Current Ability and Atrributes Research

[trebor1185]

Originally posted by Hawshiels:

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by trebor1185:

Just had another look at this and have a few questions, have you classed ML and MR as attackers, and crossing is an important attribute for full backs and does take up CA points from what I can see.

There's other ones such as wing backs and sweepers who have different important attributes to other defenders, or are you only concentrating on the main positions.

Sorry, I should have been clearer about this. I am only taking the central positions for midfielders because they work differently from the defenders and attackers.

For example, a midfielder on either the right or left gets a 'FREE' score within 'positioning' to make up for the loss in the free midfielder 'FREE' score of 'crossing' (which of course a right or left-sided midfielder would be charged for). I will post all of these though. I just thought I had a simple way of taking people through the process and introduce the complexities one stage at a time. </div></BLOCKQUOTE>

Oh right that kind of makes sense now

[Hawshiels]

And now to the differences in values between the 3 attribute types - TATT, MATT and PATT.

I am giving the most simplified fractions (ratios) here rather than producing lengthy decimals for each.

So ...

One TECHNICAL-ATTRIBUTE-POINT = One and eleven thirty sevenths of a MENTAL-ATTRIBUTE-POINT (roughly one and a third)

One MENTAL-ATTRIBUTE-POINT = seventy five ninety sixths of a PHYSICAL-ATTRIBUTE-POINT

One PHYSICAL-ATTRIBUTE-POINT = almost exactly one TECHNICAL-ATTRIBUTE-POINT. [A technical attribute point is worth very slightly more, but not enough to be noticeable]

[trebor1185]

Forgot to say in my last post, your not looking at full backs yet?

[Hawshiels]

I showed earlier the differences in the main central areas of the pitch (i.e. Defenders, Midfielders and Attackers).

However there is also something to add here about the sides of the pitch also, because these introduce thier own 'free' points areas.

The first area we'll look at is in defence.

We already saw that a central defender has the following 'free' attributes - in that they cost no CA points:

TATT

Crossing

Dribbling

Finishing

Long shots

Technique (I missed this from the earlier list)

MATT

Creativity

Off the ball

Flair

Teamwork

However, a defender on either the right or the left looks a little different.

TATT

Dribbling

Finishing

Long shots

You will notice that DRs and DLs get charged for crossing and technique.

MATT

Off the ball

Flair

Influence

Again, you will see that the DR and DL is charged for creativity but gains the 'influence' attribute scores at no cost to the CA.

Wingbacks (either side) have the following:

TATT

Finishing

Heading

Long shots

This means that as soon as a DL or DR becomes a wingback, they forfeit their free 'dribbling' attribute score for a free 'heading' attribute score.

MATT

Flair

Influence

Bravery

So, a DL/DR that becomes a wingback loses the free 'off the ball' attribute but gains a free 'bravery' attribute score.

Making sense so far? I hope so.

[Hawshiels]

So, I hope you can see what I meant earlier about positions costing a player CA points.

Each time a player introduces a new position to his list, he loses the 'free' nature of that attribute so it costs the player more because they have to spend points to get points in those attributes.

If a player covered every position on the park, he would have none of the free attribute scores and would have to pay CA points for all of them.

[trebor1185]

Yeah all looks the same as what i've got, I just hadn't noticed that the players gain and lose free attribute scores.

KUTGW

[Hawshiels]

For the avoidance of doubt .... the following do not take up CA points:

- Languages spoken by the player

- Player preferred moves

And, neither height nor weight make any difference with regards to the CA points being used.

[kolobok]

Originally posted by Hawshiels:

And now for the really cool bit - and probably the reason many people have found it hard to work out the CA formula.

There are actually attributes (other than the ones I posted above) that cost NO CA points depending on what position the player can play in. This is critical if you are going to create good training schedules without using up all the points towards the PA of a player.

So, here are the attributes that take up NO points for each position.

Defender

TATT

Crossing

Dribbling

Finishing

Long shots

MATT

Creativity

Off the ball

Flair

Teamwork

Midfielder (DM/M/AM)

TATT

Crossing

Heading

Marking

MATT

Flair

Influence

Bravery

PATT

Jumping

Attacker

TATT

Marking

Tackling

MATT

Flair

Influence

Bravery

Thast's really interesting! I am especially surprised with your findings that jumping has no effect on DM ability, teamwork has no effect on Dcs, and bravery does not affect STs.

Hmmm... I always thought that teamwork is less important for STs than DCs and usually attackers have lower stats for it. Apparently, the logic is different. Now the question is whether it's still important to develop a DC in teamwork area. It makes sense from the team prospective, but maybe not so important if I want to increase player CA in order to push his value up...

I am not really surprised that set pieces attributes play no role in players CA. As IRL, nobody buys a player for the sake of free kicks ability only. And you are right, it can be used in designing training schedule. The only problem I see there is crossing is also in that training category and we don't know how the training will be distributed across free kicks, corners, long throws and crossings. I don't really want my DC waste time training corners, crossings or long throws, but free kicks and penalties would be nice addition. If the efforts don't go in the desirable direction, I would rather train a "maxed out" DC in teamwork or maybe some attacking categories to be able to use him as TM from time to time.

To summarize things a bit. We do

[Hawshiels]

I made an error earlier that I need to correct. In a previous version (not sure how far back) the goalkeeper scores (for a non goalkeeper) took up CA points.

However, I can confirm that in 8.0.2 (not sure about other version 8's), the goalkeeper scores do not cost 'CA points' unless the player becomes trained in the goalkeeper position.

[Hawshiels]

I think that based on a couple of posts in this thread, it is important to appreciate something.

I don't think that a position having a 'free' attribute means that it is useless for that position. What SI have done is made certain positions have a 'free' attribute or two to help balance the effects of these CA scores within the game.

A mixture of engine tweaks, database changes, and these 'free' attributes, is what is giving us a relatively realistic experience within the game. I can see how and why this works and it is there also to remove any 'SUPERMAN' players from the game which is correct.

p.s. They've been doing this for a few versions now, but it is becoming better and better each time they make a change - although it makes it harder for me each time to work out what's going on in order to get a great training schedule.

[PaulC]

Can I just say that this is an ace thread?

Cheers,

Paul

[Hawshiels]

Originally posted by PaulC:

Can I just say that this is an ace thread?

Cheers,

Paul

Coming from you, that's a massive compliment for us. Cheers.

[lance101]

First of all, great work Hawshields!

I would like to say that I'm very interested in knowing how did you arrive at that conclusions? Which method you used? Starting different players with different CA and seeing which ratings it affects? Or some full fledged simultaneous equation or regression?

I have some doubts about the exactness of your results because I feel that CA does not translate to an exact amount of points, but rather to an interval... Once I tried to start a game with a player with every rating 20, and CA 30, and the game adjusted all the adjustable ratings to something around 8 (I don't recall the exact figures), didn't adjust flair, aggression or natural fitness (besides all the hidden ones) and adjusted the physical ratings (pace, accel, balance and a couple more) to about 13. Then I started a different one, with all ratings at 1 and the CA again at 30. The game adjusted all the adjustable ratings again to 8, didn't move flair, aggression and nat fitness, but this time it didn't adjust the physical ratings to 13, but rather to something like 9.

I don't remember the exact figures, but something like that happened -> there is a range of ratings which will fall under the same CA.

[Hawshiels]

Originally posted by lance101:

First of all, great work Hawshields!

I would like to say that I'm very interested in knowing how did you arrive at that conclusions? Which method you used? Starting different players with different CA and seeing which ratings it affects? Or some full fledged simultaneous equation or regression?

I have some doubts about the exactness of your results because I feel that CA does not translate to an exact amount of points, but rather to an interval... Once I tried to start a game with a player with every rating 20, and CA 30, and the game adjusted all the adjustable ratings to something around 8 (I don't recall the exact figures), didn't adjust flair, aggression or natural fitness (besides all the hidden ones) and adjusted the physical ratings (pace, accel, balance and a couple more) to about 13. Then I started a different one, with all ratings at 1 and the CA again at 30. The game adjusted all the adjustable ratings again to 8, didn't move flair, aggression and nat fitness, but this time it didn't adjust the physical ratings to 13, but rather to something like 9.

I don't remember the exact figures, but something like that happened -> there is a range of ratings which will fall under the same CA.

I had created a spreadsheet that I have used for a few years now to help build good training schedules. This made the job easier now, but essentially it uses simultaneous equations to work out the relationship between the costs of 'buying' TATT points, MATT points and PATT points.

However, it is REALLY important to understand the importance of these 'free' attributes also because this is why many people (I believe) have failed to understand the CA and how it reflects onto attributes.

I'm not sure if I understand you correctly when you speak of the 'range of ratings' but there is no limit (that I have found) to any individual attribute based on the CA or PA of a player.

[lance101]

What I mean is that, from my observation of the ratings of that player I described above, there is not just one sum of ratings that gives you a given CA, but rather a range. The player I created in these 2 different games, always had CA 30, most ratings 8, free mental ratings were unchanged, but the physical ratings in one case were 9 (because I started them at 1) and in the other case they were 13 (because I started them at 20). In both cases the player only had one position (M C).

[Hawshiels]

And now for goalkeepers ...

A goalkeeper has the following 'free' attributes.

TATT (outfield)

Every single one of them !!!! So, in other words, a goalkeeper uses no CA points for any of the outfield TATT attribute scores.

MATT

Creativity

Off the ball

Flair

Dirtiness

Work rate

PATT

Jumping

Stamina

TATT (Goalkeeping)

Eccentricity

Rushing out

Punching

It nay be interesting to note that aerial ability seems to cost about one third of the points that other attributes in this area do. I don't really understand this as well as the other areas though. This is nothing specific to goalkeepers though, because there is a difference in 'cost' of attributes like 'pace' and 'acceleration' compared to the others but I've still not got the 8.0.2 numbers for these yet.

[Mitja]

Originally posted by Law_Man:

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by Mitja:

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by PaulC:

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by Arsenal71:

Paul, i think what the users are moaning about and quite rightly is that there seems to be something wrong with the code in terms of increasing some players attributes by sometimes 3 pts and decreasings by the same amount. A couple of examples are in 8.0.1 Silva and now in 8.0.2 Elano. An example of a player decreasing, Hleb in my case is a an unfortunate example. Their stats in game do NOT reflect the db attributes. We have discussed this in Sires, and its certainly code that is the problem.

The researcher side of this is getting looked into. I know you are supposed to see a preview of what the game is going to do in your editor and it looks like this maybe isnt matching up with the game for certain players..... </div></BLOCKQUOTE>

not for certain players, for many of them. it's the code when ME matches up CA and atributes. for example, if player has high CA and "low" atributes in editor, ME will give him + 1/2 even 3, for each key atribute. also the game decreases "all round" players like ronaldo or kaka and favours those who have some bad atributes, but good CA. like silva or aimar.

but basicly it's all about how CA matches up with total sum of atributes. for example if you want to have silva's stats in game just the same as in editor, you'll have to decrease his CA a lot. and that's where researchers failed... </div></BLOCKQUOTE>

What do we think of this quote (the last one) by Mitja? I don't think he's done any empirical research into it but its food for thought, even if its just his opinion. </div></BLOCKQUOTE>

you know what I think. I just want realistic stats, that researchers find and put into db. is it so hard? I posted a lot of evidence how game changes stats in +/- 2 manner in that GQ thread. I'm not interested in any abrakadabra empirical stuff. anyone who uses editor knows what I'm talking about. there were researchers who were also amazed with this issue just like me and now they are probably asking themselfs what's the point of their work. sorry I can't accept that some 127 CA DC looks just the same as terry. and kow you'll probably say- they look alike but better CA player will play better. but I will say what's the point of atributes then!!! major bug!!

[lance101]

On the changing of players attributes by the game... I feel that in 8.0.1 and even more dramatically in 8.0.2, the game favours heavily players with good physical skills. That's why players like Cristiano Ronaldo might look worse than Silva (eventhough he has higher CA), but he will tend to outperform him (at least in my experience Cristiano Ronaldo >>> Silva, in both 8.0.1 and 8.0.2).

[lance101]

Hawshiels: just to clarify my question... are you just solving a system - x equations x weights; or are you using a regression like procedure, where you use more players than the weights you are trying to calculate, and therefore those weights you are giving are "best fits", but not necessarily exact?

[Hawshiels]

Originally posted by Mitja:

you know what I think. I just want realistic stats, that researchers find and put into db. is it so hard? I posted a lot of evidence how game changes stats in +/- 2 manner in that GQ thread. I'm not interested in any abrakadabra empirical stuff. anyone who uses editor knows what I'm talking about. there were researchers who were also amazed with this issue just like me and now they are probably asking themselfs what's the point of their work. sorry I can't accept that some 127 CA DC looks just the same as terry. and kow you'll probably say- they look alike but better CA player will play better. but I will say what's the point of atributes then!!! major bug!!

major bug???

Mitja: I like reading your posts on the forums - some of them can be interesting or entertaining or both.

But saying this here suggests that you either haven't read what has been written about the reasoning behind this (in which case, please do), or you didn't understand it (in which case, please ask for clarification).

But please don't make a statement suggesting a major bug when it is clearly NOT!!! As I have said on numerous occassions here and in the other threads, SI are trying to make the experience more realistic for us all and this seems to me like the most sensible way to do this. The researchers will not necessarily know the weightings of each of the attributes within the engine (only SI do), so SI use this technique to make it more realistic - some months after the researchers have done their job. It does make sense. Note that I don't want an argument about it (please) as I want this thread to be free of rants, moans, arguments (unless constructive) or anything else negative. I cannot enforce this obviously but I'm appealing to everyone's good side here when I respectfully ask this.

[Cleon]

If we agree or disagree, lets not be so dismissive of people's efforts in trying to put their points across. Some major effort as been put into this thread, well done all those who contributed

[Mitja]

http://img91.imageshack.us/img91/4807/guillonri0.jpg

CA 127.

this is not the lonely case. I repeat my self. I think it's mess when ME changes stats in +/- manner. I like the idea of more realistic/decreased stats. but this shall aply all players.

[lance101]

Mitja: I think that the real researchers of the game get some other software that tells them what is the "recommended" CA that a player with such ratings should have, meaning that they know that if they put all too many 20's and then CA 120, that the game would adjust the ratings down.

The problem is that the editor we get doesn't tell us that, therefore we don't know if the ratings we choose fall within the range possible with the CA we picked. A possible way around this is perhaps to just pick the ratings and not assign a CA. I haven't tested this myself, but will the game assign a CA reflective of the ratings you picked or a random CA and mess up with all your ratings? The first choice would seem ideal to me.

[isuckatfm]

Great stuff Hawshiels

I'm going to print that and have a slow read to properly absorb what you've posted.

Just from the first read through it there is one thing that jumped out:-

originally posted bt Hawshiels:-

There are actually attributes (other than the ones I posted above) that cost NO CA points depending on what position the player can play in.

Based on a little test I did this does seem to indicate the difference between the control CA model and the player development CA model (I could just be misinterprreting what you wrote from the first read so apologies if that is so). Here's what I did (it's on 8.0 so I don't know to what extent the it applies to other versions). I took a sample of players (not statistically sound but just to get an idea) and artificially increased their CA by different amounts to see how the game would readjust their attributes to reflect their new CA. What I discovered was that the CA model that controls attributes and which the game uses to readjust attributes based on CA appears to separate attributes into three distinct groups (note I didn't check hidden goalkeeper attributes which may have distorted the results, but there is still a pattern):-

Group A: All visible attributes plus Weaker Foot except Group B and Group C

Group B: Acceleration, Pace, Agility, Balance

Group C: Independent of CA Control Model (OUTFIELD VISIBLE: Aggression, Determination, Flair, Natural Fitness; HIDDEN: Dirtiness, Consistency, Important Matches, Versatility, Injury Proneness)

What I found was that when I artficially increased CA by x number of points, then ALL of the attributes in Group A increased by the same amount and ALL of the attributes in Group B increased by the same amount. For example I increased one players CA by 20 points and Passing, Technique, Corners, Anticipation, plus all of the Group A attributes increased by exactly 6 points (FM Modifier scale 1 to 100). For the same 20 point CA increase player his Group B attributes all increased by exactly 2 points (FM Modifier scale 1 to 100).

What's even stranger about this test is that position played no part in the increases. So when I took a striker and a centre half, increased their CA by the same amount, the game would automatically adjust their attributes in line with the groups A. B and C by the same amount but completely independent of position. The actual amount might vary slightly but both would have identical changes in the Group of attributes as outlined, so finishing and tackling would increase by the same amount.

When I graphed the change in Group attributes versus the Change in CA here's what it looked like (I put a linear and quadratic trendline in there to give a better idea of the general shape):-

To me there is a definite trend there that indicates that Group B attributes consume more CA points than Group A attributes for a given increase in CA.

If I understand correctly from your previous posts you tested this out by creating players and seeing how different changes in variables affected attribute levels. My theory on it is, unless I've misunderstood your approach, is that there is a model that both controls attributes for a given fixed CA and assigns attributes for a given CA for generated players (and researched ones at the start of a game). This is the model you have with TATT, MATT and PATT.

There is then a separate model which determines how attributes change for a given change in CA. This is illustrated by the graph I posted. How this might relate to training is how these available points are distributed for a given change in CA and a fixed training schedule. This is where your analysis comes into play.

One is important for judging CA based on looking at a player's profile.

Both come into play when designing training to improve attributes.

That's my theory on it anyway

TBH at the moment I am not sure if I understand what you've posted fully so I'll need to give it a proper read to figure out how what my test showed and your experiments show tie in precisely. But there does appear to be some contradictions (for example your experiments have shown that Penalty Taking does not require CA points to improve by training yet when I increase a player's CA artificially the game adjust to this new CA and increases the penalty taking attribute ).

Again to the post. I might try and use this insight to adjust my original model and see how it goes.

[Hawshiels]

Originally posted by isuckatfm:

Group A: All visible attributes plus Weaker Foot except Group B and Group C

Group B: Acceleration, Pace, Agility, Balance

Group C: Independent of CA Control Model (OUTFIELD VISIBLE: Aggression, Determination, Flair, Natural Fitness; HIDDEN: Dirtiness, Consistency, Important Matches, Versatility, Injury Proneness)

What I found was that when I artficially increased CA by x number of points, then ALL of the attributes in Group A increased by the same amount and ALL of the attributes in Group B increased by the same amount. For example I increased one players CA by 20 points and Passing, Technique, Corners, Anticipation, plus all of the Group A attributes increased by exactly 6 points (FM Modifier scale 1 to 100). For the same 20 point CA increase player his Group B attributes all increased by exactly 2 points (FM Modifier scale 1 to 100).

What's even stranger about this test is that position played no part in the increases. So when I took a striker and a centre half, increased their CA by the same amount, the game would automatically adjust their attributes in line with the groups A. B and C by the same amount but completely independent of position. The actual amount might vary slightly but both would have identical changes in the Group of attributes as outlined, so finishing and tackling would increase by the same amount.

Exactly! You may remember earlier on in the thread I mentioned that there were really just three groups of attributes and each had a weighting. These are the ones you've listed here. I do plan to (or at least did until you beat me to it) show all of these so that we can produce the best training schedule - afterall, this will be one of the most practical uses of the data. And this is why you are on the right track as far as the ongoing development of a player. I'll post more about this now to show the differences in potential values based on the CA of a player.

[kolobok]

Mitja,

With all respect to your finding, the thread goal is slightly different than simply identifying attributes-CA relationship. Personally I do not care about CA as a number since I never use editor of FMM. Hence I don't care about attributes in the editor. All I care about is better assessment if I should buy a player, let a player go, if player has a potnetial or he maxed out based on what I see in the game. So if you saying that the attributes we see in the game don't have anything to do with ability, that's one story. If they simply not the same as in editor, that's a different one.

Btw, Cleon (or anybody else who know the answer). You tought me how to get text files from the game screen. How about getting text file from the editor or FMM? If I choose, say, Barcelona in the editor and then go to players tab, I can see list of players with CA and PA. However, a trick with PrintScreen does not work there, neither does my attempt to copy data. Is there any way to get such a list? It would greatly help me to complete the part of this research that I promised to do. Otherwise, it may take me forever to extract CA and PA for 400 players by hand.

[lance101]

The ratings in the game are the same as in save game editors (like FMM), but they might be different from the ones in pre game editors, because the game adjusts them due to falling of the range allowed by the CA choosen.

[Law_Man]

Originally posted by PaulC:

Can I just say that this is an ace thread?

Cheers,

Paul

Cheers Paul And thanks to all those who've contributed

[Hawshiels]

So, now for some magic numbers ...

Firstly, to have a player with 20's in all of the attributes, he would need a CA of 270 (not quite exactly but as accurate as we need for this purpose). This CA would truly make the superman player, but in trying to reduce this happening, we are 'stuck' with an enforced maximum of 200.

And here is the really simple part. If you remove the 'free' attributes out of the 'equation', you will notice that each point deducted from the 270 is a point reduced on the in-game attributes total.

So, a player with a CA of 200 has 70 points under the maximum.

A player with a CA of 90 has 180 points below the maximum, etc.

This is why it is important to understand that the baseline (i.e. a player with CA of 1) is calculated as having 269 fewer attribute points than a player with 20 in every attribute.

It is important also to understand this. I did say earlier in the thread that I would only be using the in-game (1-20) scale, but for those that plan to check their own player attribute scores, you may want to bear the following in mind. The game (the program) uses a scale of 1-100 which means that a player could appear to have fewer attribute points (when you look at him on the 1-20 scale). This however would mean that many of his attributes were just below a threshold and 'read to increase' with a bit of training. Imagine a player as having a number of half points under his attribute scores all adding up to a number of whole points.

And here's the neat thing about this if you can be bothered looking at it in the game ...

You could buy a player (in theory) that has a certain CA that is distributed across the attributes in a way that makes them all (or most of them) worth up to four fifths of a point (just one fifth to go) more than is showing in the profile.

So, if we take this example further, you could buy this player and with a little bit of training you could suddenly notice GREEN arrows indicating that they have all gone up in a short space of time. This could potentially make a real difference to this player and yet all you have done is a little bit of training. What I don't know however is if the 1-20 scale is used in the engine or if it is the 1-100 scale. I hope I am being clear here, but if the engine uses the 1-20 scale, a player 'suddenly' hitting these milestones with a little training could suddenly make him a much better player within the game since he could have 21 (this is the maximum number of attributes this can happen with I believe) of his attributes increase by one 'overnight'. But what has actually happened is that they were all ready to move up because they were all at the four fifths of a point stage.

Don't hesitate to ask for clarification on this as it is quite complex but very important I think.

[Law_Man]

ps Hawshiels, do you think it would be better to use TAC, PAC and MAC instead of TATT, PATT and MATT? Just a pedantic point, in the face of your AMAZING research KUTGW

[isuckatfm]

@ Hawshiels

So you reckon there is one 'model' that controls the total number of attributes a given change in CA can produce, then a second 'model' which affects how they are distributed based on training.

The players I looked at to produce that graph were all AI managed teams. I'm not sure whether or not it's true but I always assumed AI managed players were set to the General training schedule by default and that's why they had an even distribution across the board in Groups A and B in my post. But if we use your approach of MATT, PATT and TATT to alter training schedules we can more accurately control how these attribute points get distributed and do it in the most efficient manner.

So to use the exact data points rather than the fitted curve, in my data a player who had an increase of 30 points in CA had an 8 point increase in each of the Group A attributes. There are 29 of these so there is 232 total attribute points that were available for distribution across Group A for those 30 points of CA. In your model these 232 attribute points then get distributed amongst Group A based on the MATT, PATT and TATT weightings and the training schedule the player is involved in.

Does that sound about right? Does that fit it in with your experience in developing players in your save games?

originally posted by kolobok:-

I do not care about CA as a number since I never use editor of FMM. Hence I don't care about attributes in the editor. All I care about is better assessment if I should buy a player, let a player go, if player has a potnetial or he maxed out based on what I see in the game. So if you saying that the attributes we see in the game don't have anything to do with ability, that's one story. If they simply not the same as in editor, that's a different one.

Btw, Cleon (or anybody else who know the answer). You tought me how to get text files from the game screen. How about getting text file from the editor or FMM? If I choose, say, Barcelona in the editor and then go to players tab, I can see list of players with CA and PA. However, a trick with PrintScreen does not work there, neither does my attempt to copy data. Is there any way to get such a list? It would greatly help me to complete the part of this research that I promised to do. Otherwise, it may take me forever to extract CA and PA for 400 players by hand.

I don't believe there is any element of the game that you can use to judge whether or not a player has reached their potential other than what scouts/coaches say.

I don't think there is a way to extract CA/PA/attributes from a database. You could do it with older versions and Genie Scout but I don't know about 08. Maybe have a look in the Editor's forum to see if there is a new Genie Scout out.

From what you've posted it sounds like you are trying to use statistics to see if there is a link between the CA and PA. Are you trying to see if you can

1. Look at a players attributes in their profile screen

2. Judge CA based on this

3. Judge PA based on CA

I don't believe there is an explicit link between CA and PA other than PA acting as a cap on the maximum a player can reach, and the difference between the two acting as a control on how much CA can increase. While playing FM I've always used 3rd party tools to look at how players CA evolves over time and I have a theory on how it works, as well as a rough idea of the possible maths involved. Since you mention that you don't use tools to look at CA/PA, maybe my insight from doing so over a few versions will help.

If you have any interest let me know and I'll post it.

[Law_Man]

Originally posted by Hawshiels:

And now for the really cool bit - and probably the reason many people have found it hard to work out the CA formula.

There are actually attributes (other than the ones I posted above) that cost NO CA points depending on what position the player can play in. This is critical if you are going to create good training schedules without using up all the points towards the PA of a player.

So, here are the attributes that take up NO points for each position.

Defender

TATT

Crossing

Dribbling

Finishing

Long shots

MATT

Creativity

Off the ball

Flair

Teamwork

Midfielder (DM/M/AM)

TATT

Crossing

Heading

Marking

MATT

Flair

Influence

Bravery

PATT

Jumping

Attacker

TATT

Marking

Tackling

MATT

Flair

Influence

Bravery

So does this mean that players can increase their visible (1-20) attribute scores in these positions through training, without taking any points up, so that those points would then be free for other attributes?

[isuckatfm]

@ Hawshiels

Sorry, your post above appeared while I was typing in my above post so I didn't see it. And it answers my question .

[Hawshiels]

Originally posted by Law_Man:

ps Hawshiels, do you think it would be better to use TAC, PAC and MAC instead of TATT, PATT and MATT? Just a pedantic point, in the face of your AMAZING research KUTGW

Anything you can do to make this more readable would be excellent. I feel as if I'm developing my own little language here but there's no point if it is confusing. So, please by all means be the interpreter.

[p.s. Do you realise that today alone there have been over 700 views of this thread. Just imagine how many more we'll have if you manage to made this stuff readable. ]

[Law_Man]

Originally posted by Hawshiels:

So, I hope you can see what I meant earlier about positions costing a player CA points.

Each time a player introduces a new position to his list, he loses the 'free' nature of that attribute so it costs the player more because they have to spend points to get points in those attributes.

If a player covered every position on the park, he would have none of the free attribute scores and would have to pay CA points for all of them.

This is quality info!

[Hawshiels]

Originally posted by Law_Man:

So does this mean that players can increase their visible (1-20) attribute scores in these positions through training, without taking any points up, so that those points would then be free for other attributes?

Exactly!

That's what I hate about you Law_Man. It took me 23 paragraphs and 20,000 words to say what you have written in this single sentence.

[kolobok]

From what you've posted it sounds like you are trying to use statistics to see if there is a link between the CA and PA. Are you trying to see if you can

1. Look at a players attributes in their profile screen

2. Judge CA based on this

3. Judge PA based on CA

I don't believe there is an explicit link between CA and PA other than PA acting as a cap on the maximum a player can reach, and the difference between the two acting as a control on how much CA can increase. While playing FM I've always used 3rd party tools to look at how players CA evolves over time and I have a theory on how it works, as well as a rough idea of the possible maths involved. Since you mention that you don't use tools to look at CA/PA, maybe my insight from doing so over a few versions will help.

If you have any interest let me know and I'll post it.

I think it's thread related, so post it

Now, here is the thing. I do want to look at players attributes in their profile screen and make the best possible conclusion about how good players are. I do have some understanding which attributes are more important for each position, but obviously I would like to know more. In particular, if I see a player with right stats all around except one or two, I would like to know how crucial it is that he has, say, low agility. Suppose I see a DC that looks almost like Terry in many stats, but nobody rates him even close. I would like to know why. Finally, I would like to have better understanding if player has a potential and how far he can go. I do believe that there is a link between CA and PA through some attributes (maybe those "free" mental attributes?). Speaking in FM terms, if I were a scout in FM, I would have judging current and potential abilities stats around 14-15 and I would like to have 20s .

[isuckatfm]

@ Hawshiels

Quick question about the positional and side costings. Did you figure out at what point the game takes this into account? Is it a certain positional rating (10/20 for example)? Is it the point where the L or R or C appears in their name?

For example if I have a DM C and retrain him as an AM C at what point does the game readjust the 'distribution costs' to reflect this?

[Law_Man]

Originally posted by Hawshiels:

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by Law_Man:

ps Hawshiels, do you think it would be better to use TAC, PAC and MAC instead of TATT, PATT and MATT? Just a pedantic point, in the face of your AMAZING research KUTGW

Anything you can do to make this more readable would be excellent. I feel as if I'm developing my own little language here but there's no point if it is confusing. So, please by all means be the interpreter.

[p.s. Do you realise that today alone there have been over 700 views of this thread. Just imagine how many more we'll have if you manage to made this stuff readable. ] </div></BLOCKQUOTE>

Exactly! I'm in the middle of my law school exams currently until the end of this week but I'm off next week so (along with some serious FM-ing) I'll get the "written bits" written up. And I'm the perfect yardstick too I think because I've got little if any, mathematical knowledge, so if I can understand it, then anyone can, and that means they'll understand what I write up. Early next week I'll email you guys with a copy of what I've written so that you can make sure its right, and suggest any amendments if needs be before I post it on the forums

Exactly! Thumbs Up

That's what I hate about you Law_Man. It took me 23 paragraphs and 20,000 words to say what you have written in this single sentence. Wink

Sorry mate, sort of comes with th'territory!

And in fairness, I could not BEGIN to do any of this maths-type malarky! So we're about even...

[Law_Man]

And hopefully (in the end) we'll come up with:

(1) training schedules by position

(2) in game attribute search filters by position

Sure there's other things that this might be sueful for too but those are the two off the top of my head!

[Law_Man]

Originally posted by kolobok:

Speaking in FM terms, if I were a scout in FM, I would have judging current and potential abilities stats around 14-15 and I would like to have 20s .

I'd have flair and creativity 20, teamwork 15 and decisions 0

[Law_Man]

Originally posted by isuckatfm:

@ Hawshiels

Quick question about the positional and side costings. Did you figure out at what point the game takes this into account? Is it a certain positional rating (10/20 for example)? Is it the point where the L or R or C appears in their name?

For example if I have a DM C and retrain him as an AM C at what point does the game readjust the 'distribution costs' to reflect this?

Apologies for the annoying multiple posts, my head is fried due to law exams.

Anyway, yeh I think I asked Hawshiels about this earlier in the thread: i.e. does being "unconvincing" in a position take up the same, more, or less, CA than being say "competent" or "accomplished" and so on...?

[lance101]

I've done some more tests to Hawshiels theory, and it all works out the way he says..

Does one player lose ability points through having one more position in itself, or just by its effects in the free attributes? (For example, if I train a player that is AM L to AM R - which has the same free attributes - will he lose any ability points that could be spent elsewhere?)

Congrats for the great job!

[kolobok]

Anyway, yeh I think I asked Hawshiels about this earlier in the thread: i.e. does being "unconvincing" in a position take up the same, more, or less, CA than being say "competent" or "accomplished" and so on...?

Let me guess (to see if I earn 0.5 point in my abilities ). It's not exactly the position that takes points from CA, but rather having higher "position important" attributes for more positions. So the higher rating for the additional position, the more CA points are "eaten" by it.

[Hawshiels]

Great question about the stage when a position 'costs' CA points.

Well, the first thing to tell you is that I know the answer and I'll explain it. The bad news is that it is more difficult to work this out within the game. Here goes.

Each position has a rating of 1-20.

The first part of this is less important to get exact but I'll explain it roughly anyway.

The stages are yellow (awkward), orange (unconvincing), orange (competent), darkened green (accomplished) and lightened green (natural). Or something like that.

However, here's the issue.

There is no affect on the 'free' attribute until a player reaches 11 out of 20. Up to 10 out of 20 for any position does not lose the 'free' attribute. But unfortunately, within the game you cannot tell when you have gone from 10 to 11 in the score. The both read 'unconvincing'. If it goes as far as 'competent' you have obviously gone too far, but there is no way to distinguish the tipping point. I would therefore suggest that orange should be taken as meaning they have lost the 'free' attribute just to be on the safe side.

Secondly, there is very little difference (3 or 4 CA points depending on position on the field) between scoring 11 for a position and scoring 20 for that position. This seems strange in some ways but that's how it works anyway.

Remember that this only counts for a new area in the field. So, a DMC being trained as a MC or AMC will make no difference to the 'free' attributes as they are all in the same area. However, as soon as a player moves into another side of the field (i.e. left, right, centre), it costs them points as soon as they reach the magic number of 11 out of 20 for the position.

[isuckatfm]

originally posted by kolobok:-

In particular, if I see a player with right stats all around except one or two, I would like to know how crucial it is that he has, say, low agility. Suppose I see a DC that looks almost like Terry in many stats, but nobody rates him even close. I would like to know why.

The only people who know the answer to that for certain are the coders of the game.

In any given scenario in the match engine the importance of an attribute is going to depend on how much weight is placed on that attribute when deciding on a 'successful' outcome.

So in your example of a defender with low agility how does the match engine judge if a striker can beat that defender is only something you or I can hypothesise about from observation. If by no one rates a defender who looks almost identical to Terry you mean what scout/coach reports say then what Hawshiels has posted about the link between CA and attributes is particularly relevant. I have messed around with player's CA/PA and in doing so have found that this is the dominant factor in the scout/coach reports. So if you assume that that scout is accurate then the reason your judgement based on looking at his attributes differs from theirs from seeing CA/PA is because of how attributes are distributed based on CA.

Scouts judge primarily on CA and not on attributes. The match engine calculates based on attributes. This is particularly relevant with regens because if the part of the game that assigns attributes to regens doesn't work effectively you end up with regens with high CA but not so great attributes for their position. You see this alot with players with really high physical attributes. I had a striker whose CA was 'leading premiership striker' according to coaches/scouts but he had really high physical attributes that consumed his CA Points and left him with finishing of 12 and technique 14. Hardly 'leading Premiership player' standard.

Basically in my experience the model for player development and the relationship between CA and PA looks something like this

The Potential Ability (the straight horizontal line) acts as the ceiling on Current Ability (the graph), and as CA approaches this value as a player gets older the rate at which it can change decreases. The Decline Point = the age at which players CA, and by extension attributes, starts to fall. This varies from position to position and from player to player.

With this particular model there is a specific variable which controls the rate at which CA reaches PA (the curvature and slope of the graph) and thus whether it is reached. The actual elements of the game that control this include playing time, quality of training, injuries, personal traits etc.

I represented it here as a continuous model but it is more likely a complex version of this involving a piecewise function. This means the game re-evaluates the controlling variables at regular intervals and adjusts the projected curve if necessary.

There are elements of the game that I have observed that reflect this model. For example youngsters tend to get large increases in physical attributes which requires more CA and you get a steeper curvature to reflect this and general development. As a player gets closer to their P, increases in CA slow. At this point attributes are more likely to get distributed amongst mental ones like anticipation, decisions etc.

Towards the end of a players career they reach the ‘Decline Point’ where CA starts to drop dramatically initially mainly with physical attributes declining. It then starts to tail off and you see a slower decline in Mental and Technical attributes.

That's my theory on it anyway . How right or wrong I am only SI knows. Also it's not something I've tested so there may well be players who buck the trend I've typically seen.

[lance101]

But don't (for instance) RB and LB have the same free attributes? It's adding left back to right back that grabs the ability points, or is it just when you reach into a position with different free atributes (say a RB becoming AM L)?

[Hawshiels]

Originally posted by lance101:

I've done some more tests to Hawshiels theory, and it all works out the way he says..

Does one player lose ability points through having one more position in itself, or just by its effects in the free attributes? (For example, if I train a player that is AM L to AM R - which has the same free attributes - will he lose any ability points that could be spent elsewhere?)

Congrats for the great job!

lance101: A player can be re-trained from left to right (but not centre) without losing the 'free' attribute. This applies to defenders, midfielders, forwards, etc. However note that if you do train a right footed AMR player as an AML, you will find that he will develop his left foot and this will cost CA points. If an AML/ML player can use either foot, he can be re-trained in the AMR/MR positions without losing any CA points.

Thanks for making me clarify this as it is again very important!

[lance101]

Thanks for the clarification!

Now we are all waiting to know which attributes are more important for each position... Can you confirm that the attributes in "group B" of isuckatfm post (pace, agility, acceleration and balance) are problably the more important ones for all outfield positions?

Keep up the good work!

[kolobok]

The Potential Ability (the straight horizontal line) acts as the ceiling on Current Ability (the graph), and as CA approaches this value as a player gets older the rate at which it can change decreases. The Decline Point = the age at which players CA, and by extension attributes, starts to fall. This varies from position to position and from player to player.

Exactly the way I think about it. The problem is that when you look at scout reports they sometimes say about 20 years old that he plays close to his full potential, and sometimes they think 24 yeras old player can get much better. So the point on the Age line at which CA reaches PA moves (as well as decline point).

Another point, why I think this research is very important is Lower League clubs often put restrictions on where you can send your scouts, plus the scouts are crap (at least attributes tell me so). In addition, as long as I realise that I can win promotion (and that could be as early as mid season), I want to be able to judge potential signings on the new scale. The scouts, however, will continue to assess players based on the current league scale until I formally win promotion. The next day after the win a player, who was highly recommended yesterday, will be not worth to sign. Again, I have understanding how to judge player plus I may have something else in mind (say sign an older player for free for one season just to make sure I have a backup), but the more the better.

Hawshiels,

It's just crossed my mind. Say a player has reached his PA (so CA=PA). Since I don't use any third party soft, I cannot know it. If I decide to retrain him to a new position, it wouldn't be possible? You plrobably have never faced the situation, but even your thoughts would be much appreciated.