The Uncertainty Principle & Dynamic Potentials(READ)

[RBKalle]

OK, I think I did misread it a little. He turned out to be an excellent midfielder despite having a PA of 130. I don't really see why that's not possible without PA. I'm guessing his attributes became more "extreme" - less-important attributes went down, and more-important attributes went up.

Actually I signed him for free when he was already 27, and he was already very close to his peak PA of 130, so his attributes didn't go up that much, except the mental ones.

I don't have a screenie of his profile from his heyday, so here's a screenshot of him anno 2011 in another save, so give or take the distribution of 5 CA points for a midfielder trained to be an all-around MC.

Nothing really impressive in continental context... still for a good 6 seasons he looked like Gerrard in my team.

However, CA is just a weighted average - it just creates a weighted total to make it easier to calculate players' abilities. Players with CA 130 are roughly on the same level, although of course they can have better-suited attributes with CA 130 (i.e. a player with 10 for everything is inferior to a player with 12 in important attributes and 6 in less-important attributes).

I would argue that if the player did develop in this way, then there exists a CA value such that the average ability of an "average" player is equivalent to your player, and that THAT CA value is the "fair" value of CA.

I suspect well-rounded players work better than "extreme" ones.

And a system without PA can of course still let your best players have low CAs.

How so?

If we allow the game to "freely" [ie. without preset boundaries] recalculate the Attributes (aka CA/HCA/PA... which are, as you explain, a weighed average of the whole attribute grid), then if a player performs well and is among the best of the team, he will always get a boost...

It looks like a loop... Good player plays fine --> attributes improved --> plays even better --> more bonuses etc...

I suppose I refer to my 10.00 rating example again - if he averages 10.00 every game, then he is the next Lionel Messi, no matter what his attitude or ability is.

Likewise, if you can make Sokolowski perform well despite a low CA, then he is the next big thing in the world of midfielders. No matter how untalented he seems to be.

The fact he performs well consistently suggests he will develop as a player. Perhaps (very) slowly if he's not that talented. But the human mind is an amazing thing - he will be absorbing information all the time, and as long as his body can physically keep up, he will keep developing.

Here we disagree...

I "made" Sokolowski, an adequate MC for Norwegian mid-top table acts, play extra well because the game allows me to do so, with the right tactic and the right settings...

But hadn't he had the PA set at 130, he would have kept on getting better and better and better. While in real life it can't happen. Because, like it or not, every player has a natural talent that he can't really "exceed".

Not as granitic as FM currently handles it, but still... Otherwise every kid who spent some time at a pro club academy should have at least made it in some capacity... And 27yo players who are ok at bottom EPL could hardly become Man Utd material two years of hard training later...

P.S. let's not forget FM gives us a whole lot of advantage, so what can work for AI players could be "disrupted" (or exploited) by human players

[Cougar2010]

Its late and I don't have the energy to cover all the old ground again but I do want to comment on two points:

"Potential" is an opinion. If a Blue Square North player moved to a Premier League side, his future would look brighter. We would revise how good we think we will become. I argue this is potential - and it changes.

Again PA gets revised for RL players version to version because its impossible for researchers to give an accurate figure (they don't have a crystal ball) and tbh what a researcher gives as PA for older players is more likely their opinion of that players maximum CA or current CA if the player is in decline.

- There exists some number representing peak ability for a person

- We don't know this value when this person is born

- Therefore a model of this person's life cannot generate a solid peak ability number at the start of a player's life

The game is not acting as a god. It has no real control over what a player can do, so it cannot "know". It just needs to create something to make the game easier.

We aren't talking about peaks, we are talking about PA and I find your comment about the game not acting as a god quite bizarre.

[x42bn6]

I suspect well-rounded players work better than "extreme" ones.

I suppose it depends how well-rounded...

One thing I've noticed about FM10 and onwards is that it's quite easy to raise small attributes.

For example, on another forum we did a silly experiment where we were able to create wonderkids but we got to pick the starting attributes. I picked "Emile Messi", a freak of nature. But the real point is that his "small" attributes (the 3s and 4s) became rather large quite quickly, while the 18s barely moved.

http://www.imagebam.com/image/07794377290382/

http://www.imagebam.com/image/545e3d77822033/

Which means I strongly suspect that smaller attributes are actually not that valuable, meaning well-rounded players may well be better.

But then again, that depends on how well-rounded...

How so?

If we allow the game to "freely" [ie. without preset boundaries] recalculate the Attributes (aka CA/HCA/PA... which are, as you explain, a weighed average of the whole attribute grid), then if a player performs well and is among the best of the team, he will always get a boost...

It looks like a loop... Good player plays fine --> attributes improved --> plays even better --> more bonuses etc...

Nah, that's where balancing comes in. All players suffer dips in form, for example, or injuries end runs of form. I'm not suggesting we hard-code these in, of course...

One idea is that we don't really learn things linearly - for example, I don't learn twice as much by playing twice as many games. It simply gets harder the more I try to learn. Some learn quicker.

Another idea is that we could develop attributes differently, at different rates. For example, we could develop worse attributes quicker with more playing time (if well-rounded players are worse), or better attributes quicker (if well-rounded players are better), so that lots of rapid development at early phases still doesn't let a player develop through the roof - the returns will be diminished.

If you like, the player has to continue to surprise you in order to get the juicy bonuses. If he is consistently brilliant and continues to exceed expectations, I would argue he is continuing to learn. There will, however, come a point where he simply can't keep surprising you - his form will taper off and he will be forced to revert to "solid" - at which point his development is forced to slow again.

And I will also argue that if he continues to surprise you and get better and better, then that's just that - he's getting better and better, and his CA rises will be nice.

Here we disagree...

I "made" Sokolowski, an adequate MC for Norwegian mid-top table acts, play extra well because the game allows me to do so, with the right tactic and the right settings...

But hadn't he had the PA set at 130, he would have kept on getting better and better and better. While in real life it can't happen. Because, like it or not, every player has a natural talent that he can't really "exceed".

Not as granitic as FM currently handles it, but still... Otherwise every kid who spent some time at a pro club academy should have at least made it in some capacity... And 27yo players who are ok at bottom EPL could hardly become Man Utd material two years of hard training later...

P.S. let's not forget FM gives us a whole lot of advantage, so what can work for AI players could be "disrupted" (or exploited) by human players

I suppose we will have to agree to disagree. But I believe the bold part is covered by some part above (i.e. players don't have to keep developing, or the rate at which they develop will slow down).

----

I'm going to take a break now, ladies and gents... My fingers hurt, and I want to get some FM done. I'll resume tomorrow!

[DaSilva9]

Perhaps what is needed is a way for the AI to judge players based on how their individual attributes fit in to certain role or positions? Judging players based on their CA is misleading because sometimes a player with lower CA will play better than a player with higher CA so perhaps instead of drastically changing the player development system SI just needs to implement a way of judging players that is more precise. Therefore those low CA players whose attributes are perfect for a certain position will have an opportunity to shine in the game environment though the problem with this system would be that it probably involves a heavier use of system resources.

[noikeee]

I agree with the thread starter. It would be very useful for the researchers that want to get precise, I mean I'm not a researcher but sometimes I've been editing the database looking for more realism and there's a player who's still getting started in top class football, his potential is a bit unknown yet but let's say I estimate it at around 125-145. Now there's 3 options:

a) set it at -7, which means that the player's PA will bounce through save games between 110 and 140. In half of the save-games he's going to be way too crap.

b) set it at -8, which means that the player's PA will bounce through save games between 130 and 160. In half of the save-games he's going to be way too good.

c) set it at an average of 135. That's unrealistic because we don't really know to that super-high degree of precision how good the player will be!

In fact I'd go so far ahead as to make it mandatory for the PA of every player to have at least 5 points each way of dynamic range. And a maximum of say, 25 (that'd let you create some players with very high ranges, ex. 50-100 PA for very unknown lower league youngsters). Let's see slightly more variety in savegames. Researchers can't see the future, their judgement of potential shouldn't be set in stone. However, it makes perfect sense for a players potential within the game to be set in stone as every player has their physical limits, very few people can become a Zidane even if they trained every second of their lives in perfect conditions. But it is extremely hard to judge exactly where is that limit - every single world class manager/scout in the world has made wrong calls regarding players' potential ability, why should some unpaid SI researchers know better than them?

As for how a player's ability develops through the game, and how the AI reacts to it, that's a different subject. I've just made a long post in another thread about one of the issues - it's far too easy to see PA within the game world. There's at least two other issues which I might take the time to explain one day:

a) the AI staff should be more aware of when the gap between a players CA and a players PA is so high that he may never be close to reach it

b) player development has gotten better over every version of FM, but there still aren't enough late bloomers. I guess a dynamic in-game PA would solve this, but I much prefer the idea of having a dynamic PA set by the researchers!

[Chris M]

No - but all that came to mind when I read it was Heisenberg compensators... which shows I am both a geek and a computer nerd.

That was what came to mind immediatly after.

We're so cool.

[Lankylars]

x42bn6, I fear that you will never see eye to eye with many of the rest of us here because of how you're choosing to define potential. You say that a player's PA is their maximum given current circumstances. Many of us are defining potential as the greatest possible maximum, taking into consideration all possible circumstances. Your "PA" is simply growth or fluctuations in CA. Your example earlier about a BSN player having a less-bright outlook than if they suddenly moved to Arsenal illustrates this. PA represents what they could achieve at Arsenal, if you will, whereas CA will vary depending on current environment. Sure, the highest CA value they might achieve in BSN is lower than in the Premier League, but their true PA doesn't take that into account - it's a global, not a local, maximum. Again, changes to the way players develop would alleviate this, not chucking PA out the window.

[aderow]

That's good - but what about players who exceed expectations? Neymar is allowed to fail expectations, but other players aren't allowed to succeed.

This is key imo.

Its very easy for a player not to reach his potential but in the current system you will never find a player who has, for example, played in the lower leagues and has made it up to the premier league and is at least decent in the premier league.

And like I said before this is probably because late bloomers are too difficult to code for.

[RBKalle]

This is key imo.

Its very easy for a player not to reach his potential but in the current system you will never find a player who has, for example, played in the lower leagues and has made it up to the premier league and is at least decent in the premier league.

And like I said before this is probably because late bloomers are too difficult to code for.

True, but I think we should stop focusing on CA/PA as if it was the one and only factor...

How many times have we seen other users complaining about Player X or Player Y not performing as well as they expected, despite said players having terrific attributes?

How many times have we found "hidden gems" who play much better than their attributes should allow them?

With the right tactic and with the right mental skills an otherwise average player can still be as good as a "better" one.

So we should just stop talking about CA/PA as if it was an univocal measurement unit for success.

Sure, the CA development model could use a tweak to allow for more mobility both down AND up, but under the right circumstances it's already possible having some sort of "late bloomers" and "from Championship to Champions League" stories.

Then again, it should be an exception, not a rule.

P.S. if CA/PA were so decisive and exclusive in determining success, we wouldn't ever be able to win domestic and continental trophies with mid-table clubs, let alone completing the various BSN/BSS to EPL challenges!

[Cougar2010]

True, but I think we should stop focusing on CA/PA as if it was the one and only factor...

I agree with that, many users seem too hung up on CA/PA and in many cases is due to them using 3rd party applications.

Without the use of 3rd party applications we only have attributes & stars - Thats the way the game was designed to be played and no where is CA/PA really mentioned.

[aderow]

True, but I think we should stop focusing on CA/PA as if it was the one and only factor...

I wasn't really focusing on CA/PA

I do agree with this statement though.

[SCIAG]

Its very easy for a player not to reach his potential but in the current system you will never find a player who has, for example, played in the lower leagues and has made it up to the premier league and is at least decent in the premier league.

I think never is a bit extreme. I remember on one version of FM, I downloaded Genie Scout and found a 26-year-old with 170 PA and about 50 CA playing in League 2. I saw the challenge, and signed him. I think I managed to get him to about CA 120 by the age of 30- he was certainly a much more useful player, and had better attributes.

I also once saw Danny Haynes languish in League 1 until his mid-twenties, and then sign for a recently promoted side (think Luke Varney). He was their top scorer iirc, or at least their main striker.

I agree with the thread starter. It would be very useful for the researchers that want to get precise, I mean I'm not a researcher but sometimes I've been editing the database looking for more realism and there's a player who's still getting started in top class football, his potential is a bit unknown yet but let's say I estimate it at around 125-145. Now there's 3 options:

a) set it at -7, which means that the player's PA will bounce through save games between 110 and 140. In half of the save-games he's going to be way too crap.

b) set it at -8, which means that the player's PA will bounce through save games between 130 and 160. In half of the save-games he's going to be way too good.

c) set it at an average of 135. That's unrealistic because we don't really know to that super-high degree of precision how good the player will be!

The point of the "-" ranges is that nobody knows how good young players are going to be. I reckon -7 would be good for the player you mention. Let's face it, about half the players of that level IRL (for every Jack Rodwell there's a Paul Huntington or Nigel Reo-Coker) don't reach the levels expected of them.

Regardless, 135 (or rather 145- I reckon you'd be better off going for the maximum you estimate rather than the median) is the maximum the player can achieve, not what he is expected to achieve or what he will achieve.

[aderow]

I think never is a bit extreme. I remember on one version of FM, I downloaded Genie Scout and found a 26-year-old with 170 PA and about 50 CA playing in League 2. I saw the challenge, and signed him. I think I managed to get him to about CA 120 by the age of 30- he was certainly a much more useful player, and had better attributes.

I also once saw Danny Haynes languish in League 1 until his mid-twenties, and then sign for a recently promoted side (think Luke Varney). He was their top scorer iirc, or at least their main striker.

You are right. Never is probably a strong word.

I based that on my personal experiences with FM. I personally have never seen that sort of player in any of my saves...then again I wasn't exactly looking for them.

I also probably shouldn't doubt that it has happened for anyone else since it hasn't happened to me.

The thing is without using genie scout, you probably would not have signed that player because your scouts would likely have advised you not to.

[x42bn6]

x42bn6, I fear that you will never see eye to eye with many of the rest of us here because of how you're choosing to define potential. You say that a player's PA is their maximum given current circumstances. Many of us are defining potential as the greatest possible maximum, taking into consideration all possible circumstances. Your "PA" is simply growth or fluctuations in CA. Your example earlier about a BSN player having a less-bright outlook than if they suddenly moved to Arsenal illustrates this. PA represents what they could achieve at Arsenal, if you will, whereas CA will vary depending on current environment. Sure, the highest CA value they might achieve in BSN is lower than in the Premier League, but their true PA doesn't take that into account - it's a global, not a local, maximum. Again, changes to the way players develop would alleviate this, not chucking PA out the window.

I am aware that PA is defined given all possible circumstances (realistically, though, it's not "all" - it's more like a 99% thing because "never" is too strong a word), but the problem is that we make this decision based on information we have "now".

How many times have we seen a player and a few years later, thought, "Wow, I never thought he'd be there?" And this refers to big jumps and big falls.

That's the real issue I have - the maximum can't be predicted. Never is too strong a word for something that is uncertain. Relaxing the "never" is more realistic.

Re this:

PA represents what they could achieve at Arsenal

The problem with the current PA system is that this is simply not true. PA for a player moving from BSN to Arsenal refers to the PA of a youngster who started at BSN level - not one that started at BSN level but moved to Arsenal. I see these numbers as different, because of the opening doors principle. If his PA is high then the fact he moves to Arsenal somewhat compensates for this - but what if it's not? See the above paragraphs.

Think about it - when you watch academy and reserves matches, why do you think X has tons of potential and why Y doesn't? We see their talent. But what does it imply for X to have lots of potential? If "X has a lot of potential", do you read that sentence as "cor blimey, that kid is only going to be as good as Rooney at his best", or "whoa, that kid is probably going to be a world-class striker at his best"? It's the latter!

And here's the issue - PA is the first sentence, while my model is the latter. When you "put a yardstick down", we allow ourselves to allow for the fact we might be wrong - in both directions. We know we will probably be wrong, so we cannot be definitive in our statements. And "only going to be as good as Rooney as his best" is partly-definitive - so it must be wrong.

To me, the only "definitive" statement is that we cannot be better than the ultimate football being in the world. Think a player with 20 in every single attribute (or 1 where appropriate, i.e. controversy). But that is bad enough to define because a) we can't really pick out a CA for that person, b) we can't envision it and c) it's probably never going to occur anyway. Either way a stronger statement is that "CA will never exceed 200" - to me, that is the only definite statement that can be made.

It does imply PA = 200 for every single player - which is why I keep hammering home the fact that PA is wrong - it implies nonsense.

[faraow]

...

PA is not wrong. The fact that you can't see nor correctly estimate PA does not mean that it doesn't exist. If you say that there is no such thing as PA, it implies that everybody has got the potential to reach the same level. You say that everybody has a PA of 200, that implies that all variance in CA is determined by environment and absolutely zero variance can be explained by genetic and/or bodiliy disposition. That hasn't been found for any single human characteristic. That is an incredible statement. Given the fact that there is huge variance in brain and body among human beings, there have to be different maximum potentials that people can reach. I can never be as good as Messi, because my brains are not as quick as his and my body is different. Ofcourse, brains are influenced by the environment and you can train your body. But that only means I could have been a better player than I am now, not that I could have been as good as Messi.

I think the problem in the game is that scouts can see potential ability. Nobody can see potential ability! You can only estimate the current ability of someone and compare it to kids of the same age. If you take the level at which they are playing and the speed of learning into account, you could estimate a future potential ability. But I think in the game the scouts "read" the potential ability from the editor and they are more or less correct dependent on the "judging potential ability" attribute. Take this striking fact: in Holland you are more likely to be scouted by a professional club if you are born in January than in December. That is, because kids from the same year of birth play in the same team, but the kids born in January are older, so they are bigger, quicker, motorically better. The kids from december aren't noticed, because they don't got the same level yet. You see it is very hard to estimate potential ability. The best we can do is compare to people of the same age, experience and league. The scouts in the game should do this to.

I do agree that the environment does play a huge role. Every third division player will get better if he plays at Arsenal. I don't think it is necessary to programme a maximum PA level into the game, but there has to be an influence of talent. Maybe learning curves, that are affected by playing level, coaches etc. But in general, the most important thing is that the level of uncertainty should be way higher. There will be a player somewhere in the non-leagues that will develop to a championship player. We don't know which one, but the game could be programmed that on average a small % of all low league players will have high PA/talent. This % should get higher in the premier league, but there is no necessity to pin down specific players.

[x42bn6]

If you had read my posts you will have found I don't disagree that a player has a limit. What I deny is the ability to determine it from the start of a player's career.

We can factor in genetics via "talent" in determining how a player develops - talented players find it easier to develop and may find it easier to do well at higher levels.

What we can estimate is where we "roughly" think they will end up - but we cannot determine with absolute certainty - as static PA implies - the ultimate maximum.

So yes, we have a maximum, but no, it makes no sense to determine what that should be at the start of his career.

The fact we have a limit is modeled without PA - because the game will of course be balanced to ensure that every player does not hit CA 200.

[faraow]

I must admit that I haven't read all the posts in the topic. But the way you put it here: agree totally . Now only hope SI reads this and improves the whole development and scouting area.

[BAD-RELIGION]

People who, like me, like to play many seasons, are finding the game a bit boring. The oldest -10 and -9 was way to better and could provide much better players. Now you have a -9 that go from 150 to 170 and a -10 that go from 170 to 200, previews we had a -9 that could go from 150 to 200 and a -10 that go from 170 to 200. every negative value could generate a 200 potential ability player, even a player with zero (0) could have a great potential when a new save started. That leveled up the players and we had a much more random game. Now what happens is, independently from what save you have, after 4 or 5 years the best players are always the same players...

Has there is less good players, and every one knows that high PA is not the only factor for a player to reach a great level the game is much leveled down. They need a good combination of determination, ambition and profissionalism to combine with the high PA, a player could have 200 PA but if they have <10 profissionalism he never get not even near their PA...

[SFraser]

No. The game doesn't really care about the future.... It's mostly concerned about the present.

Bingo.

The whole point is that real players at game start correspond to what people think at the time. Presenting a "realistic" game at game start is key to FM. Wonderkids turn out to be awesome, duds tend to be under-rated, Joe Average never goes anywhere.

That is what most people think at the time. It makes for a realistic football atmosphere for a couple of months.

Once the newgen/regen system comes in to the play the whole debate is moot other than for the question as to whether moving to a better club should raise the PA bar. It's a fantasy land, a complete lottery, and young players 5 seasons later in FM never turn out the same as young players 5 years later in real life.

So the question is "should Jack Whilshire in 5 FM years be like what people expect now, or should he be something different?"

Never underestimate the power of inaccurate prediction and unrealistic player development to sell the next game in the series.

[MelbVictory]

The problem with this is that you can see PA or CA with 3rd party editors.

If you never saw these numbers, which is how the game is designed, this wouldn't be even thought of.

Don't like the idea for that reason. When I play the game, I cannot tell you what exact CA a player is. That essentially is what you are proposing - some variability.

[jayahr]

While being a firm advocate of a fixed PA system, as far as one save is concerned, I really like and support the proposal of having a certain degree of uncertainty in these values.

Maybe not +/- x points, but +/- x% of the CA and PA, but that's details.

I think that the nature of the PA concept requires it to be fixed at the start of each game. If we change that, we may as well get rid of it, but in my view that would be a bad thing as firstly I believe that every player does indeed have a certain level of talent he cannot exceed no matter how much or well he is trained and secondly the PA concept makes it a lot easier to balance the game properly which in the end is what makes it playable.

But the suggestion in the OP is good, because it make each save a little more random and therefore less repetitive. Can't see anything wrong with that

[x42bn6]

The problem with this is that you can see PA or CA with 3rd party editors.

If you never saw these numbers, which is how the game is designed, this wouldn't be even thought of.

If we can't see X, then does that mean there is no problem with X?

[whilewolf2]

In response to OP

Just add more negative values -11 to -31 with a ten point spread for each. Some times you have a good idea of the level a player will end up playing at but the range offered by the current -1 to -10 values are too wide.

To x42bn6 the one upside of your proposal is the possibility of researched players exceeding the expectation of the researcher. While that would be nice it does seem to be a fairly small upside for what looks like a large cost in system resources used to process it. Some time in the future maybe but for now there are more pressing issues around CA/PA id like SI to look at.

Stop the scouting module peeking at PA instead make it look at CA and the direction and speed its moving at.

Don’t tie CA development to age so closely so late developers are more plentiful and able to make more dramatic improvements.

Or allow the character attributes to change overtime so Ambition, professionalism etc go up and down that would allow more varied career paths for players.

[MelbVictory]

If we can't see X, then does that mean there is no problem with X?

Of course, the system might be flawed.

But once you start looking under the hood, obviously you will find an oil leak here or there.

If you hadn't looked under the hood, you would be happily oblivious, and thus, playing the game no problems.

[TacticalGenius]

Of course, the system might be flawed.

But once you start looking under the hood, obviously you will find an oil leak here or there.

If you hadn't looked under the hood, you would be happily oblivious, and thus, playing the game no problems.

This.

Anyway, I think there is more a problem with PA ceilings for lower league players not being able to prove themselves at a higher level, rather than Neymar 'always' reaching his ability under the right circumstances.

[x42bn6]

Of course, the system might be flawed.

But once you start looking under the hood, obviously you will find an oil leak here or there.

If you hadn't looked under the hood, you would be happily oblivious, and thus, playing the game no problems.

Surely if the you're not meant to see is flawed, it should be fixed, whether the end-user sees it or not?

[Cougar2010]

Surely if the you're not meant to see is flawed, it should be fixed, whether the end-user sees it or not?

The point is the system isn't flawed. It works fairly well and just needs tweaking a little.

I know you won't agree with that x42 as you have your own ideas though.

[Lankylars]

Ugh, not this discussion again...