Ok I posted the test results in a new thread so a mod can probably close down this one. Feel free to comment on the test results if anything is unclear.
Following a discussion with some members of the forum on what would be the impact of retraining on the CA of a player I realized that there were opinions on the subject but little data.
So I decided to conduct a little test to clarify the situation.
The starting hypothesis was the following: when player gain attributes it will cost them CA points, that cost is not the same for all the player and is driven by their position. As an example it is more expensive for a defender to train tacking than it would be for a striker because that attribute is considered by the game to be useful for the position while it is considered to be much less useful in the case of the striker.
There was therefore a theoretical possibility to exploit that by retraining a player for a position very different of his. For instance you take a striker you teach him tackling which is pretty much a free attribute for him and then you turn him into a defender.
For that test you are going to take a player put him in a position different than his with the editor while erasing his natural position to simulate the perfect training that we could have achieved (like a natural defender with striker stats) and then we are going to reintroduce a position familiarity in the striker role up to natural.
I took one of the striker from my AJAX save. Based on his attributes when he is playing as a natural striker with no other roles he has a CA of 125. When I erased the striker familiarity to 1 and instead puts him as a natural defender his CA is 103. There is no magic here this is just the application of the different weights for the different attributes and it shows that our starting hypothesis is valid.
Now this is what happens when I start retraining him as a striker:
What you see in the defender/strikers columns are the position familiarity for the role.
The RCA is what the game would see as the new CA, if that value is above the CA of the player we have 2 options, either the player is maxed out in terms of PA and he will has to lose attributes to compensate until he reaches back his CA, if he is not maxed out it will cost him CA in term of developement and will not be able to reach the same level of attributes than he would have had he not been retrained.
The delta is the impact on the RCA of 1 point increase in the similarity when we retrain him.
I also did another simulation with another player with better stats and a natural CA of 144 as a striker instead of the 125 that we have in the first set of data to see if the move are consistent for different player ranges.
What we can see is that retraining someone up until 8 is free but then some costs start to appear for the better subject when we reach 9.
Then the cost of retraining is not linear with the bulk of the cost happening between 10 and 16 familiarity which means that is basically always worth it in a more normal scenario to retrain someone from accomplished to natural in a position as the cost for that upgrade is almost free.
We can also see that retraining someone with 2 opposite positions seems to take to worse of the 2 weights and not the average. It can be shown that when our subject was a pure defender he was CA103, a pure striker CA125 and when he was natural in both he was CA133. So retraining someone to an opposite position is never worth it from an optimisation point of view.
Also when you scout and you see players able to play different positions (even if they cant play them well) it will have an impact in terms of development vs someone that can only play 1. The more opposite the position the worse the impact will be. So for instance an advanced midfielder that can play as a striker will cost you virtually nothing, while a player being able to player being able to play as a DC and all the midfielder roles will cost a lot.
So the test clearly goes in the direction that some people had hinted to in their comments but at least now we have data to back it up
Poacher said: I strongly suggest ignoring any 3rd party tools like FM Scout because they don't have insight into the game code and all they calculation are based on assumption that people make which often are incorrect and outdated.
You can only trust calculation from official tools like the official pre-game editor or the official in-game editor from the devs.
Please, don't be offended but what you offer doesn't make sense. The guys have already explained you very well how it works, honestly, I don't understand why you still don't get it. Expand
I get what they say but I'm an engineer and I need to understand exactly how it works. Don't worry I will conduct the test and share the results. For the 3rd party source not being reliable I have to disagree it just depends which source and it has to be tested. For something like CA it is very easy to check if the calculation is off or not but don't worry I will double check with the ingame editor.
Thanks for all the people taking the time to comment. I think that the test that needs to be run is the following: - take a young player with a single position like a defender - check his CA and double check the values with the FM calculator. Let's assume that the guy has 90 CA and that it matches what the FM scout calculator would find. (it should) - increase the CA of the player to let's say 160 by only bumping striker attributes so only increase manually stuff like finishing / dribbling/ first touch and so on until you reach 160CA. - you should now have a defender with weird attributes with 160 CA and 160 PA. - recheck with the FM calculator that their estimate of the CA is still valid (it should). - check with the FM calculator what the CA would be if the guy was a striker. Normally if you went hardcore enough you should have something much higher like 180. - now start adding a new striker familiarity and check until when you can keep that mismatch. - we already know that until 10 familiarity it will be free. But maybe we can get him to accomplished and still have a higher de facto striker CA as the game still considers him to be more a defender than a striker.
Hope that makes sense. If nobody can do it I will run the tests when I'm back home. In a way I just want to see if there is an arbitrage opportunity.
I'm on holidays atm so I can't access the game but fore sure I will do a proper test when I'm back and share it here. I hoped that someone had done the test already 😊
So according to you the weight cost of the attributes is calculated based on a worse off scenario (like if you can play multiple position the game will always use the highest weight to determine the cost of the attribute). That is a possibility and that would indeed kill any attempt to game the system even if that would be quite unfair. Do you have any data for that affirmation or this is your gut feeling? If that were the case it would also mean that from an attribute point of view players able to play multiple positions are a bad deal (like a striker /midfielder) as you get more flexibility that you might not need but it has a cost.
You think that if you take a defender, Acceleration attribute costs cheaper for a defender than it cost for a striker and train him to play a striker then you'll get Acceleration attribute cheaply but that's not how it works when he learns playing a striker then a recalculation of CA will happen and Acceleration attribute will start costing as for a striker and that's how it works. Expand
OK interesting but when is that recalculation happening and on what basis? Because the guy is still considered by the game to be a defender so the cost cannot be the same as am for a striker. So average of the 2? Only for future points or even for already allocated points?
I found another way to express this. When a player has reached its full potential his CA=PA but that calculation is based on his role. So if you have someone that was created by the game as a defender but was retrained since he was 15 as a striker he will still achieve the same PA but you could get much more points of what is useful for a striker for that PA.
For instance those acceleration /pace points will eat up less CA so you'll be able to get more potentially creating a de facto higher PA than what the limit would imply if it makes sense. I understand that this is a very hardcore approach.
Lapidus said: I see where your confusion comes from. You suppose that it's easy for a player to forget his "natural" playing position but that's not true. A player might not play on his natural position or not train it for very long time but he won't "forget" it. As I said I don't remember that a player drops his natural playing position. Expand
OK I take from that that the player natural position is hard coded and will never change. If that is the case this is ever better for what I want to do. As the weight of the attribute cost will be completely de correlated from the position familiarity
Lapidus said: You're complicating things and talking about unrealistic scenarios and things that don't make sense.
I've already told you already it works quite simple.
If you got a pure striker from your youth intake and you want him to be as efficient as possible at the STC position then don't train him any other positions. If you want him to cover some other positions then you can train him to play there but he won't be the same efficient at the STC position as if it was the only his playing position.
P.S. I don't think it's possible for a player "to forgot" his "natural" playing position, at least I don't remember I've seen it. Expand
You don't get what I mean. I want to min/max the attributes I can put on a player. If you were to train quickness on a defender it will eat up less CA points than the same number of points on a striker because the weight of these attributes is less for the defender than for the striker. I can then play that defender as a striker. Yes he will most likely have terrible striker attributes (like finishing or first touch) but these attributes are pretty much ignored by the match engine so I don't care if my ex defender played as a striker has a finishing of 1.
So contrary to what you are saying this is not simple at all and I'm looking for hard facts. I understand that this is a very unusual min/max approach of the game but I have 1000s of hours into FM so I'm kinda bored playing normally. Especially with a match engine as broken as the one we have in FM21.
So in a nutshell it has nothing to do with normal retraining. I just want to max out the attributes points that I can load onto the player
ZaZ said: It's possible, but highly unlikely to produce anything good in a reasonable time. The initial attributes are just too different. Maybe it would take 4 or 5 years for a striker to reach the attributes of a teenager defender, which isn't very impressive at all. Also, DC is one of the positions with the lowest costs for pace, acceleration and agility, which are the most effective attributes overall.
I believe the natural position is decided at generation time, so you can't change it even if he only plays in another position. The natural position is the one that never lowers (always 20) when you stop training, while the other positions can decrease if you stop playing/training. Expand
I think it depends. Physical attributes are pretty much the only thing that matters for the match engine so if I have anybody with strong physical attributes I can retrain him to play anything basically as ultimately I don't care if my defender ends up with 10 tackling or my striker ends up with 5 finishing.
I agree that it would take time but for someone coming out from the youth intake you basically have 3 years which is quite a lot of time to get them in a decent junior role. Of course they have to get the right starting physical attributes. I will run some tests when I'm home and post them here. I had hoped that someone had gone through the process already.
Thanks but this is not what I meant. My point would be let's assume that I have a young striker coming out of my youth intake. Let's assume his CA will be 90 and his PA 160. He has a lot of room to grow. I now start training him as a defender. Is the cost of all the future points attributed using the weight for striker or defender? Because that would lead to a massive difference in potential points. Bonus question what is the exact trigger? (I have read that it could be position familiarity that would decide if the game considers you to be a striker or a defender) but if the player only trains as a defender but rarely plays it will take ages to achieve familiarity as a defender (which would be good as we could load up more cheap attributes points onto him)
I have been searching the forums for a clear answer on this but couldn't not find anything clear.
When a player gains attribute points the CA cost will depend on its position so increasing tackling for instance will be very expensive for a defender but pretty much free for a striker.
What happens if I have a striker that I retrain as a defender? Is the cost increase based on being a striker? A defender? An average of the 2?
Is it possible to min/max by keeping it natural as a striker and only accomplished as a defender so that the tackling cost is using the striker weight rather than the defender weights?
This is what I thought as well but it didn't work after testing. I downloaded into excel 300 5 stars potential regen from my save all aged 15-17. I ch9different patterns such as sum of all attributes / really high attributes /combinaisons of attributes to search for patterns. I then checked them all using fm scout to see what was the reality. Nothing that I checked worked. You have some kids with low attributes poor scout rating but a 190 potential vs a kid same age that has really high starting attributes but max PA 130.
I had posted that on the official forum but I thought that given the amount of testings that some of your guys doing you might have an answer.
I was wondering what are your tricks to sign the really good young players.
You can scout like crazy but the issue is that you then end up with hundreds of potential 5 stars players like 3 yellow + 2 white or 4 yellow + 1 white. But given that the players are so young it basically doesn't mean anything. I had players ending up 190 PA and some others 130 PA.
Of course I have 100% knowledge of these players and good scouts so the result is as accurate as it can be for a 15-17 years old.
What kills me is that it still feels like a lottery and buying these players can still be expensive. A young Brazilian can easily cost 5-10m so it's not like I can sign 50 and see what comes of it.
I have been playing FM for ages but I feel that this version is very random in that respect.
So in a nutshell, once you have scouted the whole of Brazil and Argentina and you only keep the 5 stars potential, how do you separate the good from the bad?
(I even tried uploading all the stats of the scouted players to excel to build models to find predictable patterns but to no avail.)
If you don't have any hardcore sorting way, how do you usually proceed?
So I decided to conduct a little test to clarify the situation.
The starting hypothesis was the following: when player gain attributes it will cost them CA points, that cost is not the same for all the player and is driven by their position. As an example it is more expensive for a defender to train tacking than it would be for a striker because that attribute is considered by the game to be useful for the position while it is considered to be much less useful in the case of the striker.
There was therefore a theoretical possibility to exploit that by retraining a player for a position very different of his. For instance you take a striker you teach him tackling which is pretty much a free attribute for him and then you turn him into a defender.
For that test you are going to take a player put him in a position different than his with the editor while erasing his natural position to simulate the perfect training that we could have achieved (like a natural defender with striker stats) and then we are going to reintroduce a position familiarity in the striker role up to natural.
I took one of the striker from my AJAX save. Based on his attributes when he is playing as a natural striker with no other roles he has a CA of 125. When I erased the striker familiarity to 1 and instead puts him as a natural defender his CA is 103. There is no magic here this is just the application of the different weights for the different attributes and it shows that our starting hypothesis is valid.
Now this is what happens when I start retraining him as a striker:
What you see in the defender/strikers columns are the position familiarity for the role.
The RCA is what the game would see as the new CA, if that value is above the CA of the player we have 2 options, either the player is maxed out in terms of PA and he will has to lose attributes to compensate until he reaches back his CA, if he is not maxed out it will cost him CA in term of developement and will not be able to reach the same level of attributes than he would have had he not been retrained.
The delta is the impact on the RCA of 1 point increase in the similarity when we retrain him.
I also did another simulation with another player with better stats and a natural CA of 144 as a striker instead of the 125 that we have in the first set of data to see if the move are consistent for different player ranges.
What we can see is that retraining someone up until 8 is free but then some costs start to appear for the better subject when we reach 9.
Then the cost of retraining is not linear with the bulk of the cost happening between 10 and 16 familiarity which means that is basically always worth it in a more normal scenario to retrain someone from accomplished to natural in a position as the cost for that upgrade is almost free.
We can also see that retraining someone with 2 opposite positions seems to take to worse of the 2 weights and not the average. It can be shown that when our subject was a pure defender he was CA103, a pure striker CA125 and when he was natural in both he was CA133. So retraining someone to an opposite position is never worth it from an optimisation point of view.
Also when you scout and you see players able to play different positions (even if they cant play them well) it will have an impact in terms of development vs someone that can only play 1. The more opposite the position the worse the impact will be. So for instance an advanced midfielder that can play as a striker will cost you virtually nothing, while a player being able to player being able to play as a DC and all the midfielder roles will cost a lot.
So the test clearly goes in the direction that some people had hinted to in their comments but at least now we have data to back it up
You can only trust calculation from official tools like the official pre-game editor or the official in-game editor from the devs.
Please, don't be offended but what you offer doesn't make sense. The guys have already explained you very well how it works, honestly, I don't understand why you still don't get it.
I get what they say but I'm an engineer and I need to understand exactly how it works. Don't worry I will conduct the test and share the results.
For the 3rd party source not being reliable I have to disagree it just depends which source and it has to be tested. For something like CA it is very easy to check if the calculation is off or not but don't worry I will double check with the ingame editor.
I think that the test that needs to be run is the following:
- take a young player with a single position like a defender
- check his CA and double check the values with the FM calculator.
Let's assume that the guy has 90 CA and that it matches what the FM scout calculator would find. (it should)
- increase the CA of the player to let's say 160 by only bumping striker attributes so only increase manually stuff like finishing / dribbling/ first touch and so on until you reach 160CA.
- you should now have a defender with weird attributes with 160 CA and 160 PA.
- recheck with the FM calculator that their estimate of the CA is still valid (it should).
- check with the FM calculator what the CA would be if the guy was a striker. Normally if you went hardcore enough you should have something much higher like 180.
- now start adding a new striker familiarity and check until when you can keep that mismatch.
- we already know that until 10 familiarity it will be free. But maybe we can get him to accomplished and still have a higher de facto striker CA as the game still considers him to be more a defender than a striker.
Hope that makes sense. If nobody can do it I will run the tests when I'm back home.
In a way I just want to see if there is an arbitrage opportunity.
That is a possibility and that would indeed kill any attempt to game the system even if that would be quite unfair. Do you have any data for that affirmation or this is your gut feeling?
If that were the case it would also mean that from an attribute point of view players able to play multiple positions are a bad deal (like a striker /midfielder) as you get more flexibility that you might not need but it has a cost.
You think that if you take a defender, Acceleration attribute costs cheaper for a defender than it cost for a striker and train him to play a striker then you'll get Acceleration attribute cheaply but that's not how it works when he learns playing a striker then a recalculation of CA will happen and Acceleration attribute will start costing as for a striker and that's how it works.
OK interesting but when is that recalculation happening and on what basis?
Because the guy is still considered by the game to be a defender so the cost cannot be the same as am for a striker. So average of the 2? Only for future points or even for already allocated points?
For instance those acceleration /pace points will eat up less CA so you'll be able to get more potentially creating a de facto higher PA than what the limit would imply if it makes sense.
I understand that this is a very hardcore approach.
OK I take from that that the player natural position is hard coded and will never change. If that is the case this is ever better for what I want to do. As the weight of the attribute cost will be completely de correlated from the position familiarity
I've already told you already it works quite simple.
If you got a pure striker from your youth intake and you want him to be as efficient as possible at the STC position then don't train him any other positions. If you want him to cover some other positions then you can train him to play there but he won't be the same efficient at the STC position as if it was the only his playing position.
P.S. I don't think it's possible for a player "to forgot" his "natural" playing position, at least I don't remember I've seen it.
You don't get what I mean. I want to min/max the attributes I can put on a player. If you were to train quickness on a defender it will eat up less CA points than the same number of points on a striker because the weight of these attributes is less for the defender than for the striker. I can then play that defender as a striker. Yes he will most likely have terrible striker attributes (like finishing or first touch) but these attributes are pretty much ignored by the match engine so I don't care if my ex defender played as a striker has a finishing of 1.
So contrary to what you are saying this is not simple at all and I'm looking for hard facts. I understand that this is a very unusual min/max approach of the game but I have 1000s of hours into FM so I'm kinda bored playing normally. Especially with a match engine as broken as the one we have in FM21.
So in a nutshell it has nothing to do with normal retraining. I just want to max out the attributes points that I can load onto the player
I believe the natural position is decided at generation time, so you can't change it even if he only plays in another position. The natural position is the one that never lowers (always 20) when you stop training, while the other positions can decrease if you stop playing/training.
I think it depends. Physical attributes are pretty much the only thing that matters for the match engine so if I have anybody with strong physical attributes I can retrain him to play anything basically as ultimately I don't care if my defender ends up with 10 tackling or my striker ends up with 5 finishing.
I agree that it would take time but for someone coming out from the youth intake you basically have 3 years which is quite a lot of time to get them in a decent junior role. Of course they have to get the right starting physical attributes. I will run some tests when I'm home and post them here. I had hoped that someone had gone through the process already.
My point would be let's assume that I have a young striker coming out of my youth intake. Let's assume his CA will be 90 and his PA 160. He has a lot of room to grow. I now start training him as a defender. Is the cost of all the future points attributed using the weight for striker or defender? Because that would lead to a massive difference in potential points.
Bonus question what is the exact trigger? (I have read that it could be position familiarity that would decide if the game considers you to be a striker or a defender) but if the player only trains as a defender but rarely plays it will take ages to achieve familiarity as a defender (which would be good as we could load up more cheap attributes points onto him)
When a player gains attribute points the CA cost will depend on its position so increasing tackling for instance will be very expensive for a defender but pretty much free for a striker.
What happens if I have a striker that I retrain as a defender? Is the cost increase based on being a striker? A defender? An average of the 2?
Is it possible to min/max by keeping it natural as a striker and only accomplished as a defender so that the tackling cost is using the striker weight rather than the defender weights?
Nothing that I checked worked. You have some kids with low attributes poor scout rating but a 190 potential vs a kid same age that has really high starting attributes but max PA 130.
They all appear as 5 stars potential btw.
I had posted that on the official forum but I thought that given the amount of testings that some of your guys doing you might have an answer.
I was wondering what are your tricks to sign the really good young players.
You can scout like crazy but the issue is that you then end up with hundreds of potential 5 stars players like 3 yellow + 2 white or 4 yellow + 1 white. But given that the players are so young it basically doesn't mean anything. I had players ending up 190 PA and some others 130 PA.
Of course I have 100% knowledge of these players and good scouts so the result is as accurate as it can be for a 15-17 years old.
What kills me is that it still feels like a lottery and buying these players can still be expensive. A young Brazilian can easily cost 5-10m so it's not like I can sign 50 and see what comes of it.
I have been playing FM for ages but I feel that this version is very random in that respect.
So in a nutshell, once you have scouted the whole of Brazil and Argentina and you only keep the 5 stars potential, how do you separate the good from the bad?
(I even tried uploading all the stats of the scouted players to excel to build models to find predictable patterns but to no avail.)
If you don't have any hardcore sorting way, how do you usually proceed?
Thanks for your help