Jump to content
Sports Interactive Community

Archived

This topic is now archived and is closed to further replies.

VibeTribe

Potential

Recommended Posts

hahahaha so true, there are so many of these threads now.

I'm still in the boat that until someone proves they have a better option ( and in the countless threads no one yet has shown they have) then the current system has to be kept.

Looky

Not perfect, but hey, without SI source code, it's kind of difficult to do so...

Share this post


Link to post
Share on other sites
Looky

Not perfect, but hey, without SI source code, it's kind of difficult to do so...

yeah like i say im still not convinced, you clearly have put a lot of thought and effort into what you have come up with, but i still see it as removing one restraint just to replace it with another and without a great deal of benefit considering the upheaval it would create in the game.

Share this post


Link to post
Share on other sites
Looky

Not perfect, but hey, without SI source code, it's kind of difficult to do so...

I remember this model. It's terrible, down right terrible and I'm not afraid to say that. If you had the SI source code by the way I doubt you'd even know what it was. Reasons, look back through that thread. Essentially, it would be unstable globally and would fail to replicate most circumstances in real life, like the current game does extremely well.

Share this post


Link to post
Share on other sites
yeah like i say im still not convinced, you clearly have put a lot of thought and effort into what you have come up with, but i still see it as removing one restraint just to replace it with another and without a great deal of benefit considering the upheaval it would create in the game.

I don't want you to judge my idea by looking at the amount of work. I'm aware this is a high-risk, high-workload, minimal benefit change, but I want you to judge the change on its own merits. Let SI decide where it goes on the TODO list.

Either way, there are clearly points between the current state and my model - one of those is RBKalle's idea of having PA be renewed every season, rather like being a researcher in the game - that are lower-risk, lower-workload but similar benefits.

As for the restraint bit, every model I propose will be imperfect - because a model by definition cannot be fully correct (Gödel's incompleteness theorem - the only perfect model is life itself). What matters is that it relaxes one assumption and replaces it with a weaker assumption, with controls to ensure that we get something out of it - i.e. it is better - i.e. it is possible to remove PA.

Share this post


Link to post
Share on other sites
I remember this model. It's terrible, down right terrible and I'm not afraid to say that. If you had the SI source code by the way I doubt you'd even know what it was. Reasons, look back through that thread. Essentially, it would be unstable globally and would fail to replicate most circumstances in real life, like the current game does extremely well.

Why would it be unstable? We use the normal distribution all the time. If that is unstable, you had better stop taking medicine, since it's not safe enough for you - hypothesis testing is based on the normal distribution and its related distributions.

And what would it fail to replicate in real-life?

Share this post


Link to post
Share on other sites
I don't want you to judge my idea by looking at the amount of work. I'm aware this is a high-risk, high-workload, minimal benefit change, but I want you to judge the change on its own merits. Let SI decide where it goes on the TODO list.

Either way, there are clearly points between the current state and my model - one of those is RBKalle's idea of having PA be renewed every season, rather like being a researcher in the game - that are lower-risk, lower-workload but similar benefits.

As for the restraint bit, every model I propose will be imperfect - because a model by definition cannot be fully correct (Gödel's incompleteness theorem - the only perfect model is life itself). What matters is that it relaxes one assumption and replaces it with a weaker assumption, with controls to ensure that we get something out of it - i.e. it is better - i.e. it is possible to remove PA.

If anything like what you've mentioned ends up on their to do list I'll happily never buy another product from the company. Why? Because that's the moment Football Manager would be flying off the rails.

In this case the simplest model is probably the best as having a "changing potential" is both silly and redundant if one understands the CA/PA system as it currently stands. Rather than explaining why this is I'd ask you to name for me one well known player who couldn't be successfully modelled by the CA/PA system and why he couldn't be?

Share this post


Link to post
Share on other sites
If anything like what you've mentioned ends up on their to do list I'll happily never buy another product from the company. Why? Because that's the moment Football Manager would be flying off the rails.

So "improved development for late-bloomers" and "taking into account possible uncertainties with researcher input values" shouldn't be on the to-do list?

In this case the simplest model is probably the best as having a "changing potential" is both silly and redundant if one understands the CA/PA system as it currently stands. Rather than explaining why this is I'd ask you to name for me one well known player who couldn't be successfully modelled by the CA/PA system and why he couldn't be?

Pedro, FM07. His PA was low in FM07, yet he turned out brilliant in real-life. If Pedro had performed exactly as he did in-game as he did in reality, he would not have turned out in the same way he did in real-life as in-game. Therefore the game is inaccurate.

Now I can't give you an example in FM11, as it implies I can see into the future. But what I can say is that a lot of youngsters right now in FM14 will have different PAs than now in FM11 - and some will be higher. And in those cases, PA was wrong in FM11.

Is that the PA value that is wrong, or the model? Both. The PA value is wrong, full-stop - nothing can be done about that. But I'm not going to pretend nothing can be done about the model - perhaps making up for the fact that the PA value is wrong.

Share this post


Link to post
Share on other sites
So "improved development for late-bloomers" and "taking into account possible uncertainties with researcher input values" shouldn't be on the to-do list?

Pedro, FM07. His PA was low in FM07, yet he turned out brilliant in real-life. If Pedro had performed exactly as he did in-game as he did in reality, he would not have turned out in the same way he did in real-life as in-game. Therefore the game is inaccurate.

Now I can't give you an example in FM11, as it implies I can see into the future. But what I can say is that a lot of youngsters right now in FM14 will have different PAs than now in FM11 - and some will be higher. And in those cases, PA was wrong in FM11.

Is that the PA value that is wrong, or the model? Both. The PA value is wrong, full-stop - nothing can be done about that. But I'm not going to pretend nothing can be done about the model - perhaps making up for the fact that the PA value is wrong.

There are plenty of late-blooming players in the game. The CA/PA system is actually somewhat of a specialist at this in recent years. It works especially well with regens as well.

Pedro? No, they just got his PA wrong, that's a data editor problem, not a problem with the system.

See, you're searching for the ability of future sight, but there is the problem. The issue is more to do with the projected "apparent PA" that we give to the real players in the data base. But if you take the system as what it is, and that is as a simulation where there may be errors in our judgements of the players, then it becomes hard to fault it. The problem here in the difference between our opinions is due to how we are seeing the situation. You think that the database producers aren't wrong about their potential, but instead their potential has just increased. Ignoring the definition of potential you can form this argument, but it leads to a highly complicated and potentially unstable model of ability which would be extremely hard to work into the game. On top of this it wouldn't change a damn thing as the CA/PA system already models players development very accurately.

Explain though, how is "PA value wrong"?

Share this post


Link to post
Share on other sites
I don't want you to judge my idea by looking at the amount of work. I'm aware this is a high-risk, high-workload, minimal benefit change, but I want you to judge the change on its own merits. Let SI decide where it goes on the TODO list.

Either way, there are clearly points between the current state and my model - one of those is RBKalle's idea of having PA be renewed every season, rather like being a researcher in the game - that are lower-risk, lower-workload but similar benefits.

As for the restraint bit, every model I propose will be imperfect - because a model by definition cannot be fully correct (Gödel's incompleteness theorem - the only perfect model is life itself). What matters is that it relaxes one assumption and replaces it with a weaker assumption, with controls to ensure that we get something out of it - i.e. it is better - i.e. it is possible to remove PA.

Im not judging it by how fancy it looks, but by the content, i also have no intention of telling SI where to put this on the list, or bin as i would. It still comes down to removing a, suppose to be hidden, restraint and replacing it with another. The only real problem with the PA system is when someone looks at it as a number, without knowing that number it is impossible to find situations you do not agree with, to which you will reply something about not knowing there is a problem by not looking under the hood and we can all move onto Francis Jeffers or Pedro and debate their careers over and over.

I still dont agree that a system without PA is better and i still think its a very arrogant thing for you to say when you have no real idea if your system would work or not with the FM code, as you have stated yourself.

Share this post


Link to post
Share on other sites
There are plenty of late-blooming players in the game. The CA/PA system is actually somewhat of a specialist at this in recent years. It works especially well with regens as well.

Not denying that. However, in order to late-bloom, your PA needs to be high enough. This is despite the fact that PA is "expected" and late-blooming is "unexpected". You cannot use an expected value for unexpected consequences?

Pedro? No, they just got his PA wrong, that's a data editor problem, not a problem with the system.

Is it? I'd be damned if there were systems out there that didn't assume their input data was correct, and did some statistical analysis to take that into account...

Take election polls. They take an average of many surveys, usually taking into account things like blinding and blocking, and produce a result that has a certain margin of error. They don't just take the average and wave it around, saying "Ah-ha! McCain is going to win this one!" They take into account the fact that there is bias and underlying issues with the survey (a survey is never perfect - but some are better than others). So when someone wants to simulate an election, they input various means but they also input the variances too - to reflect the fact that the input data isn't perfect - and never will be. If it were, you'd be rich.

The thing is, we know PA is wrong for many youngsters - we know this because researchers change them every research phase. So why should it be locked down, within the game, at any one time. If it changes in reality, surely the most realistic thing to do is for it to change in-game too?

See, you're searching for the ability of future sight, but there is the problem. The issue is more to do with the projected "apparent PA" that we give to the real players in the data base. But if you take the system as what it is, and that is as a simulation where there may be errors in our judgements of the players, then it becomes hard to fault it.

Not really - there are errors in the input and these need to combine with the errors in judgement.

Your argument can also be applied to every single flaw with the game ("It's alright, since we make errors in-game!") - it is no excuse.

The problem here in the difference between our opinions is due to how we are seeing the situation. You think that the database producers aren't wrong about their potential, but instead their potential has just increased. Ignoring the definition of potential you can form this argument, but it leads to a highly complicated and potentially unstable model of ability which would be extremely hard to work into the game.

See, this is the part I don't understand. You are treating a no-PA model like an F1 car with a speed limiter of 30 mph and then removing the speed limiter, watching the carnage ensue. The model I've showed, however, has the chance of seeing one player who has PA 140 to become PA 200 once every 47*365 trials (once every 47 years) - i.e. if you restarted the game 47*365 times, you'd see it happen once. Once! And I personally think the randomness is too much anyway - but it's the only way I could get pretty graphs. Reducing the standard deviation has an exponential decreasing effect on the chances of 140 => 200.

The reality is, a no-PA model that is well-designed will be rather predictable and more akin to taking the training wheels off your toddler's bike when he/she has outgrown them. Yes, they are bound to run over some elderly neighbour or surgically remove someone's wing mirror with an unfortunate burst of speed every now and then - but it's still quite predictable.

The issue is that because we don't know what the limit is, we shouldn't, well, posit a limit.

On top of this it wouldn't change a damn thing as the CA/PA system already models players development very accurately.

You've just said it would become highly unstable - is it really going to change anything?

It models it accurately for a lot of players - for some, however, it doesn't. This will help this.

No model can model it 100% - such a model would be perfect, and no model is perfect. But why shouldn't we strive for as close to 100% as possible?

Explain though, how is "PA value wrong"?

PA is an estimate, and is taken as unable to be wrong in a "positive" way by the game. If it is an estimate, it can be wrong in a "positive" way.

It is akin to modelling the top speed of my Golf GTI by the fastest it can go on a flat surface in optimal conditions, before realising that it doesn't work going downhill.

Share this post


Link to post
Share on other sites
Im not judging it by how fancy it looks, but by the content, i also have no intention of telling SI where to put this on the list, or bin as i would. It still comes down to removing a, suppose to be hidden, restraint and replacing it with another. The only real problem with the PA system is when someone looks at it as a number, without knowing that number it is impossible to find situations you do not agree with, to which you will reply something about not knowing there is a problem by not looking under the hood and we can all move onto Francis Jeffers or Pedro and debate their careers over and over.

Not really. If you have a 90/90 player averaging 9.0 every match in the Premier League, you will know something is wrong as he won't develop.

Just because we can't see something doesn't mean there isn't a problem with it. If I gave you a black box that was a calculator and it sometimes spat out wrong results, you'd say there was something wrong with this black box. If you opened the box up and found where the problem was, I don't get to hide behind the "you aren't supposed to open it up or investigate!" excuse!

I still dont agree that a system without PA is better and i still think its a very arrogant thing for you to say when you have no real idea if your system would work or not with the FM code, as you have stated yourself.

I'd like to know why you'd disagree.

It's not arrogance. I've already said my system isn't perfect and without source code it is, quite frankly, impossible to know. But you have said that you wanted to see a system without PA - I've shown you one.

I've invited people to download the thing and play around with it. People wanted to see a model that worked without limits and produced interesting graphs reflecting reality. People wanted to see a model that wouldn't blow up out of control easily - this one doesn't. People wanted to see that on average, it would be predictable - this one does.

I'm afraid I can't solve the "no source code" issue - I have to work with assumptions in the hope that I get my view across and someone on the SI side will take a look at it. That's all really. I can't build a fantastic system alone. But I've never said my system was fantastic.

It's just a proof-of-concept that shows X, Y and Z. I personally think it is better, for the multitude of reasons stated above and in other threads. Arrogance? It's really up to you to decide. I've put a model out; I've said it's not perfect; I've said it's not going to be production-perfect without access to source code - over to people to pick it apart, I think.

Share this post


Link to post
Share on other sites
you do love your car analogies :)

It's the best analogy I have. Maximum speed in optimal conditions? Check. Maximum speed in sub-optimal conditions? Check. Maximum speed in supra-optimal conditions? Ooer... Nope.

Share this post


Link to post
Share on other sites

You actually are having trouble understanding what I'm saying. Rather than getting annoyed about this I'll just accept that some people on here are going to have issues understanding what's been written, so I'll rewrite this in plain simple text for you.

PA is the ability given to a player in the database. Players may latebloom if there is PA to do so. When it happens it is unexpected because you shouldn't be able to see that PA.

As I said before the PAs that are given to players in the game aren't the true values, but instead the perceived values of this and as such under this model can and will be very wrong at times. You can assume that its right all day, but as it changes one year to another its quite clear that they get it wrong, and they get it wrong a lot.

Why should it get locked down, because its part of the system and without we'd have an essentially an infinite talent pool of players and the World would become overloaded with stupidly good players.

Onto your "they're errors", "that's like errors in the game" argument. What?! Really?! That makes no sense whatsoever. What I said was that people won't get PA right because by definition they can't. You can't see it in the game and you can't see it in real life. You can only see it in the game because you can break the rules and cheat, and there is the problem. The real Pedro might have the potential to become better, but FM11.3 virtual Pedro has the amount that he was given by some database producer and what he felt it was. It would be like you viewing a regen in the game world and being asked to give him a PA when he was 16 or 17. You could have a guess, but you're very unlikely to get it right, but that's how the numbers are decided. Hence they change year on year.

The model is dependant on performances to do this, however performance values are dependant upon leagues. There are links between values that you don't take account of which can lead to highly unstable systems.

Again, I don't know of any players it fails to accurately model. Pedro was your example and by having his PA higher, as it may very well be, it works fine. Where's the problem?

Your model is closer from perfect. It would be like replacing quantum mechanics with the plumb pudding model because its easier to demonstrate to children.

PA is not an estimate, it is a value given to players within the game. PA of generated players is by definition correct. It is a guess for real players, but once they are in the game that is them. Some can late bloom, some will flop, etc.

Share this post


Link to post
Share on other sites
Not really. If you have a 90/90 player averaging 9.0 every match in the Premier League, you will know something is wrong as he won't develop.

Just because we can't see something doesn't mean there isn't a problem with it. If I gave you a black box that was a calculator and it sometimes spat out wrong results, you'd say there was something wrong with this black box. If you opened the box up and found where the problem was, I don't get to hide behind the "you aren't supposed to open it up or investigate!" excuse!

I'd like to know why you'd disagree.

It's not arrogance. I've already said my system isn't perfect and without source code it is, quite frankly, impossible to know. But you have said that you wanted to see a system without PA - I've shown you one.

I've invited people to download the thing and play around with it. People wanted to see a model that worked without limits and produced interesting graphs reflecting reality. People wanted to see a model that wouldn't blow up out of control easily - this one doesn't. People wanted to see that on average, it would be predictable - this one does.

I'm afraid I can't solve the "no source code" issue - I have to work with assumptions in the hope that I get my view across and someone on the SI side will take a look at it. That's all really. I can't build a fantastic system alone. But I've never said my system was fantastic.

It's just a proof-of-concept that shows X, Y and Z. I personally think it is better, for the multitude of reasons stated above and in other threads. Arrogance? It's really up to you to decide. I've put a model out; I've said it's not perfect; I've said it's not going to be production-perfect without access to source code - over to people to pick it apart, I think.

The thing is if you have that blackbox and it is spitting out the right results (as is the case), then opening it up and finding a hamster in a wheel could make you question if it was doing it right or not. However, if it is working properly, who are you to say that the hamster in the wheel isn't the right way to do it?

Also, your model didn't work in any way and didn't reflect reality in any way. If that was an acceptable model, my job would be so incredibly easy. But it's not, you have no reasoning for the way you did it, you have no working system that involves real input and you haven't even compared it to a real situation. What you have is a series of glorified parabolas with some random number generators perturbing the values and calling it a model. Even beyond that it isn't thought through and there is a feedback issue (as well as many others) that would lead to an unstable global population.

Again, it's a very poorly thought out model that hasn't even been shown to replicate real life. You've just drawn some graphs on excel and given yourself a pat on the back. Also, stop hiding behind source code, you couldn't even understand it if you had it.

Share this post


Link to post
Share on other sites
Not really. If you have a 90/90 player averaging 9.0 every match in the Premier League, you will know something is wrong as he won't develop.

Just because we can't see something doesn't mean there isn't a problem with it. If I gave you a black box that was a calculator and it sometimes spat out wrong results, you'd say there was something wrong with this black box. If you opened the box up and found where the problem was, I don't get to hide behind the "you aren't supposed to open it up or investigate!" excuse!

you using an extreme example to prove your point, show me a 90/90 player that can average a 9.0 in FM, make him up if you want, its not possible unless the most extreme conditions are met. The point is, if you didnt know about PA as a number, then you would never be able to judge when someone had peaked, ive had players who have continued to develop up till they start declining due to age, not huge changes, maybe one attribute going up in a year, but its still development, again ill point to the fact that maybe 1 in every thousand or more players will reach their PA number in FM, players will get close, even 1-2 points close, but in all my time using FMRTE checking up thousands of players, top players or lower league players i have yet to see one maxed out. That means that almost every single player in the game has room to improve always, even with a limit they slow down their rate of development the older they get.

I'd like to know why you'd disagree.

It's not arrogance. I've already said my system isn't perfect and without source code it is, quite frankly, impossible to know. But you have said that you wanted to see a system without PA - I've shown you one.

I've invited people to download the thing and play around with it. People wanted to see a model that worked without limits and produced interesting graphs reflecting reality. People wanted to see a model that wouldn't blow up out of control easily - this one doesn't. People wanted to see that on average, it would be predictable - this one does.

I'm afraid I can't solve the "no source code" issue - I have to work with assumptions in the hope that I get my view across and someone on the SI side will take a look at it. That's all really. I can't build a fantastic system alone. But I've never said my system was fantastic.

It's just a proof-of-concept that shows X, Y and Z. I personally think it is better, for the multitude of reasons stated above and in other threads. Arrogance? It's really up to you to decide. I've put a model out; I've said it's not perfect; I've said it's not going to be production-perfect without access to source code - over to people to pick it apart, I think.

i disagree because i am one who thinks that each person has his or her limits at any given thing. So i personally agree with the PA system, yes there are always improvements that could be made and will be made but i dont think scrapping the idea they have is going to be any better. On top of that the system does work, its all anictodal but i could show you careers in my FM game that follow your charts and your expectations, players who peaked early, late, never at all, all of that already happens. The only difference between what your proposing and what is in place is that currently you can look up PA using a third party tool as things stand. If that was not possible this argument would never happen, because we would be unaware of this magical number and as they developed the system and the regen model peopel would not know the difference.

The only reason i say it sounds arrogant is because you put it across as your way is definately better, without being able to prove it, obviously because you dont have access to the code which is not your fault.

Share this post


Link to post
Share on other sites
You actually are having trouble understanding what I'm saying. Rather than getting annoyed about this I'll just accept that some people on here are going to have issues understanding what's been written, so I'll rewrite this in plain simple text for you.

Actually, no, I do understand. I just disagree.

PA is the ability given to a player in the database. Players may latebloom if there is PA to do so. When it happens it is unexpected because you shouldn't be able to see that PA.

Just because we cannot see something doesn't mean that there isn't anything wrong with the mechanics behind it. We are not allowed (in general) to see what goes on inside North Korea but for what little comes out of it, we know there is a serious issue, even though we don't know the full military structure of North Korea and rely on defectors who are likely biased against the regime.

I believe that there are specific reasons why players late-bloom. They don't suddenly decide to shoot up - they develop suddenly for a reason. In Luca Toni's case, for example, it was sustained development over a whole season - consistently brilliant. This could be one reason to late-bloom. Notice, however, that the expected maximum "ability" of Luca Toni isn't taken into account here. For good reason! Luca Toni was never expected to late-bloom - the researcher saw a mediocre mid-table striker at an age where most strikers peak. There was no reason to increase his PA. His expected maximum "ability" was immaterial.

In other words, a player who is consistently brilliant over a whole season should late-bloom, whether it was expected (you should buy a lottery ticket) or not (clearly most of the time).

This doesn't match the game, however. A player, say 130/130 a regen version of Luca Toni (equal in all attributes), who is consistently brilliant over a whole season, will not late-bloom, full-stop. He can't late-bloom because the researcher deemed him unlikely to develop further. But like I said, this shouldn't matter - Toni wasn't expected to develop further, but did. This 130/130 player wasn't expected to develop further and showed the same or similar level of development as Toni - should he not late-bloom too?

For a game to be realistic, if Y happens as a result of X in reality, then Y must happen as a result of X in-game. Conversely, if Y happens as a result of X in-game, then Y must happen as a result of X in reality. A bit like "Game if and only if Reality". There is a mismatch here - Toni (Y) late-bloomed as a result of consistently-brilliant performances (X) in-reality; Toni (Y) does not late-bloom as a result of consistently-brilliant performances (X) in-game - a mismatch. Therefore the game is not realistic.

As I said before the PAs that are given to players in the game aren't the true values, but instead the perceived values of this and as such under this model can and will be very wrong at times. You can assume that its right all day, but as it changes one year to another its quite clear that they get it wrong, and they get it wrong a lot.

I'm not assuming it's right. The game is! The game takes the PA value input by a researcher - a value that is by definition uncertain - and then uses it as a certain upper-bound, despite the fact that it is uncertain.

I did say that they get it wrong and get it wrong a lot. I'm not sure what you are getting at here.

Why should it get locked down, because its part of the system and without we'd have an essentially an infinite talent pool of players and the World would become overloaded with stupidly good players.

This is the part I disagree with the most. Take any match in FM. There is no upper-limit to a game's scoreline (except arithmetic overflow, 2.1 billion or something), but why do we not see scorelines like "Manchester United 20,082-583,948 Arsenal"? Just because scorelines are basically unlimited doesn't mean that the scores suddenly go haywire.

The same thing applies here. My proposal is that talent determines the speed at which a player develops - if you are talented, you find it easy to develop; if you are not talented, you will struggle. On average, the cream rises to the top.

It does not imply you can take any old player and turn him into Lionel Messi with ease. If you are a chef, you will always want a sharp knife to cut meat rather than a blunt knife. The same applies here - you are more likely to succeed in turning a talented player into Lionel Messi than a Blue Square North youth failure. It may not be impossible, but it will definitely be incredibly difficult to the extent that you are more likely to win the lottery many times over. In other words, not worth considering.

A car analogy is saying that a Golf GTI has a theoretical extremely high speed if you chuck it off a cliff. But this is hardly going to affect any racing game since falling off a cliff is unlikely to win you a race - it's not worth considering.

Onto your "they're errors", "that's like errors in the game" argument. What?! Really?! That makes no sense whatsoever. What I said was that people won't get PA right because by definition they can't. You can't see it in the game and you can't see it in real life. You can only see it in the game because you can break the rules and cheat, and there is the problem.

Again related to "just because you can't see something doesn't mean there is no problem with it".

Besides, some people work with the editor to create custom databases. It's not outlandish to figure out the mechanics.

And even if you did, you should be allowed to pick apart fallacies in the logic rather than hide behind the excuse that "you shouldn't see that."

The real Pedro might have the potential to become better, but FM11.3 virtual Pedro

I am not concerned about FM11.3 Pedro. I am looking at FM07 Pedro, with the low PA. FM07 Pedro is not allowed to become the Pedro today due to PA. FM11.3 Pedro is clearly more correct than FM07 Pedro because FM07 Pedro was defined without a crystal ball.

has the amount that he was given by some database producer and what he felt it was. It would be like you viewing a regen in the game world and being asked to give him a PA when he was 16 or 17. You could have a guess, but you're very unlikely to get it right, but that's how the numbers are decided. Hence they change year on year.

So why can we not have this "change mechanism" in the game year-on-year? Surely this is more realistic?

The model is dependant on performances to do this, however performance values are dependant upon leagues. There are links between values that you don't take account of which can lead to highly unstable systems.

Well then, perhaps SI need to do a little analysis with mathematical packages. They already do soak tests - they will be able to design stable systems. Again, a lack of limits does not imply absolute chaos.

Again, I don't know of any players it fails to accurately model. Pedro was your example and by having his PA higher, as it may very well be, it works fine. Where's the problem?

If Pedro's PA were higher in FM07, then it would be alright.

But it's not higher than it is in FM07!

Can't you see the problem right there?

It is a bit like saying, "If Arsenal had a decent goalkeeper, they would have won at least one trophy. Where's the problem?" It's right there, in front of your eyes...

Your model is closer from perfect. It would be like replacing quantum mechanics with the plumb pudding model because its easier to demonstrate to children.

Well I'm sorry I don't have time to design a system that is essentially a final-year University project (assuming access to source code). All it is is a proof-of-concept that a limitless system is predictable. There are some things I simply can't show.

Besides, if you find a flaw in the match-engine, you aren't expected to write a new match-engine from scratch!

I only wrote this because people didn't believe that something with a potentially-infinite outcome was predictable and stable. Despite my calls towards things like Brownian motion and various casino games.

PA is not an estimate, it is a value given to players within the game.

It is an estimate that is turned (incorrectly) into a concrete value in-game. It is like saying "China has approximately 1 billion people today" and then sticking 1 billion into your model-of-the-world game as a hard limit. What if you underestimated this?

PA of generated players is by definition correct. It is a guess for real players, but once they are in the game that is them. Some can late bloom, some will flop, etc.

You shouldn't distinguish between real and regen. The reason is that two players with exactly the same attributes, but one is a regen and one is real, are essentially exactly the same in the context of the game.

It is just that I use real-life examples because they are easier to understand, and because nobody cares nor knows about the late-blooming nature of Nicolo di Stefano, a regen in my FM08 game. Reality is always easier for analogies and explanations.

The issue lies with the engine, too - the input data is immaterial, per Model-View-Controller.

The thing is if you have that blackbox and it is spitting out the right results (as is the case), then opening it up and finding a hamster in a wheel could make you question if it was doing it right or not. However, if it is working properly, who are you to say that the hamster in the wheel isn't the right way to do it?

The black box is spitting lots of correct results, but it is also spitting out some wrong ones (i.e. Pedro in FM07). It is the wrong results I am interested in.

Also, your model didn't work in any way and didn't reflect reality in any way. If that was an acceptable model, my job would be so incredibly easy. But it's not, you have no reasoning for the way you did it, you have no working system that involves real input and you haven't even compared it to a real situation. What you have is a series of glorified parabolas with some random number generators perturbing the values and calling it a model. Even beyond that it isn't thought through and there is a feedback issue (as well as many others) that would lead to an unstable global population.

Of course it's not a perfect model - I've never claimed it so. All it was was a proof-of-concept that potentially infinite outcomes don't lead to unstable systems.

But like I said, I don't have time to design a full system from scratch. It will take months - with source code - to do so, and I am not employed by SI.

Even if I did, it would get shot down anyway, because let's face it, no model is perfect. Neither is SI's development engine.

Again, it's a very poorly thought out model that hasn't even been shown to replicate real life. You've just drawn some graphs on excel and given yourself a pat on the back. Also, stop hiding behind source code, you couldn't even understand it if you had it.

I've worked with complex systems before. I've worked with systems that deal with infinite outcomes, even those that never converge.

I know enough to know that the normal distribution is used so widely in the world despite the fact P(Z>10000000000000)>0 - i.e. you can get absolutely stupid numbers generated from a normally-distributed random variable, but nobody cares! Otherwise we would all have to stop doing clinical trials because there is a negligible chance of death of this medicine.

It was never intended to replace SI's incredibly complex algorithm. It is a vastly-simplified model showing how you might begin to think about things. But you cannot actually go ahead and write without access to the design documents in SI Towers - that is basic software design!

I apologise for not being able to dedicate myself for a whole year into writing a scientific paper, buying a few extremely powerful servers to crunch me some numbers and publishing it in Nature in order to design an even better model for SI. All I have is a 5-year-old laptop and not even Excel. Sometimes, all you have is an idea, a proof-of-concept and some arguments. You are essentially asking me for the impossible.

I think it's unhealthy to ask for a full-blown model, anyway - otherwise how would you suggest major improvements to the engine? "Go write one yourself!" shouldn't be the answer to all major suggestions in this forum.

you using an extreme example to prove your point, show me a 90/90 player that can average a 9.0 in FM, make him up if you want, its not possible unless the most extreme conditions are met.

That is the point! I am arguing that "extreme" conditions are possible.

The extreme"ness" is simply down to the fact that "slightly underestimated" doesn't get your mind thinking.

It doesn't have to require extreme conditions. The moment the game starts giving a player a development "boost" that is greater than expected, the game starts to become wrong.

For example, let us say a player is 90/140 as a youngster with a good attitude and surrounded by solid training facilities. He is expected to develop fairly smoothly into a mid-table Premier League player as defined by the researcher.

But what if his development is anything but smooth?

An "extreme" development curve might be the player bursting into the first-team and performing so well the manager cannot leave him out because he is on fire, scoring goals for fun. This is a bit like the "9.0" scenario above. The development no longer suits the researcher's expectations - it has, quite frankly, exceeded the normal ranges of youth development. At that point, the researcher's thinking arguably no longer applies - by the time he is a year older, he would be miles ahead of his peers and would have a nice new five-digit-per-week salary.

This could happen in reality. Unlikely perhaps, but it could happen.

You can then tone down the extremity. Imagine a youth player who gets a sustained run thanks to injuries to other players, and performs admirably. Maybe a "7.5" scenario of some sort. He has still exceeded the researcher's expectations.

The point is, if you didnt know about PA as a number, then you would never be able to judge when someone had peaked, ive had players who have continued to develop up till they start declining due to age, not huge changes, maybe one attribute going up in a year, but its still development, again ill point to the fact that maybe 1 in every thousand or more players will reach their PA number in FM, players will get close, even 1-2 points close, but in all my time using FMRTE checking up thousands of players, top players or lower league players i have yet to see one maxed out. That means that almost every single player in the game has room to improve always, even with a limit they slow down their rate of development the older they get.

A shame this doesn't apply to late-bloomers, eh? Seasons of mediocrity proceeded by a gigantic push in development.

You won't be able to judge when a player peaks in a PA-less system, anyway. In fact, it will be more mysterious, since a player could "peak" twice (peak once, plateau, rediscover old form, plateau again) or maybe even more.

i disagree because i am one who thinks that each person has his or her limits at any given thing.

You would have a limit with the PA-less system too. It just won't be written down in the black box somewhere. And the box is smart enough to get a player to peak when circumstances permit.

So i personally agree with the PA system, yes there are always improvements that could be made and will be made but i dont think scrapping the idea they have is going to be any better.

Why?

On top of that the system does work, its all anictodal but i could show you careers in my FM game that follow your charts and your expectations, players who peaked early, late, never at all, all of that already happens.

Yes, but it doesn't match reality (search "Game if and only if Reality"). That X, Y and Z happen are meaningless if the circumstances leading up to it don't match reality.

The only difference between what your proposing and what is in place is that currently you can look up PA using a third party tool as things stand. If that was not possible this argument would never happen, because we would be unaware of this magical number and as they developed the system and the regen model peopel would not know the difference.

Well, no, because you have the editor, that automatically implies some method of thinking. Plus if you read the data forums, you will infer many things.

Personally, I think you'd see issues anyway. If I somehow make FM07 Pedro perform exactly as he did in reality at that age, I will wonder why he doesn't develop into Pedro today. Heck, if he averaged 10.00 every match, he wouldn't become Pedro today. I would know there are issues because players peak for no reason - and I believe players don't peak for no reason.

The only reason i say it sounds arrogant is because you put it across as your way is definately better, without being able to prove it, obviously because you dont have access to the code which is not your fault.

I think my idea is better because it works for more players (my idea would encompass Pedro and his ability to exceed what was given to him as a PA value in FM07). I wouldn't say that Excel model is better than SI's development engine - but I never said this was the case. People wanted a model that showed that a model with possibly-infinite outcomes was predictable and stable - I showed them one that very roughly approximated that. If I had (lots of) time I could probably write something miles more concrete, and if I had access to design documents and source code I could probably cut that time down considerably - but it wouldn't be easy.

All I have is an idea, some arguments and a proof-of-concept. People are getting hung up on the proof-of-concept as being The Model That Will Overthrow SI's Development Model Because It Is Definitely Better As It Stands - it's not. It's a proof-of-concept illustrating one of my points. It's either that or a few more thousand words, which benefits absolutely no-one.

Share this post


Link to post
Share on other sites
I believe that there are specific reasons why players late-bloom. They don't suddenly decide to shoot up - they develop suddenly for a reason. In Luca Toni's case, for example, it was sustained development over a whole season - consistently brilliant. This could be one reason to late-bloom. Notice, however, that the expected maximum "ability" of Luca Toni isn't taken into account here. For good reason! Luca Toni was never expected to late-bloom - the researcher saw a mediocre mid-table striker at an age where most strikers peak. There was no reason to increase his PA. His expected maximum "ability" was immaterial.

In other words, a player who is consistently brilliant over a whole season should late-bloom, whether it was expected (you should buy a lottery ticket) or not (clearly most of the time).

The consistent season of good performances by Toni was the late bloom it didn't lead to it. your whole model is based on players improving their attributes after they play well but in any learnt skill improved performances always follow improvements in technique and understanding. Some times it can be hard to tell the two apart as practice is often unobserved performance but there is a difference.

Luca Toni isn't even a good example of a late bloomer his goal tally seams to have steadily improved as he got older and then started to tail off in his early thirties.

A better example would be Tony Cascarino who was a bit of a top flight donkey in England then moved to France in his early thirties and became a hero and earned the nickname Tony Goal.

Cascarino has written an auto biography and in it he says that the difference was being taken away from the drinking culture, feeling unencumbered by expectation and a change in his mental attitude. Professionalism, Pressure and determination would be the FM attributes that would have changed before his performances improved.

Share this post


Link to post
Share on other sites
I think it's pretty silly limiting a players potential because IRL you really can't know how good a footballer can eventually turn out to be.

Stop using FMGenieScout. Problem solved.

Share this post


Link to post
Share on other sites

There's so many problems with your post x42bn6 and I can frankly can't be bothered repeating myself entirely, because if you reread the last 6-8 posts you'd get your answer to all of that. However, I'll make this point again.

The issue, and the ridiculous assumption that is leading you to this silly standpoint is that you believe that any Joe Bloggs can suddenly become a better players by playing well and that as such the player should have a PA increase in order to allow them to become better. The issue of this is that the player is already playing better, so the question has to become why does it need to have become better if he is already playing better. Let's take an analogy of the universe. In the study of galactic dynamics, especially of elliptical galaxies one uses Luminosity profiles whilst all the theory uses mass profiles. They normally take a one to one correspondence or in some cases slightly more complicated functions relating the two. The issue here is that there is a vague link but it's not perfect and as such if one incorrectly assumes of relation it creates all sorts of theoretical issues. This is exactly the same, except now you are relating how good a player is directly to their CA/PA. However, you want them to improve by playing better which would mean that they'd already seem like they were a better player. This is where firstly a nice little feedback loop would come in and secondly the cause of a larger issue of global instability (the talent pool wouldn't go to infinite by the way) caused by how the ratings are given. In any case you have an inconsistent argument that requires players playing better than they should, hence being better players and then physically making them better. You are tangling variables in a strange way, hence making the model complicated. Actually, pointlessly complicated as it would not make the game any better.

Now, the other issue is that you keep talking about Pedro and Toni. Again, the game didn't make them any better on those older versions and so you've said that they weren't allowed to late bloom. The players who do are irrelevant, but the fact it happens is what's important. I do get that you feel that anyone should have the chance, no matter how unlikely it is, however the issue is yet again a set of poor assumptions. The first being that the players played better not because they got better (late blooming as it is in the game), but instead because they just sort of did, which in turn lead to improvement. There are Pedros and Tonis in the game, it just can't be ever Joe, Dick and Harry. As far as the CA/PA model is concerned both of them had that extra potential and they just didn't fulfil it at first.

Also, your North Korea example is just silly. You're looking like a Troll now, so I'd suggest toning down the hyperbole or just finding more car analogies.

As for young players on fire who just stopped getting better, the Premier League is full of them. Absolutely full, spend an hour looking over a few of them, I think you'll have a bit of fun.

Anyhow, back to silly assumptions. You still are treating everything in terms of an observer, however the game isn't an observer of the CA/PA. What it sets is fact for the game World. Whether it lines up with the real world is irrelevant at that point. There will be Pedros and Tonis, there will be all sorts and it mirrors real life extremely well. I don't know what starting your war against the CA/PA system, maybe you have a good young player who had a good month, stopped developing and was crap and now feel disdain towards a system that didn't make him the new Cristiano Ronaldo, who knows. Also, your model is still not a good model no matter what you throw at it, I could put together a better one than that for either side without much difficulty that is far more explanatory than that rubbish, but hey...

Share this post


Link to post
Share on other sites

It is an estimate that is turned (incorrectly) into a concrete value in-game. It is like saying "China has approximately 1 billion people today" and then sticking 1 billion into your model-of-the-world game as a hard limit. What if you underestimated this?

ROFL you've had some decent arguments in this thread, but I'm going to have to call you out on this one. If we see that China has approximately 1 billion people today, then we stick 1 billion into our model as the CA.

I think the bottom line is, you need to live with what the SI "experts," if I can call them that, have educationally guessed is each player's potential. If you can't live with that, and you use the correct definition of potential, then you will have to set every player, current and regen, to have a PA of 200. That's what SomeGuy means when he says you model will lead to a ridiculous number of "stupidly good players."

To illustrate the definition of potential again, I will use a similar analogy to that of the car, but relate it to football. Say I have some 20-year old player Joe, who a large number of experienced scouts consider to have potential X. Five years later, he's playing at X ability, but we suddenly find out he had been playing with chains around the ankles! (How would you not know this?? but go with it). Well, turns out his potential seems to be 2X. Well, great, the scouts can set his new potential to 2X! You can now take this two ways:

1) Say it turns out that it is also impossible to remove these chains. Too bad, Joe's potential is still X. Did we make a mistake in judging his ability? Yes. But does it matter to change his potential? NO, because it's a video game, that still works very hard at making as realistic of a model as possible.

2) Well, chains off or not, we know Joe can play at 2X ability because everybody knows that you play footy exactly 2x better without chains tied around your ankles than if you have them. Well, what if we also give Joe surgery and turn him cell by cell into Lionel Messi? What if we give him steroids? (btw, finding fallacy in your car analogy here, feel free to criticize mine) What's the best football player ever possible? He should have a PA of 200. Then, if it's POSSIBLE, even 1 in 1000000 games, someone can achieve this ability (either naturally or unnaturally), they should have a PA of 200 too. Again, this would lead to an infinite number of talented players.

Finally, what if player B does have an in-game PA of 200? Should he be allowed to go over 200 PA? If so, then you would still be normalizing to some average, minimum, and maximum values. There is very little point in introducing higher values. In a real-world question, what happens a few years down the road when SI is using say 250 as the max PA value in their code? Should we arbitrarily change it to 300? Clearly the best player in the game will have 250 potential by then, and we will need to give them a few more points above 250 for the variable potential.

Good day to you.

Share this post


Link to post
Share on other sites

Actually, the instability idea was based around the idea that the player could average better and as such got to become better players. It's something that you would observe if say you had a player who was great in real life but terrible in FM that year, but in a game sense he's either just having a good season or their CA/PA doesn't really say a huge amount about their overall ability. The issue occurred in that you are assuming that they are both driving eachother and then suddenly you've got a beautifully unstable system. The player becomes better so he players better and because he's playing better he'll become better, etc. You can't have linked parameters driving eachother. Well, you can, but in this instance its complex and not warranted.

Share this post


Link to post
Share on other sites
ROFL you've had some decent arguments in this thread, but I'm going to have to call you out on this one. If we see that China has approximately 1 billion people today, then we stick 1 billion into our model as the CA.

I think the bottom line is, you need to live with what the SI "experts," if I can call them that, have educationally guessed is each player's potential. If you can't live with that, and you use the correct definition of potential, then you will have to set every player, current and regen, to have a PA of 200. That's what SomeGuy means when he says you model will lead to a ridiculous number of "stupidly good players."

To illustrate the definition of potential again, I will use a similar analogy to that of the car, but relate it to football. Say I have some 20-year old player Joe, who a large number of experienced scouts consider to have potential X. Five years later, he's playing at X ability, but we suddenly find out he had been playing with chains around the ankles! (How would you not know this?? but go with it). Well, turns out his potential seems to be 2X. Well, great, the scouts can set his new potential to 2X! You can now take this two ways:

1) Say it turns out that it is also impossible to remove these chains. Too bad, Joe's potential is still X. Did we make a mistake in judging his ability? Yes. But does it matter to change his potential? NO, because it's a video game, that still works very hard at making as realistic of a model as possible.

2) Well, chains off or not, we know Joe can play at 2X ability because everybody knows that you play footy exactly 2x better without chains tied around your ankles than if you have them. Well, what if we also give Joe surgery and turn him cell by cell into Lionel Messi? What if we give him steroids? (btw, finding fallacy in your car analogy here, feel free to criticize mine) What's the best football player ever possible? He should have a PA of 200. Then, if it's POSSIBLE, even 1 in 1000000 games, someone can achieve this ability (either naturally or unnaturally), they should have a PA of 200 too. Again, this would lead to an infinite number of talented players.

Finally, what if player B does have an in-game PA of 200? Should he be allowed to go over 200 PA? If so, then you would still be normalizing to some average, minimum, and maximum values. There is very little point in introducing higher values. In a real-world question, what happens a few years down the road when SI is using say 250 as the max PA value in their code? Should we arbitrarily change it to 300? Clearly the best player in the game will have 250 potential by then, and we will need to give them a few more points above 250 for the variable potential.

Good day to you.

I love how you feel as if you've made some valid points ^^

Seriously, that's possibly the worst thing ever written down in any language throughout time

Share this post


Link to post
Share on other sites
The consistent season of good performances by Toni was the late bloom it didn't lead to it. your whole model is based on players improving their attributes after they play well but in any learnt skill improved performances always follow improvements in technique and understanding. Some times it can be hard to tell the two apart as practice is often unobserved performance but there is a difference.

I don't quite understand the bit where you say "it didn't lead to it". And Toni's form was based on many seasons.

Luca Toni isn't even a good example of a late bloomer his goal tally seams to have steadily improved as he got older and then started to tail off in his early thirties.

At 26-27, he was banging them in - in Serie B. Then he had a solid season in Serie A followed by his incredible season with Fiorentina, followed by another great one (with injuries), before again outdoing himself with Bayern Munich (then collapsing). That is a late-bloom, surely? He should have peaked at 27, but he continued to improve until he was in his 30s.

A better example would be Tony Cascarino who was a bit of a top flight donkey in England then moved to France in his early thirties and became a hero an earned the nickname Tony Goal.

Cascarino has written an auto biography and in it he says that the difference was being taken away from the drinking culture, feeling unencumbered by expectation and a change in his mental attitude. Professionalism, Pressure and determination would be the FM attributes that would have changed before his performances improved.

Certainly, an improved attitude leads to improved performances on the pitch (Mr. Balotelli, I'm looking at you), but this is exactly where the match-engine falls down - Cascarino would not have had a great PA upon leaving England, but would have exceeded that not-so-great PA after a few seasons in France.

Imagine a game where Cascarino has just left for France - his PA is pathetic and his CA is useless. The Marseille researcher is cursing his club's board for signing such an idiot. Then imagine if Cascarino's personality changes in this game (strictly-speaking, this isn't possible without tutoring... But say it is possible) - just like reality - and performs exactly how he does in reality, in the game. Surely the most realistic game in the world would recognise that:

- His PA isn't appropriate for this player, who has 61 league goals in 84 appearances

- He is a late-bloomer

- His PA shouldn't hold him back from developing (albeit perhaps not as quickly as he would have developed if he was 16)

?

There's so many problems with your post x42bn6 and I can frankly can't be bothered repeating myself entirely, because if you reread the last 6-8 posts you'd get your answer to all of that. However, I'll make this point again.

The issue, and the ridiculous assumption that is leading you to this silly standpoint is that you believe that any Joe Bloggs can suddenly become a better players by playing well and that as such the player should have a PA increase in order to allow them to become better. The issue of this is that the player is already playing better, so the question has to become why does it need to have become better if he is already playing better.

Because he's learning!

You have surely heard of "second-season syndrome" where a player who smashes all records suddenly turns into Fernando Torres at Chelsea. Why does this happen? One of the reasons is that other players suddenly realise that this player is quite dangerous, and hence they need to deal with him in a different way.

Imagine a winger whose skill is pure pace. In his debut season defenders fall to the hype and wet themselves when he runs at them. However, eventually someone figures out he's not actually that good and starts trying to do things like push him inside or crowd him out.

Two things can happen here. First, the winger might never learn from this and will continue trying to beat his man - defenders will start to copy each other and eventually his performances drop. This is down to the winger not learning - i.e. not developing.

Secondly, the winger might actually adapt to this situation. He might bulk up so he can continue beating his man anyway, shoving them out of the way. He might time his runs better or play the offside trap better. He might learn how to be devastating when cutting inside too. He might learn how to play through the middle where his pace is deadlier against centre-backs. Here, the winger is learning - i.e. developing.

Now, notice here I haven't actually mentioned "potential" or "talent" or whatever. My belief is that it doesn't really matter in some sense. It matters only because a talented player is more likely to recognise how to beat his man differently - imagine, perhaps, Cristiano Ronaldo at 16, where all he did was run around and quite often lose the ball - but Ronaldo quickly realised that he couldn't just keep dribbling. He learnt many things at Sporting CP and Manchester United. You could imagine a player - i.e. Kieran Richardson or Shaun Wright-Phillips - where "blind alley" aptly describes them because they could never adapt to improved opposition or teammates.

So to me, this says that a player who has CA 130, say, and averages 9.0 over a whole (or ideally, many) seasons consistently, should develop, regardless of whether a researcher thinks he can or not (i.e. his PA might be 130, it might be 200 - he should develop).

A player who stops developing implies he isn't learning. I'm not convinced a player who averages 9.0 despite defenders throwing more and more kitchen sinks at him over time isn't learning. Therefore he should develop. Possibly slower than he might have been as a kid. Possibly slower than Lionel Messi. But develops nevertheless.

Let's take an analogy of the universe. In the study of galactic dynamics, especially of elliptical galaxies one uses Luminosity profiles whilst all the theory uses mass profiles. They normally take a one to one correspondence or in some cases slightly more complicated functions relating the two. The issue here is that there is a vague link but it's not perfect and as such if one incorrectly assumes of relation it creates all sorts of theoretical issues. This is exactly the same, except now you are relating how good a player is directly to their CA/PA. However, you want them to improve by playing better which would mean that they'd already seem like they were a better player. This is where firstly a nice little feedback loop would come in and secondly the cause of a larger issue of global instability (the talent pool wouldn't go to infinite by the way) caused by how the ratings are given. In any case you have an inconsistent argument that requires players playing better than they should, hence being better players and then physically making them better. You are tangling variables in a strange way, hence making the model complicated. Actually, pointlessly complicated as it would not make the game any better.

I am not implying a linear relationship of any sort, however - I'm implying diminishing returns in the same way that doubling a player's CA doesn't necessarily double a player's goal returns. However, I would argue a player who keeps getting 9.0 despite getting marked by 7-8 players every game with tight marking + close-down + hard tackling + show onto weaker foot deserves an improvement akin to a domino effect. I mean, a player who averages 9.0 is doing things better than Lionel Messi! It's not out of this world that he might turn out better than Messi at his peak if he continues like that!

Let's not forget that even learning has diminishing returns. It is easy for a child to learn their mother tongue because it doesn't know much, but it is difficult for an adult to learn a new language from scratch. If you like, the higher you go, the harder it becomes to develop. We see this through player development already - the fastest development occurs between 16-21, and the slowest at around 26-28, and gets worse from there onwards. The no-PA example might imply something like "averaging 7.5 as a 16-year-old in the Premier League is absolutely fantastic; averaging 7.5 at 28 probably not" - implying you get a bigger "development bonus" at 16 than 28 if you play exactly the same. So if a player gets 9.0 every season, through development alone, he isn't going to have a "linear" sort of development.

Also, these things should in general be considered over the long-term - a 5-year moving average is less susceptible to instability.

Lastly, in reality, there are many more hoops for a player to jump through. For example, they may suffer injuries, which can set a player quite far back. They might move to a new country and fail to settle. They might fall out with the manager. And thanks to diminishing returns, a slight setback hurts the fastest movers. Imagine an F1 car encountering a speed bump as opposed to a cyclist, for example.

Now, the other issue is that you keep talking about Pedro and Toni. Again, the game didn't make them any better on those older versions and so you've said that they weren't allowed to late bloom. The players who do are irrelevant, but the fact it happens is what's important.

I agree it's important but I also think it should happen for the right reasons.

If, in a real-life race, I crash my car against my opponent, he should flip off the track and so should I. Now imagine, in my racing game, if I crash my opponent in the same way in the last sentence, he doesn't crash. It is not enough to merely say "but my opponent sometimes flips off the track anyway!" (i.e. some AIs are idiots) and "but I flip off the track sometimes anyway!" (i.e. I am a newbie) - the list of events leading up to some other event must correspond in reality in and in real-life.

For the above sentence, replace "me crashing" with "Player X plays consistently well over many seasons" and "opponent crashing" with "Player late-blooms". Certainly, in FM, it is possible for "Player X plays consistently well over many seasons" and "Player late-blooms" (as you state, as I've never denied), but the game doesn't always make it possible to "Player X plays consistently well over many seasons implies Player late-blooms".

I do get that you feel that anyone should have the chance, no matter how unlikely it is, however the issue is yet again a set of poor assumptions. The first being that the players played better not because they got better (late blooming as it is in the game), but instead because they just sort of did, which in turn lead to improvement. There are Pedros and Tonis in the game, it just can't be ever Joe, Dick and Harry. As far as the CA/PA model is concerned both of them had that extra potential and they just didn't fulfil it at first.

This wasn't the case in FM for Pedro! Pedro never had that extra potential to fill! It is like a youngster in a youth team who is clearly only there because he is huge and can run very quickly - the researcher doesn't think he has any real potential and hence doesn't give it out.

It would be the case for, say, Cherno Samba in an old FM game - he had a huge potential, but never fulfilled it. But it is different to FM07 Pedro who had a low potential, and "exceeded it".

Also, your North Korea example is just silly. You're looking like a Troll now, so I'd suggest toning down the hyperbole or just finding more car analogies.

I'm not acting like a troll. It's logic. "We aren't supposed to see CA/PA - there is no issue with it." "We aren't supposed to see inside North Korea - there is no issue with it." Essentially, it's a red herring - what we are meant to see and what has issues are two entirely different things and it is possible for all four combinations of cases to exist.

As for young players on fire who just stopped getting better, the Premier League is full of them. Absolutely full, spend an hour looking over a few of them, I think you'll have a bit of fun.

Of course there are plenty - but I'm interested in those who were never rated, yet are on fire. I'm interested in knowing why they stopped developing. I'm not convinced they stop developing for the sake of stopping development - that begs the question.

Anyhow, back to silly assumptions. You still are treating everything in terms of an observer, however the game isn't an observer of the CA/PA. What it sets is fact for the game World. Whether it lines up with the real world is irrelevant at that point. There will be Pedros and Tonis, there will be all sorts and it mirrors real life extremely well.

Of course it mirrors real-life well. But there are specific cases where it cannot mirror reality well.

However, there aren't any Pedros - players with low PA blossoming into world-class players. They simply can't become world-class players - because some researcher thought they couldn't several years back in game-time!

I don't know what starting your war against the CA/PA system, maybe you have a good young player who had a good month, stopped developing and was crap and now feel disdain towards a system that didn't make him the new Cristiano Ronaldo, who knows.

Why do you question my motives? Is it not outlandish that I've had a good think about why Pedro can't develop like he did in reality as he can in FM07?

Also, your model is still not a good model no matter what you throw at it,

I've answered some of your responses above - hopefully that changes your mind.

I could put together a better one than that for either side without much difficulty that is far more explanatory than that rubbish, but hey...

Put your money where your mouth is then.

ROFL you've had some decent arguments in this thread, but I'm going to have to call you out on this one. If we see that China has approximately 1 billion people today, then we stick 1 billion into our model as the CA.

The analogy was around uncertain measurements being used as "certain" measurements - not CA or PA or whatever. If I take the population of China, I should really include some "percentage error", for example - that captures the "uncertainty".

I think the bottom line is, you need to live with what the SI "experts," if I can call them that, have educationally guessed is each player's potential. If you can't live with that, and you use the correct definition of potential, then you will have to set every player, current and regen, to have a PA of 200. That's what SomeGuy means when he says you model will lead to a ridiculous number of "stupidly good players."

The first part is true. The second part not. If all players had PA 200, not all of them would reach CA 200. Mario Balotelli probably wouldn't, because he doesn't have a great attitude, for example.

I am not saying we remove PA and then don't change anything! I am saying we remove PA and then add a level of abstraction that ensures that a player's development has certain characteristics. For example, instead of me trying to hold you back with a rope while you try to run, I tie your feet together instead. I've removed the limiter in lieu of another hurdle in your way - leaving an open question whether you can break free of these bonds and fly, or give up and fail - but on average, you cover this athletics course in the same amount of time as if I were holding you back OR I tied your feet together.

Essentially, it replaces a number with a probability distribution.

To illustrate the definition of potential again, I will use a similar analogy to that of the car, but relate it to football. Say I have some 20-year old player Joe, who a large number of experienced scouts consider to have potential X. Five years later, he's playing at X ability, but we suddenly find out he had been playing with chains around the ankles! (How would you not know this?? but go with it). Well, turns out his potential seems to be 2X. Well, great, the scouts can set his new potential to 2X! You can now take this two ways:

1) Say it turns out that it is also impossible to remove these chains. Too bad, Joe's potential is still X. Did we make a mistake in judging his ability? Yes. But does it matter to change his potential? NO, because it's a video game, that still works very hard at making as realistic of a model as possible.

2) Well, chains off or not, we know Joe can play at 2X ability because everybody knows that you play footy exactly 2x better without chains tied around your ankles than if you have them. Well, what if we also give Joe surgery and turn him cell by cell into Lionel Messi? What if we give him steroids? (btw, finding fallacy in your car analogy here, feel free to criticize mine) What's the best football player ever possible? He should have a PA of 200. Then, if it's POSSIBLE, even 1 in 1000000 games, someone can achieve this ability (either naturally or unnaturally), they should have a PA of 200 too. Again, this would lead to an infinite number of talented players.

So if I have this right, Joe has a "disability" of some sort. If Joe's disability could be cured, then brilliant, he could become Joe "Messi" Cole. If not, he will be stuck at Joe "Joe Cole" Cole levels.

Let's enter my model and I'll explain what should happen.

Let's say Joe has a "talent" level of 130. This 130 implies that if we started a game with Joe in it, and replayed it an infinite number of times, Joe's CA peaks at 130 on average. Sometimes he might hit 200 - sometimes he suffers so many injuries his "peak" is 30. Clearly, getting a peak of 200 is going to be incredibly rare, akin to winning the lottery many times over - as is peaking at 30, since you have to be "lucky" to suffer so many injuries (i.e. the probability is low). It's important to note that 130 is talent, not PA.

Now, you might think this is obscene - how can a destined-to-be mid-table player become Lionel Messi? My argument is: Look at where the best players came from! Cristiano Ronaldo's first club was an amateur one. Messi's was Newell's Old Boys, not the best youth academy in Argentina. Luis Ronaldo, at 16, was with São Cristóvão, a club that is nowhere near the top in Brazil. Didier Drogba was with Le Mans - not a French powerhouse.

The fact is that because a youth player's career is so uncertain, "anything can happen". You cannot say with 100% certainty that 20-year-old Joe won't be better than mid-table fodder, nor top-tier amazingness, nor Lionel Messi's heir-apparent. You can imagine in my model Messi being given a talent level of 170 because the Barcelona researcher is wary about his illness, or the Sporting researcher giving Cristiano Ronaldo a talent level of 175 because although he is good, he is selfish to the extreme, or so on. In fact, for Luis Ronaldo, if you gave him a talent level of 180 or something, you would have been slapped silly by the Brazil researcher!

Here, the talent level is an approximation. It's probably wrong. Over (real) years, the researcher's perception of a player's talent becomes more accurate. For example, if Messi was given a talent level of 170, it might have gone 170 -> 175 -> 185 -> 190 over 3 seasons.

However, with this talent level rating, it means that a Messi with talent level 170 can become the best player in the world and have a "CA" of around 200. Despite the fact the Newell's researcher clearly got the talent level wrong, a user can develop Messi similar to Barcelona did to nurture his talent and exceed the Newell's researcher's opinion. Alternatively, you could sign him with Big Sam and Messi will be turned into this annoying little fella who plays in central midfield whenever Big Sam runs out of midfielders who are above 8 feet tall, rotting his existence away and perhaps peaking at CA 140 or something.

Now, back to your example - the researcher has set this 130 value and then sat at home all smug watching Joe Cole flop at Liverpool. Nothing can be done about this 130 value if it is wrong. It might actually be 120 - it might be 200. Nothing can be done about it here.

Where your argument is false is: For number 1, if he can't remove the chains, his potential essentially doesn't change; and since you cannot remove the chains, the chain itself arguably becomes part of the player's "talent". If you like, the chain is essentially an "annoying body part" that unfortunately the player can do absolutely nothing about - it is a "part of him".

I'd note that if the chain could be broken, but was unlikely, the talent value should go up (in FM12 or whatever, the next season from the start of your post) - say 180. In a perfect world, we would have perhaps a PPM stating that "this player's legs are chained together" or something - perhaps we could have some mechanism that states that while he is probably not going to get faster, he can still develop his reading of the game.

Perhaps akin to Mozart's hands developing a serious infection - he wouldn't be able to show he is a maestro - but he can still conduct music (with a stick in his mouth perhaps) and can still compose music - his talent doesn't change - but perhaps his infection could be cured. On the other hand, if we chop his hands off, then the lack of hands "becomes a part of him" - arguably his talent should go down from a "game" viewpoint (but what if we could give him a robotic hand? ...)

So for point 1), I agree.

However, for point 2), I disagree with the conclusion. Requoted for ease of me posting:

2) Well, chains off or not, we know Joe can play at 2X ability because everybody knows that you play footy exactly 2x better without chains tied around your ankles than if you have them. Well, what if we also give Joe surgery and turn him cell by cell into Lionel Messi? What if we give him steroids? (btw, finding fallacy in your car analogy here, feel free to criticize mine) What's the best football player ever possible?

This is the point I try to make all the time - we don't know what the ultimate football player is. Thanks to evolution, what is the ultimate football player changes all the time, anyway! And if football favoured gigantic 8-foot players with thighs the girth of Andy Reid's stomach, Messi arguably wouldn't be anywhere near the best player in the world. The ultimate player quite frankly, doesn't exist and is impossible to define. Therefore I think what I have quoted is essentially a red herring.

However, I do agree with what you have said above in where you've tried to claim it's nonsense. Joe plus steroids and surgery reflects Joe developing very well (in this case, unnaturally and illegally, of course) - why wouldn't he exceed talent level 130? The number "130" is only the researcher's vague guess about Joe's future. The researcher knows he's wrong - the researcher knows he might have underestimated Joe's determination to succeed, or willingness to take steroids.

He should have a PA of 200.

Essentially yes. Even without steroids and surgery. For example, maybe one day Joe discovers a trick that is easy to execute but devastatingly effective, or he runs into a purple patch that never seems to stop. You can't rule him out hitting CA 200 - although you can certainly bet your house on it that it won't happen. I argue it's possible - just really, really unlikely. And if he is willing to take steroids - that's the exception to the rule, since steroids have never been proven to be rampant in football.

Then, if it's POSSIBLE, even 1 in 1000000 games, someone can achieve this ability (either naturally or unnaturally), they should have a PA of 200 too. Again, this would lead to an infinite number of talented players.

This I disagree with whole-heartedly. Your maths is totally wrong! If there is a one-in-a-million chance of Joe becoming Messi, then in 1 million games, you will have one game where Joe becomes Messi on average.

Probability is not additive! Let us say the probability of Joe reaching CA 200 is 1 in 1,000,000 as you say. Most games have around 100,000 players in them - let us be silly and assume every single one is equivalent to Joe. The probability of at least one of them hitting CA 200 is 1-(1/1000000)^100000 = 0.095... - i.e. there is a 9.5% chance that under the assumption that the database has 100,000 Joes, 9.5% of the time you get at least one a CA 200 player. I don't know the distribution of players who would be similar to Joe but say it's a quarter - it goes down to 2.5% in this case.

Of course, this assumes the probability is one-in-a-million - in balancing, SI might think that this value is too large and will rebalance the model. Or they might even think it's too small!

Finally, what if player B does have an in-game PA of 200? Should he be allowed to go over 200 PA? If so, then you would still be normalizing to some average, minimum, and maximum values. There is very little point in introducing higher values. In a real-world question, what happens a few years down the road when SI is using say 250 as the max PA value in their code? Should we arbitrarily change it to 300? Clearly the best player in the game will have 250 potential by then, and we will need to give them a few more points above 250 for the variable potential.

By removing PA, I am saying there is no maximum, really... I sometimes say "essentially yes" because it is a lot easier to follow.

A no-PA model implies no-CA too (since PA is just maximum CA), which implies free weighting of attributes. So this notion of "maximum PA" is meaningless - we don't need PA. It is like asking the question "what is the maximum value of this attribute that doesn't exist?"

Good day to you.

Good day to you too!

Actually, the instability idea was based around the idea that the player could average better and as such got to become better players. It's something that you would observe if say you had a player who was great in real life but terrible in FM that year, but in a game sense he's either just having a good season or their CA/PA doesn't really say a huge amount about their overall ability. The issue occurred in that you are assuming that they are both driving eachother and then suddenly you've got a beautifully unstable system. The player becomes better so he players better and because he's playing better he'll become better, etc. You can't have linked parameters driving eachother. Well, you can, but in this instance its complex and not warranted.

Covered above.

Let's not forget that a player does not react by himself in the game - there are many factors working against him. If he was superhuman maybe he could overcome all these factors, but then again if he was superhuman he would arguably be able to trigger a domino effect anyway! What it simply comes down to is balance within the game, where SI will spot that this player is going crazy too often, and will tone down various factors (i.e. perhaps reduce the effect of form or something).

Share this post


Link to post
Share on other sites

I couldn't stand to read the whole thread. While I think the PA system is unrealistic, I think we can't do any better than that. But a few tweaks could help, like a much higher number of players with negative PA's. That would add depth to the game I guess.

And there are late bloomers in real life. We see many of them every season. These should somehow be implemented into the game. Perhaps a new hidden attribute determining if he can still make progress as a player or not.

One more thing, little off-topic about CAs, CA falls too fast for most players. I can understand the physical aspect (it goes down too fast too, many 32 year old players can handle 90 minutes games weekly irl) but what I don't understand is their mental attributes. I think, apart from creativity/flair/aggresiveness many mental abilities should mirror the player's experience. They should go up, rather than down. A player in his 35 would make much better decisions on the pitch compared to his decisions back when he was 23. That's called experience. And abilities like decisions/offtheball/positioning etc. should go up when the player gets older and learns from his experiences, rather then go down.

Share this post


Link to post
Share on other sites

Put together a little program that simplifies the variables in FM and creates graphs of a players CA over his career and prove to us that your PA less system works.

What I want is to be able to enter values for Talent, training facilities, coaches, Mental stats (I'm making it easy on you by combining them into one) and injure proneness hit play and watch a graph of the players CA over a fourteen season be generated. It would be great if it could generate a few graphs so we can see that its not creating wild results or the same result again and again.

Share this post


Link to post
Share on other sites

Why isn't Cascarino's PA any higher? If he's late bloomed it must have been by definition under the model. Who are you to say that it wasn't. If the researcher said it was lower then he was wrong. As for the attitude, attitude greatly affects player development in the game and a change of attitude may have made him a far better player very quickly as such. This is actually a point where the CA/PA system is far better than any other, it's speciality if you will.

Again, find a player that the CA/PA system can't describe...

Also, players with 200 PA are not that likely to ever reach it without a damn good attitude. Attitude is one of the most important things and a change in attitude by tutoring of a young player can make all the difference.

Also... You're still on about Pedro? They got his PA wrong, that's it, that's all it was! There's nothing wrong with the model in this case. What about the other thousands of players that didn't get any better eh?! Was the game wrong to not make them all randomly develop?! I mean wow, children would get this point by now and stop repeating the same inane rubbish without actually understanding what's been said. Also the rest of your post is you just repeating yourself with pointless analogies that are either irrelevent or seemingly go against your point. In either case this is just getting tiresome. Fix your model or give up!

Nileppezdel how many years have you played into the game? Late bloomers happen all the blooming time in the game. It's happened over many versions as well. They are always floating about. If you want I can describe about 30-40 cases of top end ones over my FM playing time.

Whitewolf, he never made a model, he made a pretty graph and excel and chucked a random number generator on it. Making a model would be pretty damn simple though. Actually, I'll make one, just for fun.

Share this post


Link to post
Share on other sites

x42bn6, the main issue I have with your system is it will also allow plenty of players who have been rated accurately by the researchers to then become better than they ever actually will IRL. Allowing the PA to change, either up or down, will take away from the accuracy of the game, which for the most part, is pretty much right. For every Pedro or Luca Toni that's been underrated, and every Luke Moore that's been overrated, there are also hundreds, if not, thousands of others that have been rated accurately that your system would then alter during the course of the game.

Share this post


Link to post
Share on other sites

The current system works. As it is representative of how ir works IRL.

As Some Guy says the only issue is if a researcher gets the initial 'guess' at the potential wrong. Which If we are honest isn't that often and any mistakes that are made are usually corrected in future versions. Of course determing the very maximum a player could be is the hardest thing to judge. But I still feel it is the correct way of doing things.

We arent supposed to be able to see teh PA number or for that matter the CA number. The scouts do a fairly decent job of telling you what sort of potential a player has and the players atributes give us the current ability.

If you play the game as it is intened then there isnt a problem in my opinion

Share this post


Link to post
Share on other sites

p><p>According to Tony Coton, Ali here,

That is how I view late bloomers. Not in terms of CA/PA but in how they play and what they contribute to their teams on the football field. Luca Toni was a late bloomer so in FM-terms, what if the reason for that was that his Off the Ball, Finishing, Composure, Decisions and Anticipation, Team Work and Jumping/Strength stats finally made it into the "numbers" (RL) that made all the difference? From 14 to 15 in the technical/mental attributes and from 17 to 18 in the physical ones? I don't know, but the point is that his PA actually didn't have to change at all for this to happen.

Share this post


Link to post
Share on other sites

Players do late bloom in terms of attributes, and I've posted examples previously. There are also attributes that aren't linked to CA.

Share this post


Link to post
Share on other sites
Put together a little program that simplifies the variables in FM and creates graphs of a players CA over his career and prove to us that your PA less system works.

What I want is to be able to enter values for Talent, training facilities, coaches, Mental stats (I'm making it easy on you by combining them into one) and injure proneness hit play and watch a graph of the players CA over a fourteen season be generated. It would be great if it could generate a few graphs so we can see that its not creating wild results or the same result again and again.

I already have, sort of. I've condensed all those variables into that normal distribution. What happens during a player's career is (pseudo)random anyway. I've simply assumed they are all independent (which is clearly not the case, but I do not have time to simulate several hundred games to estimate the dependency matrix to any sensible degree). For example, quality of physios will impact the length of injuries, so there's a dependency.

I think it's a valid approximation because a player in general doesn't follow a pathway that is ridiculously wild. Under the Central Limit Theorem, it will all come down to something with a mean and a variance anyway.

What does it mean to be condensed down into that normal distribution? I am assuming that there are n (doesn't really matter what n is, as long as it is greater than zero) possible and independent factors that carry equal weight, and that their average follows a normal distribution with mean 0, variance (whatever is in the spreadsheet). For example, say only training facilities and match ratings matter. Then the normally-distributed shocks are distributed as N(0,s^2) where s is the standard deviation. In other words, it assumes that the average of training facilities and match ratings is normally distributed in that way - reflecting perhaps that a player might randomly appear in an below-average place (implying a negative) and play badly (implying a negative) leading to a large negative number. Or vice-versa.

I might do something over the weekend where you can play around with "quality" to simulate a pathway. A bit like "quality of life this season". So -1 is not good, 1 is good, 0 is average, 5 is incredible, -5 is Fernando Torres, and so on - and you get one for each season. However, into this number is bundled together everything that is relevant for a player's development except talent (a proper model would factor that in too). These -1, 1, 0, 5 and -5 are merely randomly-generated numbers that correspond to the normally-distributed shocks.

http://www.x42bn6.com/bucket/Shock.xls

Hit Calculate (I think it's F9 in Excel) to generate a new graph.

The graphs I've put are deliberately exaggerated due to scale - you can play around with this. I'd lower the standard deviation, personally. But it reflects a pathway of a player's career.

Why isn't Cascarino's PA any higher? If he's late bloomed it must have been by definition under the model.

If you were the researcher who rated Cascarino's PA before he left for France, you would be slapped-down if you had given him a high PA. He simply wouldn't have had a high PA - full-stop. He should have. But then again that's a crystal ball.

Who are you to say that it wasn't. If the researcher said it was lower then he was wrong.

I'm saying the researcher was wrong! So why do we assume the researcher is correct? We do assume this - because PA is definite and cannot be exceeded!

The researcher would have given him a low PA. It was wrong. We knew when that PA was given was that it was going to be wrong, but very likely correct.

Imagine I ask you to predict a random integer between 1-10 such that it is distributed with average 5. If you won money by playing this game, you would likely pick the 5 - it's the most likely. However, you could be wrong - what if the random number is 10?

The researcher is you picking that random number. You could be wrong. You could have underestimated what it could have been. What happens in reality could be quite different.

What is a fact is that Cascarino's PA would have been low, because the researcher cannot predict the future. That is the only certainty - that the researcher is possibly wrong, but has done his best and is probably right.

Hindsight is wonderful.

As for the attitude, attitude greatly affects player development in the game and a change of attitude may have made him a far better player very quickly as such. This is actually a point where the CA/PA system is far better than any other, it's speciality if you will.

Well yes, but I don't really care about "improvement in attitude implies improvement in player ratings" - I care about "improvement in player ratings implies that a player learns more implies that player develops". Attitude is just one of the many ingredients.

Again, find a player that the CA/PA system can't describe...

Pedro! Good grief.

Also, players with 200 PA are not that likely to ever reach it without a damn good attitude. Attitude is one of the most important things and a change in attitude by tutoring of a young player can make all the difference.

That is the whole point! Players aren't guaranteed to reach their PA anyway, so why do you think there will be a feedback loop of some sort leading to instabilities? There is already some random factor that prevents players from reaching their PA (and note: this random factor isn't random for the sake of being random - it has a reason - for example, the player might succumb to many injuries). Similarly, there will be some natural barrier preventing Joe Rubbish from becoming Lionel Messi. The only difference between this system and the current system is that my system says, "what if the researcher is wrong?" Or, for the regen equivalent, "what if the randomly-generated PA is wrong?"

Also... You're still on about Pedro? They got his PA wrong, that's it, that's all it was! There's nothing wrong with the model in this case.

Yes there is! The researcher is wrong - the researcher is bound to be wrong. The game cannot assume the researcher is partially-correct by saying, "the researcher could not have underestimated Pedro's PA."

If you like, both model and researcher are incorrect, but two wrongs don't make a right.

What about the other thousands of players that didn't get any better eh?!

I've already said CA/PA works for most players, except a few. This model will close that gap somewhat.

Was the game wrong to not make them all randomly develop?!

The game is correct to let them randomly develop, but the conditions under which it happens are incorrect.

I mean wow, children would get this point by now and stop repeating the same inane rubbish without actually understanding what's been said.

Children will. Adults will look deeper.

Also the rest of your post is you just repeating yourself with pointless analogies that are either irrelevent or seemingly go against your point. In either case this is just getting tiresome. Fix your model or give up!

You know, I could say the same thing about you... The difference is, I argue the points back, whilst you tell me to "fix your model so that it's better than what SI have, which has taken over a decade of many programmers, analysts and studies to create". Apologies if I cannot do the impossible.

Nileppezdel how many years have you played into the game? Late bloomers happen all the blooming time in the game. It's happened over many versions as well. They are always floating about. If you want I can describe about 30-40 cases of top end ones over my FM playing time.

But no Pedros... Get a player who is 80/120 (or whatever Pedro was in FM07) and turn him into Pedro today.

Whitewolf, he never made a model, he made a pretty graph and excel and chucked a random number generator on it. Making a model would be pretty damn simple though. Actually, I'll make one, just for fun.

Go on then, show me. I'm still waiting.

Pick that model apart one by one, rather than call it names.

x42bn6, the main issue I have with your system is it will also allow plenty of players who have been rated accurately by the researchers to then become better than they ever actually will IRL.

When we say a researcher is "accurate", what we are saying is that the researcher has predicted, with good accuracy, at time T, what the player's ultimate peak is. However, the problem is immediately apparent - how do you know what "good accuracy" is? The accuracy of me predicting lottery numbers is the "long-term average" of the differences between my predictions and what happens in reality. However, at time T, we don't know what happens in reality! You quite frankly don't know how accurate my lottery predictions are! If you did, you and I would be filthy rich right now.

So when we say a researcher is "accurate", what we are really saying is something along the lines of "this researcher has a proven track history of getting it right". It doesn't imply that the researcher is definitely correct to any degree and certainly puts no constraints on his correctness - for the same reason (what happens in reality).

We don't know the future - there is, quite frankly, no such thing as an "accurate" researcher if "accuracy" is equivalent to "ability to predict the future". There are certainly "accurate" researchers if "accuracy" is equivalent to "has predicted youngsters accurately before" - however, I tend to call them "good" researchers, who don't jump to conclusions after one good season, for example.

So there is no such thing as an "accurate" researcher in the future sense. It is just that the player in reality may underperform or overperform the researcher's expectations - to me, an accurate game is one that allows players in the game to underperform or overperform the researcher's expectations. So no, I don't think there's anything wrong with (your quote) "to then become better than they ever actually will IRL". The moment you start your game reality and gameworld diverge - you get to play the many "what ifs" that dominate angry Internet forums. "What if Lionel Messi had moved to Blackburn Rovers?" "What if Cristiano Ronaldo had never learnt how to work as a team?" "What if England didn't play Gerrard and Lampard in midfield?" And so on.

A concrete example. Say in 2006, a young player is predicted by a researcher to be roughly a solid top player.

Scenario 1: In 2011 reality, it turns out he is the next Wayne Rooney, and in some game, he turns out to be the next Wayne Rooney.

Scenario 2: In 2011 reality, it turns out he is the next Wayne Rooney, and in some game, he turns out to be the next Emile Heskey.

Scenario 3: In 2011 reality, it turns out he is the next Emile Heskey, and in some game, he turns out to be the next Wayne Rooney.

Scenario 4: In 2011 reality, it turns out he is the next Emile Heskey, and in some game, he turns out to be the next Emile Heskey.

I personally think all 4 outcomes are acceptable, because the researcher could be wrong.

Now, the researcher might be "accurate" in your sense that he has been rated "accurately". But he might still be wrong! And in Scenario 3, reality (Heskey) is less than the game (Rooney). This implies it's alright as long as the researcher is right. But we don't know if the researcher is right. Often, in life, we ask, "But what if you are wrong?" And, quite often, it's the "wrongness" and "uncertainty" we have to cater for. If we're right - brilliant! If we're wrong - at least we can somewhat compensate for it.

Allowing the PA to change, either up or down, will take away from the accuracy of the game, which for the most part, is pretty much right. For every Pedro or Luca Toni that's been underrated, and every Luke Moore that's been overrated, there are also hundreds, if not, thousands of others that have been rated accurately that your system would then alter during the course of the game.

True, but the PA model lets Luke Young happen whilst blocking Luca Toni. This is the crux of the issue.

Imagine we reverse it a bit - instead of "PA", say researchers had to predict "MA" which is "minimum ability" - reflecting the minimum possible ability of a player at their peak. In this model, Luke Young would be impossible (he can't fail) whilst Luca Toni is possible (he can exceed).

My model allows both sides of the coin to happen.

I'm not denying that most players are roughly correct - but they sometimes underestimate. The game can somewhat compensate for overestimation (player fails to reach PA), but it cannot compensate for underestimation (peak, brick wall).

As Some Guy says the only issue is if a researcher gets the initial 'guess' at the potential wrong. Which If we are honest isn't that often

So why do researchers revise PA values over real time? Because they are wrong! And they will always be possibly wrong. Which is fine - a researcher who is always correct would be the world's richest scout, or a billionaire who is banned from buying lottery tickets.

and any mistakes that are made are usually corrected in future versions.

Doesn't help a game that has started.

Of course determing the very maximum a player could be is the hardest thing to judge.

Actually, it's impossible. You can't predict the future. You can take a guess! An estimated guess, even. But you can't predict the future.

But I still feel it is the correct way of doing things.

It's correct to predict the impossible, yes. But not realistic.

We arent supposed to be able to see teh PA number or for that matter the CA number. The scouts do a fairly decent job of telling you what sort of potential a player has and the players atributes give us the current ability.

Red herring. Scouts can get it wrong in a PA-less model too.

If you play the game as it is intened then there isnt a problem in my opinion

Red herring. Just because you're not "meant" to see something doesn't mean that there is no problem with it.

p><p>According to Tony Coton, Ali here,

That is how I view late bloomers. Not in terms of CA/PA but in how they play and what they contribute to their teams on the football field. Luca Toni was a late bloomer so in FM-terms, what if the reason for that was that his Off the Ball, Finishing, Composure, Decisions and Anticipation, Team Work and Jumping/Strength stats finally made it into the "numbers" (RL) that made all the difference? From 14 to 15 in the technical/mental attributes and from 17 to 18 in the physical ones? I don't know, but the point is that his PA actually didn't have to change at all for this to happen.

No, it doesn't, but Toni got a fairly substantial (deserved) boost that you can only go so far with faffing around with small increases.

I believe that a player's "ability" is linked to the amount of "knowledge" and "ability" he has in his mind and feet (and for some, hands). Quick feet, quick mind, strong legs, Darron Gibson's passing, etc. etc. etc. If Toni starts performing very very well, he will start to learn more - defenders will begin to treat him differently and opposition teams will start to treat the team differently too. And if Toni continues to perform well, he has learnt something - how to cope with a new and more difficult scenario. And if he has learnt something, he should develop. I'd note that "learn" could be a "go to the gym and bulk up" activity, of course, where Toni's legs get stronger after realising he has to use his strength better.

A late-bloomer must be unexpected at first, but after glancing at it for a while, you realise that it is more than just a few games of form - it's actually a player who is doing well no matter what defenders do to him. And at the end of the day, isn't this what we call development anyway? Messi and Ronaldo continued to bamboozle defences over many seasons despite being initially marked, then double-marked, then triple-marked, then triple-marked and nastily-tackled, to whole teams playing differently against them simply because they are playing. If Messi had struggled at the "double-mark" stage, you would argue he had "peaked" and "it was just a bit of form after all". He had stopped learning and developing - because he clearly wasn't learning much.

Share this post


Link to post
Share on other sites

Well, that's annoying now isn't it. I've actually built a little model (with the logic for it typed up too), I was going to use something else, but I eventually decided that excel was the best choice. It has some issues with the constants, but it does demonstrate the point. I didn't end up posting it though as the thread went dead and I can't until Tuesday now because I'm hundreds of miles from my home for a while. Long story short I'll post it for you at some point if you want.

Anyhow, let's just put some things out in the open. I really don't think this debate is going anywhere in its current form and quite frankly the way that it is being conducted is just annoying. Your posts are essentially you repeating the same waffle (and I don't mean any offense, just an observation) and to some extent (maybe a large extent) mine are following the same pattern. This essentially means that the debate has fallen into a pile of inane rubbish that I wouldn't ask anyone to have to decipher. As such I suggest that we just point out points, in simple form and try and make this a precise and target debate rather than the shotgun style that's been employed.

Let's start with:

The Claims about the current system

1. It does not produce "late-bloomers"

- This is false, they occur frequently, even 30-40-400 years into the game. Examples have been posted in previous threads and sadly I'm not on my computer to get some examples that I can post currently, I may do so later.

"but the PA model lets Luke Young happen whilst blocking Luca Toni. "

- The above point shows that this is clearly false as Luca Tonis do occur in the game. Luke Youngs also happen.

- Whilst a don't have any images of him the most striking examples are two players that I'd seen in FM10 and FM09 (they just stuck out in my memory). The first was a striker Andy Poole. He was a player who came from Soton and spent most of his early career in the lower leagues. He had a glorious season at 27 before coming the premier league at 28 before being the top goal scorer at 30, 31 and 33. He actually played for England for about 5 years during that period as well which annoyed me no end as I thought that my 5 english strikers were all a damn site better than him, but he proved me and everyone wrong in the end. He stayed kicking about in the Premier League until he was 36 before returning to Soton and leading them to back to back relegations whilst collection rather large wages for his age. A sad end to a good career.

-The second was a young midfielder I signed one season on the advice of one of my coaches. His name escapes me currently but he was one of my brightest talents for a while, he had played over 50 times at the top level as only a 19 year old, but by 21 he had actually gotten worse. Eventually I let him go on a free transfer at 23 and he kicked about as a bit of a nothing player in League One and League Two with Reading. At 27 though he won was the out of nothing the top scorer in League One and his great form carried them to back to back promotions. His heroics weren't good enough to keep them up though, but he was certainly in the shop window and was picked up by Man City for silly money at the end of the year. He even was capped for season and went on to be a vital part of their World Cup winning side under no other than Aidy Boothroyd. He was interesting in that he always had the potential but with all the chances in the World Under me he just didn't make it. But he did make it with Reading a long time later.

Actually, I can't really think of many more. Could you add the rest so I can respond to them?

Researchers are wrong and so the model is wrong

- Honestly I'm dissapointed to see this posted by anyone, let alone someone who may actually have their head screwed on (as you seem at times). The reason being that the point of PA is that under the model its taken that PA is a set limit that is outside of control. You however say that can't work however because we can't predict it.

Why is that such a problem?

- Because unless we're talking about quantum mechanics here (which we are clearly not) then even if we can't observe this in a sense, it doesn't mean it can't exist. However as an assumption its perfectly fine to set one and treat it as the value in the game world. Maybe the specific players themselves that did latebloom mightn't do so in the game, but others will and as you said we can't predict real life.

- As such the only difference between the two models is the mode by which the players latebloom and such as neither will predict which players will accurately. The only difference is who will is set before the game starts rather than caused by what happens. This then comes down to a philisophical question rather than one about the model as to a player neither will seem entirely different in this aspect.

Players having a set upper limit is wrong

- There is no reason to say this

- This is just an assumption that you yourself are making without any proof or reason

- It would be like saying that General Relativity is wrong because spacetime can't bend

Other points

- Interestingly I tested this at one point. The median difference between CA and PA for players between the ages of 23 and 32 in the game is of the order of about 20. That's enough room that on average most players can go from being an a good Championship Player (120ish) to being a good Premier League player (140ish). That's the median as well as there are players who have huge excess PA there as well.

- A good solution to researcher error would also be to just randomise all PA/CA in the game at the start by say ~10-15% depending on age. PA by ~10% for most players and as much as ~30% for young ones. CA by very little for older players and ~10% for young ones etc. They'd have to balence it properly.

- There are attributes that are not linked with CA/PA at all which can lead to large improvement of players as well

- Lastly, the instability comes from the feedback of performance to potential you want to add. Depending on the implementation. The other issue is that average ratings aren't exactly perfect in the game so if like in former FMs some positions getting better rating for no reason than others (Defensive midfielders in FM08, wingers in FM09 etc.) then you could have a collapse of one position over another.

Hopefully the points are clearer, but the stories are a bit waffly.

Share this post


Link to post
Share on other sites

I Looked up Pedro on FM 2008 CA 75 PA -6 = 90-120, average of 110. Under X42bn6's model his going to reach about 130 CA peak with S set to 10. X42bn6 you wanted to lower the standard deviation. I had to change it to S 30 before I could generate a graph where Pedro ended up with CA 160. With S set at 30 the graphs deviate wildly from the expected curve.

When a players as underestimated as Pedro was the model would need to allow players to over and underachieve very significantly if it was to have a chance of making Pedros FM career track his real life one. Hundreds of well researched players would have CA changes that make them significantly different from their real life counterparts.

Share this post


Link to post
Share on other sites
Well, that's annoying now isn't it. I've actually built a little model (with the logic for it typed up too), I was going to use something else, but I eventually decided that excel was the best choice. It has some issues with the constants, but it does demonstrate the point. I didn't end up posting it though as the thread went dead and I can't until Tuesday now because I'm hundreds of miles from my home for a while. Long story short I'll post it for you at some point if you want.

Anyhow, let's just put some things out in the open. I really don't think this debate is going anywhere in its current form and quite frankly the way that it is being conducted is just annoying. Your posts are essentially you repeating the same waffle (and I don't mean any offense, just an observation) and to some extent (maybe a large extent) mine are following the same pattern. This essentially means that the debate has fallen into a pile of inane rubbish that I wouldn't ask anyone to have to decipher. As such I suggest that we just point out points, in simple form and try and make this a precise and target debate rather than the shotgun style that's been employed.

Let's start with:

The Claims about the current system

1. It does not produce "late-bloomers"

I never said it doesn't produce late-bloomers.

I said it doesn't produce late-bloomers when in a certain set of circumstances.

The set of circumstances being: The player has reached his PA, and continues to perform very well.

"but the PA model lets Luke Young happen whilst blocking Luca Toni. "

- The above point shows that this is clearly false as Luca Tonis do occur in the game. Luke Youngs also happen.

"Luca Toni career paths" are possible, but "Luca Toni" as defined in FM06 (or whenever his goalscoring purple patch happened) is not.

CA/PA 140/145, say, turning into a Golden Boot winner and scoring goals for fun season after season, and moving to Bayern Munich, drastically improving, isn't possible.

Researchers are wrong and so the model is wrong

- Honestly I'm dissapointed to see this posted by anyone, let alone someone who may actually have their head screwed on (as you seem at times). The reason being that the point of PA is that under the model its taken that PA is a set limit that is outside of control. You however say that can't work however because we can't predict it.

Why is that such a problem?

Because a researcher's opinion is taken into gospel. If a researcher can underestimate a player, then the game should allow the simulation of players that underestimated the researcher's opinion.

If I predict that it will likely rain lightly tomorrow with my weather machine, then it could rain very heavily or not rain at all. The PA model doesn't let "rain heavily" happen.

- Because unless we're talking about quantum mechanics here (which we are clearly not) then even if we can't observe this in a sense, it doesn't mean it can't exist.

Surely that is, if any thing, a justification that the unknown is possible, and hence the unknown possibility that a researcher underestimated a player is, well, possible? Hence being allowed to exceed PA is, well, possible?

However as an assumption its perfectly fine to set one and treat it as the value in the game world.

Why is that assumption fairly fine? The only reason is that it is a simplification - i.e. it is incorrect, but it makes things easy.

Maybe the specific players themselves that did latebloom mightn't do so in the game, but others will and as you said we can't predict real life.

That's moving the goalposts. Some players can late-bloom - some can't. I believe all players in reality have the possibility to late-bloom - regardless of what the opinion of a researcher on that player is at some moment in time. Late-blooms are, after all, unexpected.

- As such the only difference between the two models is the mode by which the players latebloom and such as neither will predict which players will accurately. The only difference is who will is set before the game starts rather than caused by what happens. This then comes down to a philisophical question rather than one about the model as to a player neither will seem entirely different in this aspect.

The model itself is philosophical - models don't come down to code and code alone.

And for the bold - it is absolutely impossible to predict accurately anyway, since you can't know the future. You only know if you are accurate once the event comes to pass (or doesn't). Right now, you don't know what is going to happen a second in the future - no model is going to help you with that one.

As for the underline - I believe that it is not correct to set this late-blooming status at the start of the simulation because sustained excellent performances (leading to late-blooming) can happen regardless of PA. PA essentially "states" the future, which is incorrect, because quite frankly the model can't know the future, especially with a human being involved.

Players having a set upper limit is wrong

I never said that we have no limits.

I said it's incorrect to state right now the opinion of a player's limit and using that as a definite limit.

We have limits. We just don't know what they are going to be, or when they will happen.

I state that it makes no sense to say, "Ah in 1995, I don't think X is going to become a Premier League superstar, so I'm going to define it so it's impossible."

Other points

- Interestingly I tested this at one point. The median difference between CA and PA for players between the ages of 23 and 32 in the game is of the order of about 20. That's enough room that on average most players can go from being an a good Championship Player (120ish) to being a good Premier League player (140ish). That's the median as well as there are players who have huge excess PA there as well.

Interesting test but I'm not sure that means anything since we all know that the majority of 23-year-olds are 4-5 years off their peak. And why did you pick the median?

- A good solution to researcher error would also be to just randomise all PA/CA in the game at the start by say ~10-15% depending on age. PA by ~10% for most players and as much as ~30% for young ones. CA by very little for older players and ~10% for young ones etc. They'd have to balence it properly.

Why have such arbitrary limits? Pulling these percentages essentially out of thin air throws more assumptions into the model, making it less realistic.

A better model is to relax or remove assumptions, including the assumption the researcher's opinion of the maximum is indeed the maximum, and letting the engine work its magic.

- There are attributes that are not linked with CA/PA at all which can lead to large improvement of players as well

Irrelevant because my concern lies with those whose PAs are arguably too low, not that some attributes are useful for development.

- Lastly, the instability comes from the feedback of performance to potential you want to add. Depending on the implementation.

That is where balance comes in.

The other issue is that average ratings aren't exactly perfect in the game so if like in former FMs some positions getting better rating for no reason than others (Defensive midfielders in FM08, wingers in FM09 etc.) then you could have a collapse of one position over another.

That is an issue with the rating system, not any other model.

It is like complaining that Darron Gibson is too good in FM10 because you apply the corner exploit and Gibson is the lurker - the issue lies with the exploit, not Gibson.

The same here - if the development engine is flawed because the ratings system is flawed, the problem lies with the ratings system. Of course, this doesn't mean that the development engine is perfect, but it is not because of the ratings system.

I Looked up Pedro on FM 2008 CA 75 PA -6 = 90-120, average of 110. Under X42bn6's model his going to reach about 130 CA peak with S set to 10. X42bn6 you wanted to lower the standard deviation. I had to change it to S 30 before I could generate a graph where Pedro ended up with CA 160. With S set at 30 the graphs deviate wildly from the expected curve.

When a players as underestimated as Pedro was the model would need to allow players to over and underachieve very significantly if it was to have a chance of making Pedros FM career track his real life one. Hundreds of well researched players would have CA changes that make them significantly different from their real life counterparts.

The real reason why I didn't really like that model and was really wondering if it were a good idea to show was because you need to simulate it thousands and thousands of times before you get any sensible distribution. At s=10, it is very rare for a player to go very high. I stated that talent=140 had roughly a 1 in 47*365 chance of reaching CA 200, for example - you need thousands and thousands of simulations on average before you get a Pedro.

It is definitely possible to get one with s=1, for example - you'll just be waiting for a couple of real-life years.

Now, is this realistic? Like I said, players like Pedro are rare - they aren't commonplace. A young winger whose career appears to be going nowhere will likely end up having a career that goes nowhere - the likes of Pedro are "unlikely" to begin with.

The reason I put the s parameter in is partly for this reason - to get an interesting graph in a reasonable amount of time, I had to make it more random.

I don't think it really matters though - what matters is the methodology, not how jagged the graph is. That a sum of several normal distributions can create lots of stable graphs and a few (important: rare) weird ones, that look plausible.

In some ways, you've proved that it is possible to have a very stable model - you needed to increase s to 30 before it started wildly going all over the place. So imagine if our model is modeled such that s is less than 30 - it is; in some ways, for your interests, it would be stable enough.

Share this post


Link to post
Share on other sites

The real reason why I didn't really like that model and was really wondering if it were a good idea to show was because you need to simulate it thousands and thousands of times before you get any sensible distribution. At s=10, it is very rare for a player to go very high. I stated that talent=140 had roughly a 1 in 47*365 chance of reaching CA 200, for example - you need thousands and thousands of simulations on average before you get a Pedro.

It is definitely possible to get one with s=1, for example - you'll just be waiting for a couple of real-life years.

Now, is this realistic? Like I said, players like Pedro are rare - they aren't commonplace. A young winger whose career appears to be going nowhere will likely end up having a career that goes nowhere - the likes of Pedro are "unlikely" to begin with.

The reason I put the s parameter in is partly for this reason - to get an interesting graph in a reasonable amount of time, I had to make it more random.

I don't think it really matters though - what matters is the methodology, not how jagged the graph is. That a sum of several normal distributions can create lots of stable graphs and a few (important: rare) weird ones, that look plausible.

In some ways, you've proved that it is possible to have a very stable model - you needed to increase s to 30 before it started wildly going all over the place. So imagine if our model is modeled such that s is less than 30 - it is; in some ways, for your interests, it would be stable enough.

I already conceded in another thread that it could work in theory. What your models demonstrating is just how hard its going to be to build a balanced model. With s=10 you get a 60 point swing every 16732 players. 59 point swing more often 58 even more often etc you will see a lot well researched players deviate significantly from real life. If FM12 was to be released with unlimited potential implemented like that you would see the forums full of complaints about unrealistic player development. Why has

Reo-cocker won the Ballon d'Or? It's 2014 and Wayne Rooney is playing for Scunthorpe whats going on?

You could avoid that by tying development closely to age so well established players couldn't have big CA swings but wouldn't that prevent established players late blooming?

You could tie development tightly to performances but if the players CA is underestimated to begin with how will he perform well enough to develop?

Share this post


Link to post
Share on other sites

You again bring up Luca Toni and point out that the issue is to you not that players don't latebloom, but rather that he didn't. That doesn't make sense. Let's just boil this down:

You want:

- There to be latebloomers

- There to be the ability for most players to improve later in their careers

The Facts:

- Players do latebloom in the game

- Most players have a fairly decent stock of spare PA

However you now are in the position of now complaining that not everyone can latebloom. This now comes to an argument of the system rather than the result. You desire that any player who performs well will latebloom when sadly this is clearly not the case. You point to Pedro and Toni as examples of failure of the system, however as already covered (and I will get back to this) Pedro was just a youngster that was underrated and Toni falls in the catagory of a player who seemed to never have had the required ability. You Toni specifically as an example of the failure of the system, however that is more of a database issue than one of the model, as clearly shown by the abundance of latebloomers. Whether they are those that do in real life is irrelevent as even under your "model" they probably wouldn't have been the ones anyhow. On the point of players who played consistently well and never went on to be a "latebloomer" you only have to look at the premier league signings over the years. Countless players have come out of nothing in the championship and have come to the Premier League being expected to be far better players, however they seem to be nothing more than a flash in the pan and never play like they did again, even returning to lower leagues. You have your Rasiaks, Ellingtons, Howards and so forth. The English Leagues are littered with this type of player. Under your model they would seemingly all go on to become superstars when so many end up simply hitting the roof and falling off rapidly. Maybe you're conducting your argument the wrong way as in FM the only thing I'd argue is an issue is the fact that players generally don't drop off when they're young like Ellington, Ronaldinho and such.

You like to keep going on about Pedro and Luca Toni simply because they couldn't become better. I know you're frustrated that they themselves wouldn't have been able to in the game, but at the end of the day the model does get the right result as a whole, even if it isn't for the specific players. As already mentioned not even your version would do so. This issue is down to not being able to know their PA and the solution to the innacuracies would just be to chuck an error margin on at the start as I said because you're right, we can't know. I know you want it to be fully accurate, but that's not possible, so instead its better to put some random flavour in it.

As for the reason to take the median, do you think I'd be bothered averaging that many values. The median is nice as easy if you have a dataset. Also, 23-32 is because most players won't improve drastically after 23. Also, I think it was actually between 23-32 (that's 24-31) when I did it, but anyhow. I think I also did a 25-30 range as well, but I don't remember what it was. It wasn't less than 10 though and I have feeling it was 15, but I'm still not around my computer to test this.

At the end of the day you're never going to agree with me because it seems you've made this your mission on this forum. I feel sorry for you as I know that you're never going to enjoy the game to its fullest whilst the game is in the form it is now. Sadly for you the vast majority of people like the current system because of the reasons I've stated and as such your cries will hopefully fall on death ears. If SI did choose to go in your direction I would be extremely dissapointed but in the long run I'll trust their judgement on this one. This argument is getting tiresom and repetative and I think we should just agree to dissagree on this one as neither of us is likely to back down.

Also, for whitewolf, CA and performance isn't directly correlated. There is a relation where you'd expect that an increase in CA would increase performance, however x CA doesn't map to y performance and vice versa. He is right that you can have a player massively overperform, but it is also true that you can have players who simply by having a better distribution of their abilities plays much better than other players. An example of this would be my captain on FM10, one Simon Gillett, an English striker. I checked his CA/PA after I was already well into FM11 as part of the checks I did and it was 169/171. I had 3 other strikers with a higher CA/PA including Ben Bradbury with 185/186, Geoff Clark with 192/194 and Tony Talbot with 195/198. All 3 strikers were not as good as Gillett who would average about 8.4, score of the order of 3 every 2 games and set up a further 4 every 3 games. He was just a beast. He wasn't as talented as the others, but he was quick and had a nose for goal as well as a way about him in the air. Did he need to improve? No... Does Defoe need to be improved due to his form? No, because he plays like Defor with CA/PA of ~150ish. This is probably another issue that our good friend is having because he cares too much about CA/PA and not enough about performance. It is never about CA/PA, it's about performance. One of my best every centrebacks, multiple champions league winner, vice-captain, Francis Dickoh, had a CA of 115 and a PA of 118. Did it mean he was rubbish? No, he was fantastic, did that mean he needed to improve? He averaged over 7.5... At the end of the day the issue is down to how people see CA/PA, not what they are or how they work.

Have a good day guys!

Share this post


Link to post
Share on other sites

  • Recently Browsing   0 members

    No registered users viewing this page.

×
×
  • Create New...