of the squad, how well it is coached, motivated and managed etc. and, to a greater extent than we generally acknowledge, random chance.

    The use of Transfermarkt values as a proxy for how good players are is, potentially, an elegant way of calibrating how good each squad in the league is and then, in turn, predicting final league table positioning.

    I say "potentially" because the question is whether the methodology described works. I would suggest that a correlation (coefficient) between squad values and final league table position of somewhere in the 0.6 - 0.7 range, which I think is what @seriously_red‌ is suggesting, is highly significant. If you had a model for predicting stock prices or exchange rates, for example, with anywhere near that degree of explanatory power you'd not only be very happy you'd be very wealthy too.

    Of course, @seriously_red‌ 's model is not perfect, if it were the correlation coefficient would be 1.0. It's not 1.0 because of the factors I noted above and, of course, because there are errors in the underlying Transfermarkt data, i.e. a lot of values will be wrong, and this will create additional noise in the relationship.

    There is a very important health warning that needs to go with statistical analysis of this type though. The errors or unexplained effects I've talked about don't matter if we are dealing with a very large data set; those errors cancel each other out. If we were forecasting the league table position of twenty teams in twenty leagues in each of five years, say, then @seriously_red‌ 's model would give us a real edge, at least on the basis of the evidence presented.

    However, over small data sets the model is likely to be much less reliable, as idiosyncratic factors become more important. Put another way, there is less opportunity for errors to cancel out. Of course, this is especially true if we are focusing on one team in one season. In Charlton's case, errors in the Transfermarkt valuation of players might prove to be very significant as may be the blend of the squad assembled and Bob Peeters ability to get the team to perform.

    My own assessment, for what little it's worth, is that this summer's activity in the transfer market has been very encouraging and @seriously_red‌ 's analysis is one way of calibrating the potential improvement which has taken place. However, at this stage, I remain no more than cautiously optimistic about the prospects for next season. There are many unknowns, perhaps more so than for most other teams in the league, and it's simply too early to judge how we might fare even though it's probably reasonable to hope for a better outcome than last season. We'll know more 5 - 10 games in. Fingers crossed.

    PS My comment about predictability is deliberately oversimplified to ease explanation. To be of value any "model" needs to do better than the current market odds.

    this more or less sums up my thinking on this debate and a very good introduction to correlation and simple statistical analysis. seriously_red's use of transfermarkt as a proxy for real life is not an unreasonable one, but for this data set is prone to potential error. we can improve the model, but we would need to identify other indicators which also have an affect on performance (such as management style, refereeing, weather, ....and many other potential indicators). there is an interesting (yes Henry - stats are really interesting and tell us lots of things, watch 'The Joy of Stats' as a starter) piece of research to be done to see which indicators make a difference when taken together. it might come up with some unexpected and interesting results. then again, it might not. that is the fun of research.

    anyway seriously_red's rough and ready model does hold water if treated with some suspicion. but that is the point of stats. they are used to predict trends and identify similar cases based upon historical data. any specific case might break that trend because there will always be outliers. that doesn't matter, it just provides data to improve the model.

  • Think I'll simply wait and see how our new players perform on the pitch in a Charlton shirt rather than rely on any old lies, damned lies and statistics...

  • FYI the 60% quote was my error .

    Finally an admission that you got it badly wrong.

    This discussion is over

    : - )


  • FYI the 60% quote was my error .

    Finally an admission that you got it badly wrong.

    This discussion is over

    : - )

    When you say it's over, do you mean that you have read and absorbed the contributions of @Mundell Fleming‌ and @henrythecat‌ who both clearly state that a 74.9% correlation is significant - in fact Mundell confidently states that this level of correlation applied to forex or equity trading models would make one very rich.
    The point which you miss is that it looks likely that we will continue to secure the services of a number of unknown players via this new scouting network. If they are rated at €2m then it is likely they will make a difference and we will climb the table - what's not to like?

    Far from being over, the debate might continue particularly if the league table next season shows the same relationship to notional squad values. For it can be used to question the rise and fall of other clubs, particularly as FFP kicks in. And challenge the bookies and nonsense in the daily mail.

    So feel free to engage, or don't engage but please don't try to control and insult.
  • Sponsored links:



  • FYI the 60% quote was my error .

    Finally an admission that you got it badly wrong.

    This discussion is over

    : - )

    When you say it's over, do you mean that you have read and absorbed the contributions of @Mundell Fleming‌ and @henrythecat‌ who both clearly state that a 74.9% correlation is significant - in fact Mundell confidently states that this level of correlation applied to forex or equity trading models would make one very rich.
    The point which you miss is that it looks likely that we will continue to secure the services of a number of unknown players via this new scouting network. If they are rated at €2m then it is likely they will make a difference and we will climb the table - what's not to like?

    Far from being over, the debate might continue particularly if the league table next season shows the same relationship to notional squad values. For it can be used to question the rise and fall of other clubs, particularly as FFP kicks in. And challenge the bookies and nonsense in the daily mail.

    So feel free to engage, or don't engage but please don't try to control and insult.

    Do you ever read what you write or what others say?

    I said 60% wasn't a significant correlation. It isn't. It is "less than moderate" by your own admission. You couldn't even calculate the correlation correctly by your own admission.

    Mundell and Henrythecat very politely ripped to shreds your suggestion that transfermarket was reliable or that your analysis, based as it is on one team over one season, was valid but you seem to have only read the bits that you think agree with and have ignored the rest.

    You are so quick to dismiss others points as "The rest is noise" so you don't have to address them but you don't seem able to read what people actually say. You seem to find that and other insults quite acceptable when it is you making them. I remember your personal attacks on me about the museum and the attempt to attack me when I called the Trusts finance story (you know, the one you let the Club rewrite in "content and style") disappointing.

    Mundell and Henrythecat are actually agreeing with me that your reliance on Transfermarket isn't valid.

    Bye, bye.
  • @Mundell Fleming‌ very clearly states a correlation coefficient of 60-75% is highly significant"
    Or henrythecat states that using transfermarkt as a proxy "is not an unreasonable one"
    The actual number was .749 but if it was a tad less it makes no difference to the argument.
    I fail to see why you keep citing 60% as that was an error which has been acknowledged and has no relevance. Is it because you have nothing else?
    I know not how many seasons transfermarkt has been going but it has built data over time and my particular piece of work is about ALL 24 teams in the Championship... 650 players. Not just one club as you assert. Perhaps I will rerun after 10 games next season to see if the correlation is stronger or weaker but would put a tenner on it being higher than 50%. Are you up for that?
    The graph I uploaded clearly illustrated the trend with some outliers. The thing about outliers is that they do NOT disprove the overall correlation.
    So let us be very clear: whether one calls 75% correlation moderate or significant it is still a clear factor and might depend upon the application. And it certainly does not mean insignificant.
  • Blimey, I'm 78.9% certain I'm ready for the season to start
  • Still prefer watching this than reading about stats

    http://www.youtube.com/watch?v=LlTqh02qV_o

  • FYI the 60% quote was my error .

    Finally an admission that you got it badly wrong.

    This discussion is over

    : - )

    When you say it's over, do you mean that you have read and absorbed the contributions of @Mundell Fleming‌ and @henrythecat‌ who both clearly state that a 74.9% correlation is significant - in fact Mundell confidently states that this level of correlation applied to forex or equity trading models would make one very rich.
    The point which you miss is that it looks likely that we will continue to secure the services of a number of unknown players via this new scouting network. If they are rated at €2m then it is likely they will make a difference and we will climb the table - what's not to like?

    Far from being over, the debate might continue particularly if the league table next season shows the same relationship to notional squad values. For it can be used to question the rise and fall of other clubs, particularly as FFP kicks in. And challenge the bookies and nonsense in the daily mail.

    So feel free to engage, or don't engage but please don't try to control and insult.

    Do you ever read what you write or what others say?

    I said 60% wasn't a significant correlation. It isn't. It is "less than moderate" by your own admission. You couldn't even calculate the correlation correctly by your own admission.

    Mundell and Henrythecat very politely ripped to shreds your suggestion that transfermarket was reliable or that your analysis, based as it is on one team over one season, was valid but you seem to have only read the bits that you think agree with and have ignored the rest.

    You are so quick to dismiss others points as "The rest is noise" so you don't have to address them but you don't seem able to read what people actually say. You seem to find that and other insults quite acceptable when it is you making them. I remember your personal attacks on me about the museum and the attempt to attack me when I called the Trusts finance story (you know, the one you let the Club rewrite in "content and style") disappointing.

    Mundell and Henrythecat are actually agreeing with me that your reliance on Transfermarket isn't valid.

    Bye, bye.
    actually Henry, I wasn't agreeing with you. From what I have read I would say your understanding of stats and how they are used has a number of weaknesses, particularly your dismissal that stats are pointless. They are not and a lot of the world is driven on data analysis. But I will leave that to one side.

    What I said (and using my best academic and research based voice) is that transfermarkts is not an unreasonable data set to use in the general, but could be prone to error due to the small size of data set. a .6 -.7 correlation could be significant, but is only one factor in determining accuracy. lots of double negatives, and get out clauses to cover all eventualities.

    so I think seriously_red's assertion (and he can correct me if wrong) is that the daily mail's article is wrong because, as a rough measure, transfermarkt's data suggests that we have improved adn are above a number of other sides in the league. that seems a fair point to me. it doesn't suggest that it will happen in reality but as the daily mail's analysis is just a guess, at least his is built upon data (albeit limited)

    where you are right Henry is that if you want to predict more accurately, you would need look at a wider set of indicators. after that there is some interesting work to find out how the indicators react with each other. But to do that would be a significant effort and life is too short. So I shall just wait and see how we get on when we play football matches.

  • FYI the 60% quote was my error .

    Finally an admission that you got it badly wrong.

    This discussion is over

    : - )

    When you say it's over, do you mean that you have read and absorbed the contributions of @Mundell Fleming‌ and @henrythecat‌ who both clearly state that a 74.9% correlation is significant - in fact Mundell confidently states that this level of correlation applied to forex or equity trading models would make one very rich.
    The point which you miss is that it looks likely that we will continue to secure the services of a number of unknown players via this new scouting network. If they are rated at €2m then it is likely they will make a difference and we will climb the table - what's not to like?

    Far from being over, the debate might continue particularly if the league table next season shows the same relationship to notional squad values. For it can be used to question the rise and fall of other clubs, particularly as FFP kicks in. And challenge the bookies and nonsense in the daily mail.

    So feel free to engage, or don't engage but please don't try to control and insult.

    Do you ever read what you write or what others say?

    I said 60% wasn't a significant correlation. It isn't. It is "less than moderate" by your own admission. You couldn't even calculate the correlation correctly by your own admission.

    Mundell and Henrythecat very politely ripped to shreds your suggestion that transfermarket was reliable or that your analysis, based as it is on one team over one season, was valid but you seem to have only read the bits that you think agree with and have ignored the rest.

    You are so quick to dismiss others points as "The rest is noise" so you don't have to address them but you don't seem able to read what people actually say. You seem to find that and other insults quite acceptable when it is you making them. I remember your personal attacks on me about the museum and the attempt to attack me when I called the Trusts finance story (you know, the one you let the Club rewrite in "content and style") disappointing.

    Mundell and Henrythecat are actually agreeing with me that your reliance on Transfermarket isn't valid.

    Bye, bye.
    actually Henry, I wasn't agreeing with you. From what I have read I would say your understanding of stats and how they are used has a number of weaknesses, particularly your dismissal that stats are pointless. They are not and a lot of the world is driven on data analysis. But I will leave that to one side.

    What I said (and using my best academic and research based voice) is that transfermarkts is not an unreasonable data set to use in the general, but could be prone to error due to the small size of data set. a .6 -.7 correlation could be significant, but is only one factor in determining accuracy. lots of double negatives, and get out clauses to cover all eventualities.

    so I think seriously_red's assertion (and he can correct me if wrong) is that the daily mail's article is wrong because, as a rough measure, transfermarkt's data suggests that we have improved adn are above a number of other sides in the league. that seems a fair point to me. it doesn't suggest that it will happen in reality but as the daily mail's analysis is just a guess, at least his is built upon data (albeit limited)

    where you are right Henry is that if you want to predict more accurately, you would need look at a wider set of indicators. after that there is some interesting work to find out how the indicators react with each other. But to do that would be a significant effort and life is too short. So I shall just wait and see how we get on when we play football matches.
    What I said was "Mundell and Henrythecat are actually agreeing with me that your reliance on Transfermarket isn't valid.

    What you said was "where you are right Henry is that if you want to predict more accurately, you would need look at a wider set of indicators. "

    So we agree, SR's reliance on one discredited source of info is not valid.

    I never said Stats weren't interesting just football's joys is it glorious unpredictability and that relying solely on stats was a bit sad Which you agreed with when you said

    "after that there is some interesting work to find out how the indicators react with each other. But to do that would be a significant effort and life is too short. So I shall just wait and see how we get on when we play football matches."

    Which has been my point from the start. It is the football matches that will prove the value of the players not made up stats.

    Stats are fine. I have an A Level in statistics (very old) so I do understand them, I just don't rely on them and especially not people who suddenly shift the stats from 60% to 74% when it suits them without an explanation beyond "I got it wrong". Hardly a reliable interrogation then, was it : - )

  • FYI the 60% quote was my error .

    Finally an admission that you got it badly wrong.

    This discussion is over

    : - )

    When you say it's over, do you mean that you have read and absorbed the contributions of @Mundell Fleming‌ and @henrythecat‌ who both clearly state that a 74.9% correlation is significant - in fact Mundell confidently states that this level of correlation applied to forex or equity trading models would make one very rich.
    The point which you miss is that it looks likely that we will continue to secure the services of a number of unknown players via this new scouting network. If they are rated at €2m then it is likely they will make a difference and we will climb the table - what's not to like?

    Far from being over, the debate might continue particularly if the league table next season shows the same relationship to notional squad values. For it can be used to question the rise and fall of other clubs, particularly as FFP kicks in. And challenge the bookies and nonsense in the daily mail.

    So feel free to engage, or don't engage but please don't try to control and insult.

    Do you ever read what you write or what others say?

    I said 60% wasn't a significant correlation. It isn't. It is "less than moderate" by your own admission. You couldn't even calculate the correlation correctly by your own admission.

    Mundell and Henrythecat very politely ripped to shreds your suggestion that transfermarket was reliable or that your analysis, based as it is on one team over one season, was valid but you seem to have only read the bits that you think agree with and have ignored the rest.

    You are so quick to dismiss others points as "The rest is noise" so you don't have to address them but you don't seem able to read what people actually say. You seem to find that and other insults quite acceptable when it is you making them. I remember your personal attacks on me about the museum and the attempt to attack me when I called the Trusts finance story (you know, the one you let the Club rewrite in "content and style") disappointing.

    Mundell and Henrythecat are actually agreeing with me that your reliance on Transfermarket isn't valid.

    Bye, bye.
    actually Henry, I wasn't agreeing with you. From what I have read I would say your understanding of stats and how they are used has a number of weaknesses, particularly your dismissal that stats are pointless. They are not and a lot of the world is driven on data analysis. But I will leave that to one side.

    What I said (and using my best academic and research based voice) is that transfermarkts is not an unreasonable data set to use in the general, but could be prone to error due to the small size of data set. a .6 -.7 correlation could be significant, but is only one factor in determining accuracy. lots of double negatives, and get out clauses to cover all eventualities.

    so I think seriously_red's assertion (and he can correct me if wrong) is that the daily mail's article is wrong because, as a rough measure, transfermarkt's data suggests that we have improved adn are above a number of other sides in the league. that seems a fair point to me. it doesn't suggest that it will happen in reality but as the daily mail's analysis is just a guess, at least his is built upon data (albeit limited)

    where you are right Henry is that if you want to predict more accurately, you would need look at a wider set of indicators. after that there is some interesting work to find out how the indicators react with each other. But to do that would be a significant effort and life is too short. So I shall just wait and see how we get on when we play football matches.
    What I said was "Mundell and Henrythecat are actually agreeing with me that your reliance on Transfermarket isn't valid.

    What you said was "where you are right Henry is that if you want to predict more accurately, you would need look at a wider set of indicators. "

    So we agree, SR's reliance on one discredited source of info is not valid.

    I never said Stats weren't interesting just football's joys is it glorious unpredictability and that relying solely on stats was a bit sad Which you agreed with when you said

    "after that there is some interesting work to find out how the indicators react with each other. But to do that would be a significant effort and life is too short. So I shall just wait and see how we get on when we play football matches."

    Which has been my point from the start. It is the football matches that will prove the value of the players not made up stats.

    Stats are fine. I have an A Level in statistics (very old) so I do understand them, I just don't rely on them and especially not people who suddenly shift the stats from 60% to 74% when it suits them without an explanation beyond "I got it wrong". Hardly a reliable interrogation then, was it : - )

    i didn't say using transfermarkt isn't valid. i said it is not an unreasonable, if somewhat weak, model. that is very different from not valid. it is not perfect by a long way but better than the made up guff from the daily mail. i believe that was seriously_red's point.

    you said: "I feel sorry for those that have to use statistics and algorithms to get their kicks, they really are missing out on the fun" which does not mention of football there, just about stats.

    you seem to have some kind of issue with him, and he with you which has clouded this discussion. he said - here is some evidence that the daily mail is wrong. you said - it is only real matches that changes the value of players. both of those statements are true because they relate to different issues. on that, i'll leave this discussion because, as i have said, life is too short.
  • Sorry but you are quoting selectively there. I said

    "It is the uncertainty and unpredictable nature of sport and football in particular that make it so exciting, so frustrating, so glorious.
    I feel sorry for those that have to use statistics and algorithms to get their kicks, they really are missing out on the fun."

    By missing the first line you totally change the context.

    No one had defended the Daily Mail piece which is poor by any standard but saying that a "weak model" (your words) is better does not make that model any better. We don't need a model or a correlation to say the mail article is poorly written although only time will tell if it is accurate.

    But like you I will leave it there as i suspect that we agree anyway other than I seem to have offended you, as someone who like stats, which was not my intention. Stats are great in context, I enjoyed Soccernomics which is full of them but they are in this context and in my opinion a limited tool for predicting the success of Charlton next season.
  • edited July 2014
    @Henry Irving‌ the stat was .749 - it didn't move and certainly not a "sudden shift"... I quoted it as 70%+ and then the precise number.

    Your insistence that I changed it without explanation is hardly a reasonable nor rational interpretation of my very clear statement that I had cited (in error) another correlation on July 1st - that being the relationship between club expenditure and league position.

    Of course one could look at all factors but that was never the intention. As @henrythecat‌ states I was both challenging the Daily Mail rubbishing of our chances and your assertion that transfermarkt is as unreliable as the Mail.

    So there's no conspiracy to deceive with misleading or made up numbers - just a way of trying to understand the direction of travel of CAFC.

    More years will help prove / analyse transfermarkt (or another better database) but by the time a few seasons data has been gathered we will already know what the team looks like.

    Life is short so enjoy the weekend and the season. Let's hope these new signings live up to their transfermarkt valuations ;)
  • Lies Damn Lies and Statistics
  • Sponsored links:


  • Sorry but you are quoting selectively there.

    This, kids, is something called irony. :-)
  • This has got to be one of the most boring threads ever, just cannot be bothered to read it all, just two people trying to out do each other
  • "Chances: Narrowly avoided the drop last season and then promptly binned the manager who saved them, while many of their better players have also departed. Once held up as the model of how to run a football club, Charlton are now a disaster waiting to happen."

    Well, well, well. Let us find the idiot responsible for writing this piece. Although it is early days, I am confident we are not 'a disaster waiting to happen'

    1st class plonker - who are you?
  • cabbles said:

    "Chances: Narrowly avoided the drop last season and then promptly binned the manager who saved them, while many of their better players have also departed. Once held up as the model of how to run a football club, Charlton are now a disaster waiting to happen."

    Well, well, well. Let us find the idiot responsible for writing this piece. Although it is early days, I am confident we are not 'a disaster waiting to happen'

    1st class plonker - who / where are you?

    Fixed it for you... Block 'A' is coming for him!!
  • cabbles said:

    "Chances: Narrowly avoided the drop last season and then promptly binned the manager who saved them, while many of their better players have also departed. Once held up as the model of how to run a football club, Charlton are now a disaster waiting to happen."

    Well, well, well. Let us find the idiot responsible for writing this piece. Although it is early days, I am confident we are not 'a disaster waiting to happen'

    1st class plonker - who / where are you?

    Fixed it for you... Block 'A' is coming for him!!
    Be great if we could find out his name. I fancy sending his editor an email telling him what a prat he is
  • the footballing sage that is Joey Barton said we'd be relegated on his twitter feed the week b4 the first game

    I sent him a tweet last night asking if he still thought we'd be going down but, strangely, he hasn't responded...
  • Think maybe we should be a little less sensitive about what splash heads like Barton and ill informed journalists think. Let the results do the talking.
Sign In or Register to comment.

Roland Out Forever!