The 8.0.2 New Game Attributes "Problem"

[Mitja]

but that's exactly what I'm trying to say.

attributes should be the same as in db and becouse of "When the researcher filled in that player he may have been told the CA was too high for the 14's, but we suspect there may have been a bug in his research tool that meant he wasnt told that. The editor supplied with the game for your use has no such feature." they are not.

logicly, researchers do their finding by stats first and then they asign CA. so it means that players were given too high or low CA, basicly. I hope you don't try to say that computor's right (CA) and researchers (stats) are wrong. this issue shows all weaknesses of CA/PA system, which was discussed many times here. I know thisshouldn't be my concern and that you don't want to talk about it here.

as for looking players on the pitch, there are just too many things that influence how a player will perform.

[Ched]

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

as for looking players on the pitch, there are just too many things that influence how a player will perform. </div></BLOCKQUOTE>

That's why you should compare a whole season's performance between a supposedly over-rated player and a supposedly under-rated player - the stats will show you the point we are trying to make.

I struggle to understand how this argument is going on - it's very simple:

-Some players have a CA that causes their stats to increase - the players performance over a season IS NOT AFFECTED.

-Some players have a CA that causes their stats to decrease - the players performance over a season IS NOT AFFECTED.

-Again i site the examples of ronaldo and elano - elano does not EVER come close to replicating ronaldo's form - for a simple reason - his COMBINATION of stats work better for a player in his position.

-The only people who can possibly have a complaint is the poor reasearchers who are getting stick off cretins because ronaldo's dribbling is not 20 (look in the 2008 data issue forum to see what i mean).

This really is a non-issue.

I can see what's annoying you mitja, it just isn't as significant as it appears.

NB: this isn't a limitation of the CA/PA system - merely a bug in the researchers tool that appears to have prevented them from setting a CA that is concordant with the players stats. No doubt this will be fixed.

[postal postie]

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

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

the only technical stats that elano has that are much better than ronaldos, according to the screen shots on the first page, are his crossing and freekicks and corners.

his passing is also far superior.

i would hardly say that the crossing and free kicks and corners stats affect the plyaers out field abilities that much.

as for the other technical abilities. dribblings the same, finishing is 1 point difference so hardly a big deal, ronaldos heading is much better, and technique and first touch are hardly miles apart.

take into account the hidden stats and also the fact that in my game ronaldo scores about 1 goal every 2.3 games and gets almost as many assits then its hardly the case that ronaldo is much worse than elano. </div></BLOCKQUOTE>

Did you read my post? I'm not at all complaining about Ronaldo being worse than Elano. Clearly Ronaldo performs much better in the game. What I am complaining about is that due to these new adjustments of the stats, Ronaldo and Elano seems to be pretty much equally good players. Elano has a few better technical stats, where Ronaldo has some better physical stats. So, since the player stats is what we mostly use to judge how good a player is, then wouldn't you expect that at least the stats of a player, that is by many (including the FM database) considered among the best in the world, would actually look like more than "almost as good as Elano". </div></BLOCKQUOTE>

in terms of stats taken in the singular they may look like equals.

but in my games at least they are very far from equal with ronaldo scoring and setting up loads and elano strungling to get 1 goal.

you seem to be stuck on thinking that only technical stats make ronaldo look technically good when playing.

alot of his ability is down to his pysical stats.

his ability to dribble past people is enhanced by his pace. doesn't mean he is exceptionaly good at dribbling.

his ability to do the tricks could also be enhanced by his balance and strength. again that doesn't mean he should have 20/20 for his flair/creativity.

there is more to the way the players play on FM than just the stats taken as singulars and this is definetly shown in their performacnes.

[postal postie]

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

but that's exactly what I'm trying to say.

attributes should be the same as in db and becouse of "When the researcher filled in that player he may have been told the CA was too high for the 14's, but we suspect there may have been a bug in his research tool that meant he wasnt told that. The editor supplied with the game for your use has no such feature." they are not.

logicly, researchers do their finding by stats first and then they asign CA. so it means that players were given too high or low CA, basicly. I hope you don't try to say that computor's right (CA) and researchers (stats) are wrong. this issue shows all weaknesses of CA/PA system, which was discussed many times here. I know thisshouldn't be my concern and that you don't want to talk about it here.

as for looking players on the pitch, there are just too many things that influence how a player will perform. </div></BLOCKQUOTE>

the problem i can see is that if the attributes are exactly the same in the db as the game then you run a greater risk of having stats and CA/PA not matching due to human error.

the way paul c explains it, this human error is merged out by the stats increasing or decreasing in the relevant parts so that it matches the CA/PA.

in this respect it must surely be better.

[PaulC]

Its fundamentally better. There is now a nice correlation between performance in the game and CA - which is kind of important!

Before we made this as tight as it is now there were CA 100 players running riot all over the place....."super players" I think they were called on here. I am thinking pre FM days here....those who have been playing a few years will understand what I mean.

The irony is that the "purists" who are getting all het up about the issue in this thread would be up in arms if we allowed the "Braaten effect" to occur in the current game.

[TeeWee]

The obvious idea would then be to not adjust the (presumably accurate) stats to accomodate CA-level when creating the game, but the other way round, to adjust the CA-number to the stats.

Since this idea seems so obvious at first glance, I'm quite certain it is deeply, fundamentally flawed. I just can't see it at first glance.

[solan]

How come then, if player's attributes aren't affecting the game, has Elano played ten games for Brazil, and Kaka, (clearly the better player) made two substitute appearances? Both have got the same natural position. It seems strange to me.

[DaveRH]

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

The obvious idea would then be to not adjust the (presumably accurate) stats to accomodate CA-level when creating the game, but the other way round, to adjust the CA-number to the stats.

Since this idea seems so obvious at first glance, I'm quite certain it is deeply, fundamentally flawed. I just can't see it at first glance. </div></BLOCKQUOTE>

No - I think this is where the bug in the researchers tools has come in. The researchers have been submitting attributes and CAs which don't match and the engine has been boosting/decreasing stats to make up the shortfall.

Remember some players have '0' value stats in the database which get randomly assigned according to their CA and position, so making the engine ignore the discrepancy would not really be sensible.

[Vicente14]

<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 Vicente14:

I think the match engine is fine (aside from the having 20,000 shots on goal and the keeper getting MotM, is that fixed yet? I have not played more than a few months yet). anyway, my point:

I agree that too many goals are scored etc. I bet if you ran the game for 20 seasons on holiday mode and checked through the stats the goals total numbers will be way too high, like it's always been that with a quality striker they will always bag about 30 goals a season when usually that is relatively rare.

I don't have time to go much further, my food will be ready soon, but in summary, retrain the researchers, they're the problem. There is too many super highly rated players, yeah they're good but I feel ratings 17+ should be quite rare, so that like 20 is reserved for legendary types, your Maradona's, Pele's Di Stefano's, Best's etc. Get them to just wind it down a bit. Aguero for example can obliterate any league right off the bat and with training it gets worse but in the scope of the game his stats are around right. If you get what I mean.

The lack of 17+ stats can be compensated by having more 1-3 ratings as they are about as rare as 17+ is common. I get the feeling if I were to put myself in the game I'd be knocking around 8 or 9 for most stats, with some at 10+, which isn't right to me, I should be barely out the 5's at best.

So leave the match engine alone in summary. The real "attributes problem" is with the researchers and not the game. </div></BLOCKQUOTE>

exactly

researhers have double standards. any player with any atribute more then 18 should be an exellant player. yestrday I was comparing my gamst pedersen and bentley (highests stats are natural fitness and corners, both 16) with some other wingers. they are crap and useless in terms of the databese made players. there are hundreds of young argentinian, brazilian and spanish nobody's with all important stats over 16. I know those country's are more technicaly gifted then norvegians. but it kills every enjoyment from me. like you said 18+ stats should be reserved for top class players. now we have many of silva's which are better then maradona and zidane together. and on the other hand player's like cristiano that look ridicoulous even when you set them to 190 CA.

why can't we have player stats same as in editor???? we could at least make our own database then! all becouse of CA/PA system </div></BLOCKQUOTE>

I didn't really mean they have double standards, just that they, I dunno, get overeagre or lose sight of what 20 actually means. To me it means 20 is what the best ever guys had, not just, the best that are playing now. I think the devs feel the same way too. You can see it in the regens, to be blunt they're usually quite awful. That is awful based on the current database. I bet if you ran a game on holiday for 30 seasons you'd see a marked shift in everything. I bet it would become much more realistic, with as I said, stats across the board going universally down. I'm not going to bother doing that but it would be something interesting to do. I'd like to hope the QA guys do it during beta.

Reducing the stats would help the Fatty Ronaldo, Pele, Maradona phenoms shine instead of being absorbed into a cesspool of overachieving, overrated players.

Either that or raise the PA to 400 or something... I'm pretty sure this is totally off topic now anyway since I'm sure they said the matchengine is fine.

Good day.

[azeri4life]

So apparently, nothing can be done about this. Will FM09 at least have this fixed? Paul said there's a bug with the researchers' tool so hopefully FM09's stats will be better than what they are in this patch.

[TeeWee]

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

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

The obvious idea would then be to not adjust the (presumably accurate) stats to accomodate CA-level when creating the game, but the other way round, to adjust the CA-number to the stats.

Since this idea seems so obvious at first glance, I'm quite certain it is deeply, fundamentally flawed. I just can't see it at first glance. </div></BLOCKQUOTE>

No - I think this is where the bug in the researchers tools has come in. The researchers have been submitting attributes and CAs which don't match and the engine has been boosting/decreasing stats to make up the shortfall.

</div></BLOCKQUOTE>

Oh, I'm not disputing the cause. But since it's possible to submit mismatches attribute values and CA, why is raising/lowering stats preferred to raising/lowering CA to compensate? I mean, CA is simply a weighted sum based on the attributes. It makes more sense to me to recalculate the CA rather than adjusting stats so that the calculation is correct (as an aside: how does the engine decide which attributes to adjust?)

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Remember some players have '0' value stats in the database which get randomly assigned according to their CA and position, so making the engine ignore the discrepancy would not really be sensible. </div></BLOCKQUOTE>

Still doesn't make sense to me. I'm sure an algorithm can be written which assigns random values to 0-value stats so that it exactly fills the CA value is possible; Better than retroactively raising/lowering the value of the prefilled stats.

So:

CAE = CA in Editor

CAP = CA based purely on prefilled values

CAR = CA remaining for remaining 0-value attributes and is equal to CAE - CAP

ACA = Actual CA, calculated based on the actual stats of the player.

The game engine then generates the 0-value attributes in such a way that it's weighted sum equals CAR. By definition, ACA = ACE. No problem

There are 2 cases where the algorithm would fail:

CAR is too small (negative or too small when every 0-value stat is filled with 1), then fill every 0-value stat with 1 and recalculate ACA. Logically, ACA > CAE.

If CAR is too big to be filled even when all remaining stats are 20, fill them anyway and recalculate ACA. Logically, ACA < CAE.

I see no reason why this is not to be preferred to the adjusting of prefilled stats. This is so obvious to me, I must be missing something. I just don't see it

[postal postie]

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

The obvious idea would then be to not adjust the (presumably accurate) stats to accomodate CA-level when creating the game, but the other way round, to adjust the CA-number to the stats.

Since this idea seems so obvious at first glance, I'm quite certain it is deeply, fundamentally flawed. I just can't see it at first glance. </div></BLOCKQUOTE>

why?

that's just the the otherway round of whats happening now.

if that happened you could quite easilly say why isnt it the other way round.

players would get a CA/PA boost

and then people would be moaning that the CA changes from whats in the database and that it should be what it is.

[postal postie]

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

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

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

The obvious idea would then be to not adjust the (presumably accurate) stats to accomodate CA-level when creating the game, but the other way round, to adjust the CA-number to the stats.

Since this idea seems so obvious at first glance, I'm quite certain it is deeply, fundamentally flawed. I just can't see it at first glance. </div></BLOCKQUOTE>

No - I think this is where the bug in the researchers tools has come in. The researchers have been submitting attributes and CAs which don't match and the engine has been boosting/decreasing stats to make up the shortfall.

</div></BLOCKQUOTE>

Oh, I'm not disputing the cause. But since it's possible to submit mismatches attribute values and CA, why is raising/lowering stats preferred to raising/lowering CA to compensate? I mean, CA is simply a weighted sum based on the attributes. It makes more sense to me to recalculate the CA rather than adjusting stats so that the calculation is correct (as an aside: how does the engine decide which attributes to adjust?)

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Remember some players have '0' value stats in the database which get randomly assigned according to their CA and position, so making the engine ignore the discrepancy would not really be sensible. </div></BLOCKQUOTE>

Still doesn't make sense to me. I'm sure an algorithm can be written which assigns random values to 0-value stats so that it exactly fills the CA value is possible; Better than retroactively raising/lowering the value of the prefilled stats.

So:

CAE = CA in Editor

CAP = CA based purely on prefilled values

CAR = CA remaining for remaining 0-value attributes and is equal to CAE - CAP

ACA = Actual CA, calculated based on the actual stats of the player.

The game engine then generates the 0-value attributes in such a way that it's weighted sum equals CAR. By definition, ACA = ACE. No problem

There are 2 cases where the algorithm would fail:

CAR is too small (negative or too small when every 0-value stat is filled with 1), then fill every 0-value stat with 1 and recalculate ACA. Logically, ACA > CAE.

If CAR is too big to be filled even when all remaining stats are 20, fill them anyway and recalculate ACA. Logically, ACA < CAE.

I see no reason why this is not to be preferred to the adjusting of prefilled stats. This is so obvious to me, I must be missing something. I just don't see it </div></BLOCKQUOTE>

but then we dont take into account human error where a person gives a player a stat that is too high or low when looking at the CA/PA of the player

as it stands there are loads of player stats, hidden and visible. many possibilities for human error.

instead we'll get a situation where someone gives a player a few stats which are slightly too high, and then all of a sudden the CA comes up to match it and we have average players in real life being world beaters.

there is one PA/CA for each player. this makes it easier to gage how good a player is in terms of CA/PA.

when you have loads of stats it makes it more difficult to get the 'balance' right.

so it makes sense that the stats get smoothed out slightly to mach the CA/PA

[DaveRH]

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

I see no reason why this is not to be preferred to the adjusting of prefilled stats. This is so obvious to me, I must be missing something. I just don't see it </div></BLOCKQUOTE>

The idea is sound, but remember the more complicated your algorithms, the more likely it is for human error or a minor error in coding to have a major effect.

Speaking from a perspective in which I have to write relatively simple code in PHP/ASP to process data from various databases believe me when I say that the simpler you make calculations and data storage the better it is!

On a slight side note, in your solution you would be storing data in the database which would have to be determined manually somehow by the researchers who you can assume that they don't know the player well. If they have issued a value of 0 (aka random assign) to an attribute they don't know the player well at all.

Surely it would be harder for them to determine your CAE, CAP, CAR and ACA values?

I find the single CA and then calculate up/down method to be a little more elegant!

[permaximum]

It would be good if this popular researchers' tool was not bugged and they could do their job like it meant to be.

Or.....

It would be good if SI fixed that bug and allow researchers to enter the attributes again in 2 days.So they could have fun of their work and players could enjoy much more accurate attributes.

But we are so sorry for devs cuz they have a lot of work to do instead of finishing their unfinisjed job...

Hope cm-cm03-04/fm 2005-2008 age won't be end at 2009...

Cheers.

Just a CM/FM fan....

[TeeWee]

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

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

Still doesn't make sense to me. I'm sure an algorithm can be written which assigns random values to 0-value stats so that it exactly fills the CA value is possible; Better than retroactively raising/lowering the value of the prefilled stats.

So:

CAE = CA in Editor

CAP = CA based purely on prefilled values

CAR = CA remaining for remaining 0-value attributes and is equal to CAE - CAP

ACA = Actual CA, calculated based on the actual stats of the player.

The game engine then generates the 0-value attributes in such a way that it's weighted sum equals CAR. By definition, ACA = ACE. No problem

There are 2 cases where the algorithm would fail:

CAR is too small (negative or too small when every 0-value stat is filled with 1), then fill every 0-value stat with 1 and recalculate ACA. Logically, ACA > CAE.

If CAR is too big to be filled even when all remaining stats are 20, fill them anyway and recalculate ACA. Logically, ACA < CAE.

I see no reason why this is not to be preferred to the adjusting of prefilled stats. This is so obvious to me, I must be missing something. I just don't see it </div></BLOCKQUOTE>

but then we dont take into account human error where a person gives a player a stat that is too high or low when looking at the CA/PA of the player </div></BLOCKQUOTE>

Of course there's human error. The whole issue is caused by human error, that is, stats which added up don't match the CA value. The question is twofold: where is the human error and what do we do to fix it?

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">as it stands there are loads of player stats, hidden and visible. many possibilities for human error.

instead we'll get a situation where someone gives a player a few stats which are slightly too high, and then all of a sudden the CA comes up to match it and we have average players in real life being world beaters. </div></BLOCKQUOTE>

this is what I don't understand. What does CA have to do with the performance? CA is just an abstraction based on the individual stats. In other words, CA is based on stat-values, not the other way round. Right?

Assuming a player with all values filled in the editor (no '0'), all fairly average numbers (12 all around the board), including hidden stats. Plus a CA which is lower than would allow such scores.

We now have 2 options:

- raise CA; player will, if I understand the engine correctly, play according to the ability of the 12's.

- lower CA; player will now have lower stats and perform worse than the values in the editor would suggest before starting the game

I have a natural preference for the first, because it doesn't affect the performance of players where the researcher is sure of it.

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">there is one PA/CA for each player. this makes it easier to gage how good a player is in terms of CA/PA.

when you have loads of stats it makes it more difficult to get the 'balance' right.

so it makes sense that the stats get smoothed out slightly to mach the CA/PA </div></BLOCKQUOTE>

This depends heavily on your point of view on which number is more accurate. I'm not a researcher, nor do I know the way one works; I assume the individual stats to be more accurate; you assume the CA stat to be more accurate.

[TeeWee]

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

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

I see no reason why this is not to be preferred to the adjusting of prefilled stats. This is so obvious to me, I must be missing something. I just don't see it </div></BLOCKQUOTE>

The idea is sound, but remember the more complicated your algorithms, the more likely it is for human error or a minor error in coding to have a major effect.

Speaking from a perspective in which I have to write relatively simple code in PHP/ASP to process data from various databases believe me when I say that the simpler you make calculations and data storage the better it is!

On a slight side note, in your solution you would be storing data in the database which would have to be determined manually somehow by the researchers who you can assume that they don't know the player well. If they have issued a value of 0 (aka random assign) to an attribute they don't know the player well at all.

Surely it would be harder for them to determine your CAE, CAP, CAR and ACA values?

I find the single CA and then calculate up/down method to be a little more elegant! </div></BLOCKQUOTE>

I work in the IT industry as well, so I share your concern on the simplicity of the algorithm. BUT: you always have to decide on which stats you adjust; the only complications that are added here are:

- only adjust 0-value stats to accomodate a correct CA level

- if all else fails, approximate CA as far as possible using only 0-value stats; then, readjust CA.

Come to think of it, how does the engine decide to improve the player stats ingame? Does it think "Hey, it's time to raise the CA-number a few points; let's see which stats should now get raised", or the other way round: "Hmm, training and first team play suggest that dribbling and technique should be raised a point. Let's recalculate CA"

If the first, then I understand the developer's reasoning in adjusting stats on loading; I do find the second to be much more intuitive.

[DaveRH]

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Come to think of it, how does the engine decide to improve the player stats ingame? Does it think "Hey, it's time to raise the CA-number a few points; let's see which stats should now get raised", or the other way round: "Hmm, training and first team play suggest that dribbling and technique should be raised a point. Let's recalculate CA"

If the first, then I understand the developer's reasoning in adjusting stats on loading; I do find the second to be much more intuitive. </div></BLOCKQUOTE>

I'd agree, but given that the re-generation of stats happens at the start of the game (and will never happen again) then you can't make it dependant upon training.

[postal postie]

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

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

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

Still doesn't make sense to me. I'm sure an algorithm can be written which assigns random values to 0-value stats so that it exactly fills the CA value is possible; Better than retroactively raising/lowering the value of the prefilled stats.

So:

CAE = CA in Editor

CAP = CA based purely on prefilled values

CAR = CA remaining for remaining 0-value attributes and is equal to CAE - CAP

ACA = Actual CA, calculated based on the actual stats of the player.

The game engine then generates the 0-value attributes in such a way that it's weighted sum equals CAR. By definition, ACA = ACE. No problem

There are 2 cases where the algorithm would fail:

CAR is too small (negative or too small when every 0-value stat is filled with 1), then fill every 0-value stat with 1 and recalculate ACA. Logically, ACA > CAE.

If CAR is too big to be filled even when all remaining stats are 20, fill them anyway and recalculate ACA. Logically, ACA < CAE.

I see no reason why this is not to be preferred to the adjusting of prefilled stats. This is so obvious to me, I must be missing something. I just don't see it </div></BLOCKQUOTE>

but then we dont take into account human error where a person gives a player a stat that is too high or low when looking at the CA/PA of the player </div></BLOCKQUOTE>

Of course there's human error. The whole issue is caused by human error, that is, stats which added up don't match the CA value. The question is twofold: where is the human error and what do we do to fix it?

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">as it stands there are loads of player stats, hidden and visible. many possibilities for human error.

instead we'll get a situation where someone gives a player a few stats which are slightly too high, and then all of a sudden the CA comes up to match it and we have average players in real life being world beaters. </div></BLOCKQUOTE>

this is what I don't understand. What does CA have to do with the performance? CA is just an abstraction based on the individual stats. In other words, CA is based on stat-values, not the other way round. Right?

Assuming a player with all values filled in the editor (no '0'), all fairly average numbers (12 all around the board), including hidden stats. Plus a CA which is lower than would allow such scores.

We now have 2 options:

- raise CA; player will, if I understand the engine correctly, play according to the ability of the 12's.

- lower CA; player will now have lower stats and perform worse than the values in the editor would suggest before starting the game

I have a natural preference for the first, because it doesn't affect the performance of players where the researcher is sure of it.

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">there is one PA/CA for each player. this makes it easier to gage how good a player is in terms of CA/PA.

when you have loads of stats it makes it more difficult to get the 'balance' right.

so it makes sense that the stats get smoothed out slightly to mach the CA/PA </div></BLOCKQUOTE>

This depends heavily on your point of view on which number is more accurate. I'm not a researcher, nor do I know the way one works; I assume the individual stats to be more accurate; you assume the CA stat to be more accurate. </div></BLOCKQUOTE>

from what i can see the PA/CA is a better and more accurate bases for how good the player is overall.

so you can say that you think player A has a CA of 180.

but then it would take you far longer to get the players other stats 'just so' so that the players performance match a player of CA 180.

you could say player B has a CA of 120 but the stats may make him play better than his CA would suggest. like what paul c said about a previous player called braatan i think.

so although there are no wholesale changes to the stats (which would suggest the researchers got it basically right anyway) you would still find it easier to say that a player should play to a standard of CA180 than a player would play to the standard of *then list 20 different stats*

in my mind anyway

for the human error bit. i would say that there is more chance for human error in multiple stats than there is in one stat (CA).

your previous posts suggested that rather than changing the stats to match the CA that you should change the CA to match the stats.

this would surely create more problems where a players CA is over inflated/under valued due to human error in inputting 30 or so different stats?

[EastStandArab]

Yet another reason why 8.0.1 is better.

[TeeWee]

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

from what i can see the PA/CA is a better and more accurate bases for how good the player is overall.

so you can say that you think player A has a CA of 180.

but then it would take you far longer to get the players other stats 'just so' so that the players performance match a player of CA 180.

you could say player B has a CA of 120 but the stats may make him play better than his CA would suggest. like what paul c said about a previous player called braatan i think.

so although there are no wholesale changes to the stats (which would suggest the researchers got it basically right anyway) you would still find it easier to say that a player should play to a standard of CA180 than a player would play to the standard of *then list 20 different stats*

in my mind anyway

for the human error bit. i would say that there is more chance for human error in multiple stats than there is in one stat (CA).

your previous posts suggested that rather than changing the stats to match the CA that you should change the CA to match the stats.

this would surely create more problems where a players CA is over inflated/under valued due to human error in inputting 30 or so different stats? </div></BLOCKQUOTE>

The difference between us seems to be that I believe that the concept that CA represents is, by its very nature, a logical result based on the actual attributes, a derivation if you will. Therefore, if CA and attributes do not match, in my mind, I must assume that CA is given incorrectly and that you should recalculate the CA instead of tweaking attributes until they match CA.

In other words, at game start, "I have 12 in dribbling, crossing and pace, and from that, I calculate that my CA is 120" is a more logical sentence than "I have a CA of 120, therefore I must recalculate my values in dribbling, crossing and pace to be 12."

It seems to me that you think it's easier to judge a player as a whole, on a 1-200 scale. I, on the other hand, believe that the researchers will focus much more on the key characteristics (e.g. dribbling, technique, pace, acceleration for a winger) and mark these correctly on a 1-20 scale. The CA-number will be filled in later to weigh in how well you'd think the player is in general, which is fuzzy precisely because of the unknown stats. The ones that are filled are filled in consciously and are meant to be exactly that.

But then again, I don't know anything about the actual research process in this game, so I may be completely wrong here.

[Hawshiels]

Have you seen this thread ?

Current Ability and Attributes Research

You may find that this helps your understanding.

[Mitja]

TeeWee you're not wrong. CA is artificial number, attributes are those "who make player". at least IRE. zidane was so good becouse he had some skils and skils make ability, not the other way around.

I don't know how much you guys used editor, but I think I'm quite experienced with editing db. and this atribute changing was allways a problem but not so big like right now. if you wanted that players have their researched stats the same as in editor, you only needed to play with CA. but I think that should be SI's work not ours.

as Paul said next year they'll introduce new tool for more accurate handling of stats. that should probably meen they're not too happy with how things work right now, right?

TeeWee, I think we're looking at this issue more from real life perspective and they are looking more from gaming perspective. I don't know who's right and who's wrong...

...lasciamo stare!

[TeeWee]

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

Have you seen this thread ?

Current Ability and Attributes Research

You may find that this helps your understanding.

</div></BLOCKQUOTE>

Thanks for the link. I'd seen it before and I have used it as a base for my thoughts. The article answers how and why the engine changes attributes to match CA; the one thing it doesn't answer is why the engine changes attributes to match CA instead of the other way round.

I still haven't heard a good argument why CA should be the important number instead of the attribute values. The only reason I've heard so far is postal postie's assertion that the attribute values form a less reliable part of the researchers' output than the CA number.

[TeeWee]

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

TeeWee you're not wrong. CA is artificial number, attributes are those "who make player". at least IRE. zidane was so good becouse he had some skils and skils make ability, not the other way around.

I don't know how much you guys used editor, but I think I'm quite experienced with editing db. and this atribute changing was allways a problem but not so big like right now. if you wanted that players have their researched stats the same as in editor, you only needed to play with CA. but I think that should be SI's work not ours.

as Paul said next year they'll introduce new tool for more accurate handling of stats. that should probably meen they're not too happy with how things work right now, right?

TeeWee, I think we're looking at this issue more from real life perspective and they are looking more from gaming perspective. I don't know who's right and who's wrong...

...lasciamo stare! </div></BLOCKQUOTE>

Exactly. CA is weighted sum of the (hidden) attributes; I still have no reason why CA should drive the values of the attributes.

In fact, I'd go so far as to say that the CA value in the editors should be used simply to generate the 0-values. And if there's an unresolvable mismatch between attributes and CA, well, recalculate CA. I just don't understand why it's more logical to change attributes to match CA instead!

Perhaps we (as a collective) have become so focused on the CA-concept that we're taking the CA-number as the driving number in the game. Instead, it should be a tool for the game to determine whether or not a player can develop more (to match to PA), but nothing more.

[PaulC]

We will never ever change CA instead of attributes when balancing the two because CA is the reference point for the hierarchy of teams in the game as well as the overall guidelines for quality of leagues etc.

That makes sense from both a research point of view and a dev one, in my opinion.

Paul

[DaveRH]

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

In other words, at game start, "I have 12 in dribbling, crossing and pace, and from that, I calculate that my CA is 120" is a more logical sentence than "I have a CA of 120, therefore I must recalculate my values in dribbling, crossing and pace to be 12."

</div></BLOCKQUOTE>

That's true only if you have all of the values needed to calculate that final CA figure. The system must account for 0 values.

Lets say that player A is young player in the Danish league who the Danish researcher has not had an opportunity to see.

We know from the media that he's generally considered a quick player so we can feel safe giving him values of 13-15 for pace and acceleration. We can't find any statistics on goals scored/attempts, so we give him 0 values in finishing, composure etc.

We know he's good enough to play for team X week in and week out but not good enough to stand out so the player can safely be given a fairly average CA for the combination of that league and club (say 100 for arguements sake).

The engine will have to fill in the 0 values itself, but it can not just an arbitary number in case it makes the player disproportionately good.

Hence, CA is used to distribute the "spare" points around the unfilled attributes.

If you don't have this then you scupper a large proportion of the "half-known" players in the database.

[TeeWee]

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

We will never ever change CA instead of attributes when balancing the two because CA is the reference point for the hierarchy of teams in the game as well as the overall guidelines for quality of leagues etc.

That makes sense from both a research point of view and a dev one, in my opinion.

Paul </div></BLOCKQUOTE>

Thanks for this reply. I can understand this argument and, using CA as a reference for leagues, teams etc, may have repercussions that I haven't thought of. I'll mull on this for a bit. Not that I don't believe you, but because I want to understand the process.

[TeeWee]

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

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

In other words, at game start, "I have 12 in dribbling, crossing and pace, and from that, I calculate that my CA is 120" is a more logical sentence than "I have a CA of 120, therefore I must recalculate my values in dribbling, crossing and pace to be 12."

</div></BLOCKQUOTE>

That's true only if you have all of the values needed to calculate that final CA figure. The system must account for 0 values.

Lets say that player A is young player in the Danish league who the Danish researcher has not had an opportunity to see.

We know from the media that he's generally considered a quick player so we can feel safe giving him values of 13-15 for pace and acceleration. We can't find any statistics on goals scored/attempts, so we give him 0 values in finishing, composure etc.

We know he's good enough to play for team X week in and week out but not good enough to stand out so the player can safely be given a fairly average CA for the combination of that league and club (say 100 for arguements sake).

The engine will have to fill in the 0 values itself, but it can not just an arbitary number in case it makes the player disproportionately good.

Hence, CA is used to distribute the "spare" points around the unfilled attributes.

If you don't have this then you scupper a large proportion of the "half-known" players in the database. </div></BLOCKQUOTE>

I am taking the 0-values in consideration! See the definitions concerning CAE, CAP, CAR and ACA I've given, and especially CAR.

I'm claiming that CAE and prefilled attributes should be a given in the engine, and that the engine generates the 0-value stats to fill up CAR exactly, or as close as possible, and, if it's mathematically impossible to create a player with the right CAR, to not change the attributes, but the (A)CA instead.

I see that PaulC has given a reply I'm mulling over. I'll have to think whether or not the argument rings true to my ears. It has some resonance, but it's not an intuitive thought.