Player Development is a sham

[milnerpoint]

The same Jeffers regularly derided as a flop and Wenger constantly teased by it by opposition fans? Good performances? Good grief. There is a reason he is mentioned in this thread and that is because he never lived up to his promising appearances as a youngster.

his performances for Everton before he moved to Arsenal were very good, good enough to make Wenger spend £10m on him, and good enough for people to believe he was the real deal, which he wasnt, clearly.

As for the part where you mentioned "Arsenal would not have spent £10m on him", that is probably the worst argument ever put forward - replace Jeffers with Verón, Arsenal with Manchester United and £10m with £28.1m. Nobody is going to use Verón's price tag to justify his "good performances" - those were very few and far between!

Who said the Veron signing was good for Man Utd? His performances before signing for them were world class and he was regarded as one of the best in the world, he simply did not adapt well to english football, different senario completely.

He got a few injuries and couldn't displace Henry and Wiltord, didn't he? One of the key requirements for development is that you play first-team football and that you perform - the fact he couldn't get a game ahead of them suggests he was never effective enough to do so. Therefore he misses at least one of these requirements, meaning his development will stagnate. To me, this is the real reason why Jeffers is considered a flop - he couldn't get a lot of first-team football at Arsenal and when he did, he failed to perform. And then his barren streak continued as he fell back down the ranks, reducing his chances of bouncing back up. And to me, this is a sensible explanation on why perhaps Jeffers stopped developing. Absolutely nothing to do with his "potential" or anything.

completely anecdotal evidence as he did flop and never progessed he peaked early and never got any better, even after moving and getting first team football he still flopped.

The game will know the maximum scoreline of an Arsenal-Manchester United match, let's say it's 30-30 as in no team can score more than 30 goals. Sounds reasonable. Then say two human beings decide to rig the game and Arsenal play 0-0-0-5-5 with Almunia in goal. At 0-30, the Manchester United user quickly finds that Almunia cannot stop saving goals despite the fact he has all the wrong instructions too. So is the computer right here? Of course it's right - because it is impossible to be wrong, assuming the computer is all-knowing. However, to me, this looks almost like a proof by contradiction... Assume the computer is all-knowing, construct a theoretical scenario, conclude with a ridiculous conclusion... Therefore the assumption is wrong.

A completely nonsese situation, anyway you cannot start without a goalkeeper, so your proving nothing there at all. Anyway whos to say a team cannot score more than 30 goals if that ridiculous situation did occur, why would there have to be a limit at all on the score line? Talent is limited, what your describing is unrelated.

The game cannot know the future - it simply doesn't. In using PA, it immediately folds all future information into this value, despite the fact the future circumstances can vary quite dramatically to what was folded inside in the first place.

the game knows that with the right conditions player x can be x as good, not about knowing the future.

Oddly enough, researchers can be wrong - and the game should be able to compensate for this.

it does, by Si updating the database 2-3 times a year.

He didn't stop developing because he hit his potential - he stopped developing because he suffered injuries, lost his place in the team and didn't perform. You could argue that this is "potential". Let's say for argument's sake that it was PA 120.

But what if Jeffers succeeded? What if he dethroned Henry and became and Arsenal legend, scoring goals galore and making Henry wet himself and fall into mid-table obscurity in Italy because he couldn't get a game? You wouldn't think his PA was 120 - quite frankly it should be higher.

What does this mean? It simply means that how Jeffers developed at Arsenal should have an impact on his PA. Now this is fairly controversial, since the game is supposedly all-knowing... However, this is exactly what I mean when the game throws up scenarios that are unexpected. The game can't know the future - because it quite frankly can't. Therefore it should stop pretending to know, and focus on doing things - it is more realistic.

again completely anecdotal because he never made it proper at the top. But in that senarion his profile again would have been update on regular basis, but again a pointless argument,

What if I use the editor? The inbuilt one.

you cannot use the in game editor to check on the PA of regens.

Covered above - that it is wrong to assume the game cannot be wrong, since no system is correct. The system can - and is - wrong. So is my suggestion. It just comes down to whether SI should move to a better system or not.

Who says your system is better, that is a rather arrogant statement to make. You are yet to convince most people on here and more importantly any of the dev's that what your proposing is any better than what they are using at the moment, when it all boils down all you are proposing is removing one limit to replace it with another, with no real benefit except in the very few extream occasions that might happen once every 10,000 regen players created.

[sjm]

A lot of missing the point here. The reason players have a PA is to ensure a good spread of player abilities in the game.

If every player could become Lionel Messi, and suddenly every player did become Lionel Messi, the game would suck.

That's really all there is too it - it's purely for balancing purposes, and really doesn't matter. It only matters if you look at PA, don't understand why it's there or how it works and then get hung up on it.

I'm sure there are numerous other ways of balancing the game, but the net result is the same, and the PA approach is computationally cheap, so why go for an overly complicated bug-prone super algorithm when you can just use a little bit of basic statistics, a bell curve and be done?

Edit: and the issue of disagreeing with researchers (= players that already exist vs regens) is by the by, as you can just edit these yourself if you don't agree, so it's beside the point anyway. Youth development is all about how good the regens are anyway.

[DiamondFlair]

PA is needed if for no other reason the game balance. If every player had the same potential in the game, nearly every human player would find signing players pointless. Since any of your youths is capable of being the next Messi with the correct management, you never need to sign a player again. A good quarter of the game is totally pointless after season one.

A similar problem would occur with the PC teams. Managing to get the balance right when each players potential is unlimited would be almost impossible making the game either boringly easier or far to difficult to be fun. The players actions are not enough to impact development because you CAN alway do the right thing to get the right result, whereas in real life even the best youth managers in the world have sometimes failed to be able to turn a player that clearly has talent into a star simply because they went of the rails in spite of everything. The only random factor would be injuries.

PA is in no way ideal, but i unfortunately feel its the best option.

[shwan]

Everyone has limits.

Every year the likes of Man Utd, Arsenal, Chelsea, Liverpool etc release players - Why do you think this is?

disagree, they release them not because the have reached thier limits at 18 or 20 !! are you joking !! CR was been called just a selfish dribbler at 20 !! some player improve faster, some will be just good players, and that is not enough for big club.

in any way, all of them develop but at different rate , to stop them and say sorry they reach their limit is ridiculous concept, SI need to replace PA bu PDR (potential development rate ) to stop it completely is just unbelievable, players will still improve in their passing, strength, mentally ..etc. they all improve but at different rate.. to put cap on it !! is ridiculous

[SFraser]

Covered above - that it is wrong to assume the game cannot be wrong, since no system is correct. The system can - and is - wrong. So is my suggestion. It just comes down to whether SI should move to a better system or not.

I don't think Godels Incompleteness Theorem applies here.

The question is if you ignore the system and focus only on the results that occur ingame, what is the difference between FM and real life?

I would argue that there is very little difference, players have a reasonably obscure potential that may or may not be accurate and can peak out at a whole manner of unexpected levels. Is this similar to what happens in real life? Does this need more accurate simulation? I think that in practical terms it works. There are tweaks that could be considered, possibly, but overall it does the job it's supposed to do.

[x42bn6]

his performances for Everton before he moved to Arsenal were very good, good enough to make Wenger spend £10m on him, and good enough for people to believe he was the real deal, which he wasnt, clearly.

Yes, but you said that Jeffers was good for Arsenal (otherwise Wenger wouldn't have spent that amount of money on him)!

Who said the Veron signing was good for Man Utd? His performances before signing for them were world class and he was regarded as one of the best in the world, he simply did not adapt well to english football, different senario completely.

Exactly, and I used this nonsensical example to show your logic was incorrect (proof by contradiction).

completely anecdotal evidence as he did flop and never progessed he peaked early and never got any better, even after moving and getting first team football he still flopped.

It's not personal experience, so it's not anecdotal.

The ages of 16-22 are rather important for development and Jeffers went on a downward spiral never to recover. He peaked early - I am interested why he peaked early - since the game models peaks, a more accurate game would model how peaks occur.

A completely nonsese situation, anyway you cannot start without a goalkeeper, so your proving nothing there at all. Anyway whos to say a team cannot score more than 30 goals if that ridiculous situation did occur, why would there have to be a limit at all on the score line? Talent is limited, what your describing is unrelated.

0-0-0-5-5 is all players in the top 5 lines of attack - and one goalkeeper.

By arguments of the "God" idea, it is possible to impose a "maximum scoreline" idea. PA is just basically the maximum-envisioned potential ability of a player - "maximum scoreline" is the maximum-envisioned scoreline of a game. To balance the game engine, you could argue that "maximum scoreline" is possible - imagine a researcher researching scores instead, envisioning that Arsenal-Manchester United never exceeds 30-30 - a reasonable assumption, given no team in the top division has scored 30 goals before. It also helps balance the game - also a plus. The issue is that in limiting yourself to 30-30, you put an assumption based on previous information - and it is more accurate to not use that assumption.

We all know a game has a limit - the final score for one thing. But it is possible to envision that the game should have potentially finished 0-3 instead of 0-1, for example - arguably missing the maximum scoreline.

That you find this nonsensical is exactly the point - why is maximum scoreline considered nonsensical, but PA is not?

the game knows that with the right conditions player x can be x as good, not about knowing the future.

The game knows, but because the game can simply not be wrong, not down to some incredibly-gifted thought process in the design of the game.

A similar argument would be me hard-coding into my racing game the top speed of the F1 car to be 20 mph - by this argument, you can't say the game is wrong, because my game is "God" and knows that with the right conditions the F1 car can have a maximum speed of 20 mph, not knowing the future. In other words, that the game cannot be wrong is not down to it being correct - it is forced. The game cannot be wrong under this assumption (call it A) of a pre-defined limit - but the assumption A is wrong to begin with.

it does, by Si updating the database 2-3 times a year.

That doesn't help the game as it stands. When a game starts, what happens outside does not affect it.

Why should we not have some method of reproducing the researcher process in-game?

We all know researchers cannot be correct, as the future cannot be predicted. So surely the game would be even better if some process to "fix" the researchers' data over time, within the game itself, could be better?

Yes, in two different games, a player might turn out differently - but then again, that is exactly the crux of the issue, where a player's development and peak depends on how he does over time.

again completely anecdotal because he never made it proper at the top. But in that senarion his profile again would have been update on regular basis, but again a pointless argument,

Read up on what anecdotal means...

The aim of the game is to reproduce realistically reality into the game. In other words, if, say, Jeffers after moving to Arsenal performed exactly as he did in-game as he did in reality, the game and reality should match identically. In other words, say Jeffers' "average rating" matched in reality and in-game, and the "average rating" mechanism matched in reality and in-game (and so on and so forth), then Jeffers should flop in-game in exactly the same way as he did in reality.

However, this implies the opposite - if Jeffers indeed succeeded at Arsenal, perhaps dethroning Henry as an Arsenal legend, then Jeffers would have developed very well indeed, into perhaps a world-class striker or something. However, this should be able to be replicated in-game - perhaps CA 170, for example, depending on how CA 170 corresponds to reality.

This is what I mean by "the peak depends on how you develop in the future" - you cannot lock yourself into a value at the start of generation in-game - it can go up and down.

you cannot use the in game editor to check on the PA of regens.

No, but PA matters to me, in that sense (the quote was: "If there was no such thing as the scout tools, no-one would care about PA.").

Who says your system is better, that is a rather arrogant statement to make. You are yet to convince most people on here and more importantly any of the dev's that what your proposing is any better than what they are using at the moment, when it all boils down all you are proposing is removing one limit to replace it with another, with no real benefit except in the very few extream occasions that might happen once every 10,000 regen players created.

I personally think it is better, but mostly because I haven't been able to get any meaningful feedback on my proposed system anyway.

Riz Remes has already said the current system isn't 100% perfect but my system hasn't been commented upon. I am waiting for people to say things about it!

My system doesn't impose a limit per se - but it does impose an assumption that player grown on average is parabolic, something that can be relaxed to make it more realistic (i.e. a skewed parabola, or a non-parametric system). But there are no limits - the normal distribution is unbounded. If the first number is a number in the millions, then that is possible - as is the billions, trillions and so on, resulting in ridiculous growth. The good news is that if you care about this happening, then you may as well stop taking medicine - you hold yourself higher to statistical testing than clinical trials.

And please don't pull statistics out of your mouth - it will affect a lot more than a very few extreme occasions. It will affect every player because it is a new development model. Is it better? Well then, please comment. I've already explained why "a limit is one thing, specifically defining a limit is another" a million times.

A lot of missing the point here. The reason players have a PA is to ensure a good spread of player abilities in the game.

If every player could become Lionel Messi, and suddenly every player did become Lionel Messi, the game would suck.

That's really all there is too it - it's purely for balancing purposes, and really doesn't matter. It only matters if you look at PA, don't understand why it's there or how it works and then get hung up on it.

False dilemma much? What if the PA system is actually flawed? Surely that is a possibility?

The system will of course be balanced because the probability of every player becoming Messi is ridiculously small, and this needs to be statistically negligible too in-game.

Yes, I agree PA is for balancing purposes - why balance a complicated feature when a simple one will do? However, by making a feature simple, you introduce assumptions, which are less accurate. If SI are striving for accuracy, then they will need to move away from PA as a balancing limit. With Moore's Law, computationally it won't matter soon enough anyway.

I'm sure there are numerous other ways of balancing the game, but the net result is the same, and the PA approach is computationally cheap, so why go for an overly complicated bug-prone super algorithm when you can just use a little bit of basic statistics, a bell curve and be done?

If it's bug-prone, then it's badly designed. I'm not expecting SI to release badly-designed software as a new feature like that. It will be balanced.

Edit: and the issue of disagreeing with researchers (= players that already exist vs regens) is by the by, as you can just edit these yourself if you don't agree, so it's beside the point anyway. Youth development is all about how good the regens are anyway.

Not sure I agree with that though... I do not want to introduce my own bias into editing players.

As it stands, you yourself as a manager have little influence in developing players to their maximum, since every player has a limit. The world's best managers would develop the same set of youngsters to the same maxima. I personally do not think this is correct, since some managers will be able to get even more out of their players, even if they are limited - Keane and Fletcher spring to mind, for example - perhaps not the most talented of players in the world, but make up for it with excellent attitudes and mentality.

I don't think Godels Incompleteness Theorem applies here.

Gödel applies because you cannot use the FM engine to justify the FM engine. Therefore to say the game is "correct" because the game is "God" is fallacious - the game cannot justify itself. The game must be judged based on external metrics and reasoning. Consequently this means that the game can arguably never be correct, because any game consisting of any number of rules can still not be justified to be correct, because it can't judge itself. In terms of software design and modelling, this means that no piece of software is correct.

The question is if you ignore the system and focus only on the results that occur ingame, what is the difference between FM and real life?

The realism of a game follows two logical phrases - if X implies Y in-game, then X implies Y in reality; and if X implies Y in reality, then X implies Y in-game. If you like, if X happens, whether it be in-game and in reality, then Y should happen and Y should be equivalent for in-game and in reality situations. In other words, there should be a 1:1 mapping between all causes and conclusions in-game and in reality.

Of course, Gödel suggests this is impossible, but that just suggests that the more rules the game satisfies that match reality, the more accurate the game is.

Go back to your point. If something happens in-game, then there will exist some scenario in reality that matches the game. For example, if a player averages 10.00 in-game, one possibility is that this player scores 6 goals and gets 4 assists per match all the time. Another example - if a player suffers a cruciate ligament injury in-game, the real-life example is that the player suffers exactly the same injury and is out for the same amount of time, and suffers the same levels of post-recovery deterioration, and so on.

So there really is no difference between the game and real life in an ideal situation.

I just think that PA contradicts this in some circumstance - a player who is 50/50 and averaging 10.00 in the Premier League (rare, I know, but say it happens) will develop, full-stop, in reality; but won't in-game. Less-extreme examples of course are possible, like the OP at 150/150 rather than 145/150, for example.

The reason is that PA does not take into account all the information that is gained as a result of the player developing - development implies peak changes, in the same way that a theoretical Jeffers who did sustain his youth performances for the Arsenal first-team would have turned out better, since he would have actually developed during his tender years, and at a very high level indeed (i.e. Arsenal).

I would argue that there is very little difference, players have a reasonably obscure potential that may or may not be accurate and can peak out at a whole manner of unexpected levels. Is this similar to what happens in real life? Does this need more accurate simulation? I think that in practical terms it works. There are tweaks that could be considered, possibly, but overall it does the job it's supposed to do.

The examples I cite are:

- Pedro: In FM07, he had a low PA - nobody thought he would turn out to be as he did today. However, going by the realism argument, if Pedro broke into the Barcelona first-team in exactly the same way as he did in reality, and performed exactly in the same way as he did in reality, he should develop in exactly the same way as reality. The reason why PA fails here is because the PA assigned to Pedro was based on the fact he was languishing in Barcelona's reserve teams with little prospect of developing further. It was more than likely that Pedro would continue to receive the same treatment until eventually Barcelona released or sold him. What the game didn't take into account was the fact that the bit in italics is an assumption that might possibly not hold - in reality, Pep threw Pedro into the fire, and he performed. The idea is simple - sometimes, the assumptions made in generating PA are not appropriate because the future cannot be foretold. Once the reality diverges from these assumptions, PA is wrong - and sometimes, these assumptions might be wrong in a positive way - i.e. they were too pessimistic.

- Luca Toni: Toni was a mid-table striker who was rather mediocre, but suddenly out of the blue performed very, very well on a long-term basis, sparking arguably a late-bloomer profile, despite the fact he should have peaked beforehand. Toni's PA was based on the fact he would continue to become a mediocre mid-table striker - the moment those assumptions didn't hold, PA should have moved - and in this case, it should have gone up. The same argument applies for many late-bloomers, such as Drogba.

Practically, yes, it works for the majority of players, but not all of them - Gödel implies it won't be possible for all cases anyway, but we should strive for as many as possible. I think the OP's case is another one - a youngster who is performing well, peaking at a young age? A youngster who is performing well in reality will continue to develop - perhaps slowly - he will stop developing when his natural development and other development can no longer sustain the level required to actually develop (i.e. if there is no reason to develop, then don't - and this reason becomes harder to satisfy as you get older in the same way that a youngster develops quicker than an adult). His actual potential doesn't really matter - potential is a wishy-washy argument that nobody can define nor know as it is based on the future and various what-ifs. What we do know is that some players develop quicker than others, and a less wishy-washy argument is "talent" and no PA. An even less wishy-washy argument would break down talent even further. And so on and so forth. It is more realistic this way.

There are a lot of pluses to removing PA, anyway. Removing PA will also remove CA, and will allow manager-based ratings of players. Removing CA removes attribute constraints and therefore players like Walcott no longer need to suffer from having lower physical attributes or have their other attributes cut because they are "extreme" players. Removing PA and introducing talent means that you the manager can have a much better say in how a player peaks and it is fully dependent on how you develop them - as it stands, all top managers would develop this player to the same level every time (i.e. PA is reached) - something I do not think is accurate since, like I have said, how a player peaks is dependent on how he develops, in addition to how he is right now.

[sjm]

False dilemma much? What if the PA system is actually flawed? Surely that is a possibility?

The system will of course be balanced because the probability of every player becoming Messi is ridiculously small, and this needs to be statistically negligible too in-game.

Yes, I agree PA is for balancing purposes - why balance a complicated feature when a simple one will do? However, by making a feature simple, you introduce assumptions, which are less accurate. If SI are striving for accuracy, then they will need to move away from PA as a balancing limit. With Moore's Law, computationally it won't matter soon enough anyway.

I've left the rest for others to reply to, but if you've ever been involved in anything complicated, like for example designing software, you will understand why a simple system is desirable. This has nothing to do with Moore's Law and everything to do with software design. Complexity is difficult to debug and maintain, so there's no point in introducing it for little/no benefit. I don't want to sound harsh, but you really have no idea about the practical deadline-driven day to day life of software design.

The whole idea of PA is just to keep the level of players more or less constant. Sure, the player development still isn't perfect, but that's not because of PA. There really would be no benefit to essentially simulating the current method with a more complicated method. Whether you apply the statistical leveling a priori or later on doesn't make any difference from the outside, but makes a big difference in terms of efficient design, computational overheads and code maintainability.

[isignedupfornorealreason]

I can see both sides of the argument. Mind you I do wonder if it's possible to code in some elements similar to Pro Evolution Soccer's development graphs. That is; could a random number be triggered that would change the 'growth spurt' of a player. At the moment it's 15-23 years of age then growth either goes very slowly or not at all (right?). So having a RNG select between three variables that would shift the goalposts to allow for either, early-development, standard development and late-bloomers surely shouldn't be too difficult?

I'd also be impressed if PA was coded in to be variable based on perfomances. So someone reaching their PA cap should be able to develop more if say, they reach that cap before they've finished their 'growth' peroid and provided they're getting good first team run outs (and depending on their level obviously...)

Whatever the case though I have the feeling only human managers would be able to take advantage of it. AI managers tend to be shocking at developing youngsters in my experience... =/

[Lazaru5]

I can see both sides of the argument. Mind you I do wonder if it's possible to code in some elements similar to Pro Evolution Soccer's development graphs. That is; could a random number be triggered that would change the 'growth spurt' of a player. At the moment it's 15-23 years of age then growth either goes very slowly or not at all (right?). So having a RNG select between three variables that would shift the goalposts to allow for either, early-development, standard development and late-bloomers surely shouldn't be too difficult?

I'd also be impressed if PA was coded in to be variable based on perfomances. So someone reaching their PA cap should be able to develop more if say, they reach that cap before they've finished their 'growth' peroid and provided they're getting good first team run outs (and depending on their level obviously...)

Whatever the case though I have the feeling only human managers would be able to take advantage of it. AI managers tend to be shocking at developing youngsters in my experience... =/

Now this idea I like!

Two different things, a PA (i.e. a fixed limit) and a Growth Period (where a player will continue to develop regardless or beyond his PA). I'm sure that something like this could be introduced at some point, where the player's age limits his development rather than a superficial numerical value (that may or may not be wrong anyway). This is probably the best suggestion yet as to how to allow players to develop beyond set limits.

[Avelives]

Now this idea I like!

Two different things, a PA (i.e. a fixed limit) and a Growth Period (where a player will continue to develop regardless or beyond his PA). I'm sure that something like this could be introduced at some point, where the player's age limits his development rather than a superficial numerical value (that may or may not be wrong anyway). This is probably the best suggestion yet as to how to allow players to develop beyond set limits.

Good suggestion as it is, I can say from my own experience as a long time Pro Evo fanboy that it quickly becomes a very predictable and easy to manipulate system for humans and (as hinted at) not for the AI who end up woefully lost and with teams of utter crud.

In fairness I think the AI limitations is probably the main reason for the somewhat static PA system, otherwise it wouldnt be fair. As humans we can adapt to work with any randomness, the AI is coded to respond and cant be coded to respond to random systems with any accuracy.