This is part 2 of the Meta Attributes Weighted Training Session analysis. Today we're looking at the best general all around training sessions. I'm also going to be talking about number of training sessions and how we can be efficient.
TL;DR in Conclusions.
How it's measured(Same as last post) I am using weights based on the @ZaZ's under 50 minus 25 weights from the linked post. I adjusted some weights to fit first the tactic and some of my personal tastes. You can find my weights here. Last post I only did Physicals and this post will be everything other than physicals, using all attribute weights above 35 in a formula for each position and applying it for each position.
The Results There were only 4 training sessions with a average score above 3. I colored the best performer of each position green and the 2nd best as yellow. The only conclusion I'll draw from this is that Defending and Attacking are clearly miles above the rest of the training sessions.
How many of each session should be included? According to the beginning of EBFM's Training video, Part 11. He has an experiment listing the amount of gains when you have X amount of training in a week. This is with the training on its own so it doesn't tell the story of how the sessions work with EACH OTHER, but I believe it's valuable data anyway. Link to this spreadsheet is here.
If you look at Defending and Ground Defense, the most amount of efficient gains occur at 4 sessions each week, while for Physical it's more efficient at around 2 sessions a week but it's notably efficient even at just 1 session a week in terms of growth of attributes.
You can find the combined data for this spreadsheet here.
Conclusions
TL;DR Attacking and Defense are by far the best, and the most efficient gains occur at 4 of each non-physical session per week and 2 of each Physical session per week. With these findings you have the foundation of the Meta Attribute Weighted Training Session Training Schedule. Put these 11 on your training schedule pizza and you can put any toppings on it and you'll probably do well.
What's next? But cheese pizza is boring and we still have 4-7 sessions that we can fill. I'll eventually find what toppings each position likes and see if there is some sort of common denominator. But first, I want to explore what happened to the Match Practice training and why it's not performing in these results when it was practically the best in EBFM's findings. That will be Part 3. In the mean time, I'm hungry for pizza.
Looking forward to seeing the next part of this, was a big fan of Max's schedule but this takes it to the next level!
Can I ask one thing? With the diminishing returns from running multiples of the same sessions, the addition of the second physical session gives 1.4 in attribute growth over the season, is that right?
I'm thinking 1 physical session and 1 ground defence would be more efficient 2 physical sessions, 11.5 growth versus 9.1 but see you went with the 2 physical so I'm wondering if I'm misunderstanding the data
A lot of points great points. I'll start off by saying about this that the growth in CA/Attributes screenshot uses Defending, Physical Training, and Ground Defense. I think that you can extrapolate this data and apply it to 3 different categories for attribute growth.
The 1st type is I what will call General Training which includes all of the trainings under General tab except Physical.
The Physical Training which includes all high-intensity trainings that focus on Physical attribute growth.
The last and by far the biggest group of trainings are specialized trainings. These are all the trainings under Attacking, Defending, Technical, and Goalkeeping training.
I've included the trainings I have so far because they are clearly above the rest in the results. That's why they are the foundation of the Meta Attribute Weighted Training Sessions. However finding the most efficient specialized trainings will take some work as the results are pretty close and some trainings synergize better with more players better than others. To be honest, other than the Foundational 11 there isn't a "correct answer" and the goal of the posts will be to allow players to figure out their perfect schedule for themselves although I will give my best answer as well.
On the topic of physical training, everything has to do with the my personal weights and how it effects the results I get. You can read how I made the weights for my experiments below or just read the TL;DR if you can trust the weights I've cooked .
How the Weights for my Training experiments were formed
I made weights that are based on the Chinese Machine Learning FM Study. It's probably the most quality test on attributes we might ever see for some time due to sample size and the process that is AI machine learning. You can find the weights from the experiment called ykykyk balanced here from @Mark's post on it.
Here is an example of the Full Backs for ykykyk balanced.
There was a problem with just copying these weights 1 for 1 for me.
I feel that Physicals were not weighted important enough. When I was playing FM using these weights a lot of players I was being recommended(as a lower league club) were over the hill and were really declining the league in the 2 most important attributes Pace and Acceleration, but had great Technical/Mental skill to "make up" for it.
Then I used the weights that @ZaZ created all ratings under 50 minus 25 weights which @Markbalanced here. These weights fixed this problem for me by decreasing weights on the less important attributes which coincidentally really fixed the importance for the Physical Attributes.
Here is an example of the Full Backs for all ratings under 50 minus 25.
However, the problem with these weights is that I'm using the top tactic in the Patch 24.2.0 (v2.0) - Tactic Testing Table. The Machine Learning Weights are based off of this tactic from FM22. The meta tactics for FM24 are different from 2 years ago with the biggest difference being the role of the Fullback. The IWB played more like a recycle the ball DM which shows in these weights. Now the Fullback has the responsibilities of a Winger, and needs more weights to represent this.
Here is an example of changes I made to my weights to attempt to accomplish this which you can find here .
For example, I raised some weights in things like Passing, Crossing, and Technique to what I felt best emulated the role of a FB in that tactic.
I made tweaks like this to these weights to fit the top tactic for FM 24 for not just Fullback but the other roles as well(but should work other 4231 tactics).
TL;DR The weights are my attempt at mixing Chinese Machine Learning with prioritizing Physical Attributes to fit FM 24.
Because of that, almost all of the weights above 70 in my weights are trained with any sort of Physical training. I have screenshots here underlining these attributes in green.
Looking at these weights, Physical Attributes are weighted as importantly as the center of the universe. Because of that I try to maximize their growth as much as reasonably possible instead of being efficient with just 1 or 2 sessions.
If the results from physical training showed that more is better then I would do more, but it seems the maximum training sessions that grow Physical sessions in a week is 3. So that's why in the foundational 11 training schedule there are 3 physical sessions in it.
Han106 said: On the topic of physical training, everything has to do with the my personal weights and how it effects the results I get. You can read how I made the weights for my experiments below or just read the TL;DR if you can trust the weights I've cooked .
How the Weights for my Training experiments were formed
I made weights that are based on the Chinese Machine Learning FM Study. It's probably the most quality test on attributes we might ever see for some time due to sample size and the process that is AI machine learning. You can find the weights from the experiment called ykykyk balanced here from @Mark's post on it.
Here is an example of the Full Backs for ykykyk balanced.
There was a problem with just copying these weights 1 for 1 for me.
I feel that Physicals were not weighted important enough. When I was playing FM using these weights a lot of players I was being recommended(as a lower league club) were over the hill and were really declining the league in the 2 most important attributes Pace and Acceleration, but had great Technical/Mental skill to "make up" for it.
Then I used the weights that @ZaZ created all ratings under 50 minus 25 weights which @Markbalanced here. These weights fixed this problem for me by decreasing weights on the less important attributes which coincidentally really fixed the importance for the Physical Attributes.
Here is an example of the Full Backs for all ratings under 50 minus 25.
However, the problem with these weights is that I'm using the top tactic in the Patch 24.2.0 (v2.0) - Tactic Testing Table. The Machine Learning Weights are based off of this tactic from FM22. The meta tactics for FM24 are different from 2 years ago with the biggest difference being the role of the Fullback. The IWB played more like a recycle the ball DM which shows in these weights. Now the Fullback has the responsibilities of a Winger, and needs more weights to represent this.
Here is an example of changes I made to my weights to attempt to accomplish this which you can find here .
For example, I raised some weights in things like Passing, Crossing, and Technique to what I felt best emulated the role of a FB in that tactic.
I made tweaks like this to these weights to fit the top tactic for FM 24 for not just Fullback but the other roles as well(but should work other 4231 tactics).
TL;DR The weights are my attempt at mixing Chinese Machine Learning with prioritizing Physical Attributes to fit FM 24.
Because of that, almost all of the weights above 70 in my weights are trained with any sort of Physical training. I have screenshots here underlining these attributes in green.
Looking at these weights, Physical Attributes are weighted as importantly as the center of the universe. Because of that I try to maximize their growth as much as reasonably possible instead of being efficient with just 1 or 2 sessions.
If the results from physical training showed that more is better then I would do more, but it seems the maximum training sessions that grow Physical sessions in a week is 3. So that's why in the foundational 11 training schedule there are 3 physical sessions in it. Expand
I do still think that the ratings still view Stamina as important, but my mindset is that I don't think that Stamina is at the level of importance of say Pace and Acceleration for all positions.
I based the Stamina ratings on movement data from the tactic, which involved the stats Miles Traveled per 90, High Intenstity Sprints per 90 and Presses per 90 relative to Fullbacks. I also used my own experiences using the tactic in my sunday-league no badge pentagon journeyman save. I have been able to manage the health of the squad by using rotation and by using a lower intensity version of the tactic when I'm dominating the match. Usually, the only position have to sub in my matches are the fullbacks, whose Stamina weight is already at the max in my ratings. I personally think if they aren't in Red by minute 80 or so, the weights have gotten me the players that fit this goal while also maximizing pace/acceleration as much as possible.
Of course not all players might not agree with every number from the weights for every position, however as long as you find a system of analyzing players that works that's all that matters. I do hope that the ratings are close enough to people's priorities so that they still learn something regardless. For Stamina in particular, I don't think changing the weights to their original would've affected the physical training rankings from part 1.
Han106 said: But cheese pizza is boring and we still have 4-7 sessions that we can fill. Expand
Max said that we don't need to fill the empty spaces and he is right. I had worse training results when I used his training schedule and filled wednesday. That's why I'm looking forward to your results. Match practice performed quite well in my tests.
Meriten said: Max said that we don't need to fill the empty spaces and he is right. I had worse training results when I used his training schedule and filled wednesday. That's why I'm looking forward to your results. Match practice performed quite well in my tests. Expand
2 points I'll make regarding this.
Point 1: Convenience Not Maximization EBFM's training schedule is made to be plug and play, convenient and efficient within its restrictions, not to maximize training growth. I will touch on this in my 2nd point.
The convenience is the reason why he doesn't fill in more trainings. There are 21 slots available in a training week. 11 are used for trainings. 6 are taken up by Match days, and it doesn't you actually have a match that day. For example, I play in a small league in India where I only have 1 match a week. 4 are used for if you have an away match and have to travel. It doesn't consider whether or not you have a extra slot available because you have a home match that week, or you live in England and Away matches don't require 2 training slots worth of travel.
It shows growth when you have X amount of sessions in a training week and the results are clear that more is better. How much more I don't know, but you can see that 18 sessions a week does have a measurable increase compared to 10 and very likely 11 sessions. This is why it is worth it to try and fill up slots where available.
Hmm, I did see someone's post on running tests on it and saying that more might not be better so I decided to try and test it myself before moving forward through these tests.
From the test file, I have these players found in FM Genie Scout using the best.grf. I will simply run the tests 5 times for 4 teams and I will report hopefully soon.
I'm using EBFM's testing league with these changes made to it.
Set 1 MC to DM Name Set to DM Set focus for position in Tactic Change tactic to Katana Week 1 set to Rest Import EBFM Schedule Set all training weeks to EBFM
Updated file as I didn't set position focus for Team D
EDIT: Version 3 18:59 PM PT:
Team A/B AMC position focus was listed as Shadow Striker instead of Attacking Mid One of Team B's names was DM instead of DMC Stay on Feet was enabled for all team
Ok the spreadsheet is complete for the baseline tests of EBFM. It has stats like difference between Starting% and End%, Total %, Potential Rating, difference between End % and Potential %, and injury stats.
something popped into my mind while watching EBFM's series that I guess might be relevant here. A lot of these tests look to maximise CA growth per time from training. Given that: -Attributes 'cost' different amounts of CA depending on position. -Players grow faster the larger the difference between their CA and PA. -We can estimate the quality of a player with the attribute weightings.
I think it might be better to maximise the value (weighted attribute) while minimising CA growth.
Another way of looking at is that we want to 'spend' our CA->PA growth in the most efficient way to maximise the weighted attribute values. My feeling is that broad/general training programs like EBFMs will tend to grow most attributes evenly. This is probably the fastest way to grow CA overall, but I think you will get higher quality players if you can focus the attribute growth on highly weighted attributes.
@roman Well, I would need to add more sessions and do the experiments again, i just made a baseline of just doing EBFM's so that we can compare later to not just mine, but test other training schedules in the community as well. I'm a bit tired so I will do so more tests tomorrow. But yes I agree with your thoughts on efficiency of training.
Looking at the data point, Diff from Baseline, I feel that the growth of GKs and FBs are low comparatively to the rest of the positions. Looking at Room to grow, which took the Current % after a year compared to FM Genie Scout's Pot Rating(which may or may not be reliable), DCs and DMs are all lagging behind compared to their attacking position counterparts.
Attacking unit players seem to be doing very well in all metrics. On the other hand, I believe Defending/Goalkeeping players might need more. I'll try adding 2 sessions on the Wednesday designed to try and improve these metrics for those 2 groups. But would moving Defensive players to the Attacking unit work as well?
roman said: This is probably the fastest way to grow CA overall, but I think you will get higher quality players if you can focus the attribute growth on highly weighted attributes. Expand
There is a big problem. One example: Anticipation is good but Decision is bad in context of CA efficiency according to the test results on fm-arena. But there is no training session which trains more Anticipation but less Decision.
Excerpt from MaxEBFM results:
Another example is Acceleration/Pace (both are good) and Agility (bad). I tried to beat the system by using a good and a bad coach but that doesn't work either. Anticipation/Decision is trained by the same coach and Acceleration/Pace/Agility is trained by the same coach too.
Another problem is that we don't know all good attributes. Stamina e.g. is bad according to fm-arena test results but good according to ykykyk.
Meriten said: There is a big problem. One example: Anticipation is good but Decision is bad in context of CA efficiency according to the test results on fm-arena. But there is no training session which trains more Anticipation but less Decision.
Excerpt from MaxEBFM results:
Another example is Acceleration/Pace (both are good) and Agility (bad). I tried to beat the system by using a good and a bad coach but that doesn't work either. Anticipation/Decision is trained by the same coach and Acceleration/Pace/Agility is trained by the same coach too.
Another problem is that we don't know all good attributes. Stamina e.g. is bad according to fm-arena test results but good according to ykykyk. Expand
This is only a thought and totally unproven, but the machine learning was using overall, that is combinations whereas fm arena used single attributes.
Now I know that the machine learning was done with just one tactic over one year, but to me it seems more reliable.
The changes by year for attributes seem marginal. As Rampaging Roy Slavin would say, make of that what you will.
Han106 said: On the topic of physical training, everything has to do with the my personal weights and how it effects the results I get. You can read how I made the weights for my experiments below or just read the TL;DR if you can trust the weights I've cooked .
How the Weights for my Training experiments were formed
I made weights that are based on the Chinese Machine Learning FM Study. It's probably the most quality test on attributes we might ever see for some time due to sample size and the process that is AI machine learning. You can find the weights from the experiment called ykykyk balanced here from @Mark's post on it.
Here is an example of the Full Backs for ykykyk balanced.
There was a problem with just copying these weights 1 for 1 for me.
I feel that Physicals were not weighted important enough. When I was playing FM using these weights a lot of players I was being recommended(as a lower league club) were over the hill and were really declining the league in the 2 most important attributes Pace and Acceleration, but had great Technical/Mental skill to "make up" for it.
Then I used the weights that @ZaZ created all ratings under 50 minus 25 weights which @Markbalanced here. These weights fixed this problem for me by decreasing weights on the less important attributes which coincidentally really fixed the importance for the Physical Attributes.
Here is an example of the Full Backs for all ratings under 50 minus 25.
However, the problem with these weights is that I'm using the top tactic in the Patch 24.2.0 (v2.0) - Tactic Testing Table. The Machine Learning Weights are based off of this tactic from FM22. The meta tactics for FM24 are different from 2 years ago with the biggest difference being the role of the Fullback. The IWB played more like a recycle the ball DM which shows in these weights. Now the Fullback has the responsibilities of a Winger, and needs more weights to represent this.
Here is an example of changes I made to my weights to attempt to accomplish this which you can find here .
For example, I raised some weights in things like Passing, Crossing, and Technique to what I felt best emulated the role of a FB in that tactic.
I made tweaks like this to these weights to fit the top tactic for FM 24 for not just Fullback but the other roles as well(but should work other 4231 tactics).
TL;DR The weights are my attempt at mixing Chinese Machine Learning with prioritizing Physical Attributes to fit FM 24.
Because of that, almost all of the weights above 70 in my weights are trained with any sort of Physical training. I have screenshots here underlining these attributes in green.
Looking at these weights, Physical Attributes are weighted as importantly as the center of the universe. Because of that I try to maximize their growth as much as reasonably possible instead of being efficient with just 1 or 2 sessions.
If the results from physical training showed that more is better then I would do more, but it seems the maximum training sessions that grow Physical sessions in a week is 3. So that's why in the foundational 11 training schedule there are 3 physical sessions in it. Expand
I have adjusted these and tested them for a few seasons, seems that you do find the right players for the Katana v3.1 tactic on the tests.
doru228 said: I have adjusted these and tested them for a few seasons, seems that you do find the right players for the Katana v3.1 tactic on the tests.
Han106 said: Well I'm glad that my weight balance has some merit to it! Thanks! Expand
well you done 75% of the work, i just tweaked it a little bit.
@Han106 there's something that i am wondering tho, if i should include consistency in the ratings, since a consistent player does get better ratings ingame (tested with same player, just consistency was modified over one season. difference was like 1.5 points between 1 consistency and 20, and about 0.75 between 10 and 20.) results were like this:
1 consistency 6.50 10 consistency 7.12 20 consistency 7.76 all of the above over 40 games simulated, frozen attributes and no injuries on
doru228 said: well you done 75% of the work, i just tweaked it a little bit.
@Han106 there's something that i am wondering tho, if i should include consistency in the ratings, since a consistent player does get better ratings ingame (tested with same player, just consistency was modified over one season. difference was like 1.5 points between 1 consistency and 20, and about 0.75 between 10 and 20.) results were like this:
1 consistency 6.50 10 consistency 7.12 20 consistency 7.76 all of the above over 40 games simulated, frozen attributes and no injuries on Expand
You can include it in yours. However I don't include weights that aren't visible to the player. Seems great otherwise
Meriten said: There is a big problem. One example: Anticipation is good but Decision is bad in context of CA efficiency according to the test results on fm-arena. But there is no training session which trains more Anticipation but less Decision.
Excerpt from MaxEBFM results:
Another example is Acceleration/Pace (both are good) and Agility (bad). I tried to beat the system by using a good and a bad coach but that doesn't work either. Anticipation/Decision is trained by the same coach and Acceleration/Pace/Agility is trained by the same coach too.
Another problem is that we don't know all good attributes. Stamina e.g. is bad according to fm-arena test results but good according to ykykyk. Expand
Just as a reminder, the fact that one attribute did not get impressive results in the table of FM-Arena does not mean that attribute is useless. For example, stamina might be impactful only for two or three players, like DM and WBs, which represents less than 30% of the team. In that case, even if Stamina is as impactful as Pace and Acceleration for those positions, it will still give worse results than attributes that have impact across all roles.
ZaZ said: Just as a reminder, the fact that one attribute did not get impressive results in the table of FM-Arena does not mean that attribute is useless. For example, stamina might be impactful only for two or three players, like DM and WBs, which represents less than 30% of the team. In that case, even if Stamina is as impactful as Pace and Acceleration for those positions, it will still give worse results than attributes that have impact across all roles. Expand
No way that Stamina can be impactful as Acceleration and Pace for any positions.
In the FM-Arena test Stamina attribute was increased by 5 points for every positions including the DMs and WBs and that improved the result just by 2 points from 60 pts to 62 pts, which honestly can be just the RNG but when Acceleration or Pace attributes were increased by 5 then it improved the result by 21 points from 60 pts to 81 pts.
So no way Stamina attribute or any other attribute can be as impactful as Pace and Acceleration for any position.
ZaZ said: Just as a reminder, the fact that one attribute did not get impressive results in the table of FM-Arena does not mean that attribute is useless. For example, stamina might be impactful only for two or three players, like DM and WBs, which represents less than 30% of the team. In that case, even if Stamina is as impactful as Pace and Acceleration for those positions, it will still give worse results than attributes that have impact across all roles. Expand
Decision is for every position more expensive than Anticipation (except SC) but the results on the tables are worse in comparision to Anticipation (FM22/FM23/FM24). Maybe there is one position (or even two positions) where Decision has a better cost-benefit ratio than Anticipation. But that doesn't matter. Why? Because you can't train Decision just for one position. You always must train Decision for all positions:
So a high value for Decision in a training session is always a bad thing.
Meriten said: Decision is for every position more expensive than Anticipation (except SC) but the results on the tables are worse in comparision to Anticipation (FM22/FM23/FM24). Maybe there is one position (or even two positions) where Decision has a better cost-benefit ratio than Anticipation. But that doesn't matter. Why? Because you can't train Decision just for one position. You always must train Decision for all positions:
So a high value for Decision in a training session is always a bad thing. Expand
If you give a player Additional Focus of Final Third they will train both Composure and Decisions.
Lapidus said: No way that Stamina can be impactful as Acceleration and Pace for any positions.
In the FM-Arena test Stamina attribute was increased by 5 points for every positions including the DMs and WBs and that improved the result just by 2 points from 60 pts to 62 pts, which honestly can be just the RNG but when Acceleration or Pace attributes were increased by 5 then it improved the result by 21 points from 60 pts to 81 pts.
So no way Stamina attribute or any other attribute can be as impactful as Pace and Acceleration for any position.
By the way, I am testing the schedule below right now. Keep in mind that it requires training with half intensity, else it will lead to a high number of injuries. I also discovered a glitch on training that allows you to "super rest" players (almost 40% more recovery of condition and fatigue than just full rest schedule), but I am still investigating it to understand exactly how it works.
By the way, I am testing the schedule below right now. Keep in mind that it requires training with half intensity, else it will lead to a high number of injuries. I also discovered a glitch on training that allows you to "super rest" players (almost 40% more recovery of condition and fatigue than just full rest schedule), but I am still investigating it to understand exactly how it works. Expand
What do you think about individual training? Are they worth using?
This is part 2 of the Meta Attributes Weighted Training Session analysis. Today we're looking at the best general all around training sessions. I'm also going to be talking about number of training sessions and how we can be efficient.
TL;DR in Conclusions.
How it's measured(Same as last post)
I am using weights based on the @ZaZ's under 50 minus 25 weights from the linked post. I adjusted some weights to fit first the tactic and some of my personal tastes. You can find my weights here. Last post I only did Physicals and this post will be everything other than physicals, using all attribute weights above 35 in a formula for each position and applying it for each position.
The Results
There were only 4 training sessions with a average score above 3. I colored the best performer of each position green and the 2nd best as yellow. The only conclusion I'll draw from this is that Defending and Attacking are clearly miles above the rest of the training sessions.
How many of each session should be included?
According to the beginning of EBFM's Training video, Part 11. He has an experiment listing the amount of gains when you have X amount of training in a week. This is with the training on its own so it doesn't tell the story of how the sessions work with EACH OTHER, but I believe it's valuable data anyway. Link to this spreadsheet is here.
If you look at Defending and Ground Defense, the most amount of efficient gains occur at 4 sessions each week, while for Physical it's more efficient at around 2 sessions a week but it's notably efficient even at just 1 session a week in terms of growth of attributes.
You can find the combined data for this spreadsheet here.
Conclusions
TL;DR Attacking and Defense are by far the best, and the most efficient gains occur at 4 of each non-physical session per week and 2 of each Physical session per week. With these findings you have the foundation of the Meta Attribute Weighted Training Session Training Schedule. Put these 11 on your training schedule pizza and you can put any toppings on it and you'll probably do well.
What's next?
But cheese pizza is boring and we still have 4-7 sessions that we can fill. I'll eventually find what toppings each position likes and see if there is some sort of common denominator. But first, I want to explore what happened to the Match Practice training and why it's not performing in these results when it was practically the best in EBFM's findings. That will be Part 3. In the mean time, I'm hungry for pizza.
Looking forward to seeing the next part of this, was a big fan of Max's schedule but this takes it to the next level!
Can I ask one thing? With the diminishing returns from running multiples of the same sessions, the addition of the second physical session gives 1.4 in attribute growth over the season, is that right?
I'm thinking 1 physical session and 1 ground defence would be more efficient 2 physical sessions, 11.5 growth versus 9.1 but see you went with the 2 physical so I'm wondering if I'm misunderstanding the data
You can't just add them up
A lot of points great points. I'll start off by saying about this that the growth in CA/Attributes screenshot uses Defending, Physical Training, and Ground Defense. I think that you can extrapolate this data and apply it to 3 different categories for attribute growth.
The 1st type is I what will call General Training which includes all of the trainings under General tab except Physical.
The Physical Training which includes all high-intensity trainings that focus on Physical attribute growth.
The last and by far the biggest group of trainings are specialized trainings. These are all the trainings under Attacking, Defending, Technical, and Goalkeeping training.
I've included the trainings I have so far because they are clearly above the rest in the results. That's why they are the foundation of the Meta Attribute Weighted Training Sessions. However finding the most efficient specialized trainings will take some work as the results are pretty close and some trainings synergize better with more players better than others. To be honest, other than the Foundational 11 there isn't a "correct answer" and the goal of the posts will be to allow players to figure out their perfect schedule for themselves although I will give my best answer as well.
I'll give my take on the Physical training amount after work today.
On the topic of physical training, everything has to do with the my personal weights and how it effects the results I get. You can read how I made the weights for my experiments below or just read the TL;DR if you can trust the weights I've cooked
.
.
How the Weights for my Training experiments were formed I made weights that are based on the Chinese Machine Learning FM Study. It's probably the most quality test on attributes we might ever see for some time due to sample size and the process that is AI machine learning. You can find the weights from the experiment called ykykyk balanced here from @Mark's post on it.
Here is an example of the Full Backs for ykykyk balanced.
There was a problem with just copying these weights 1 for 1 for me.
I feel that Physicals were not weighted important enough. When I was playing FM using these weights a lot of players I was being recommended(as a lower league club) were over the hill and were really declining the league in the 2 most important attributes Pace and Acceleration, but had great Technical/Mental skill to "make up" for it.
Then I used the weights that @ZaZ created all ratings under 50 minus 25 weights which @Mark balanced here. These weights fixed this problem for me by decreasing weights on the less important attributes which coincidentally really fixed the importance for the Physical Attributes.
Here is an example of the Full Backs for all ratings under 50 minus 25.
However, the problem with these weights is that I'm using the top tactic in the Patch 24.2.0 (v2.0) - Tactic Testing Table. The Machine Learning Weights are based off of this tactic from FM22. The meta tactics for FM24 are different from 2 years ago with the biggest difference being the role of the Fullback. The IWB played more like a recycle the ball DM which shows in these weights. Now the Fullback has the responsibilities of a Winger, and needs more weights to represent this.
Here is an example of changes I made to my weights to attempt to accomplish this which you can find here
For example, I raised some weights in things like Passing, Crossing, and Technique to what I felt best emulated the role of a FB in that tactic.
I made tweaks like this to these weights to fit the top tactic for FM 24 for not just Fullback but the other roles as well(but should work other 4231 tactics).
TL;DR The weights are my attempt at mixing Chinese Machine Learning with prioritizing Physical Attributes to fit FM 24.
Because of that, almost all of the weights above 70 in my weights are trained with any sort of Physical training. I have screenshots here underlining these attributes in green.
Looking at these weights, Physical Attributes are weighted as importantly as the center of the universe. Because of that I try to maximize their growth as much as reasonably possible instead of being efficient with just 1 or 2 sessions.
If the results from physical training showed that more is better then I would do more, but it seems the maximum training sessions that grow Physical sessions in a week is 3. So that's why in the foundational 11 training schedule there are 3 physical sessions in it.
Han106 said: On the topic of physical training, everything has to do with the my personal weights and how it effects the results I get. You can read how I made the weights for my experiments below or just read the TL;DR if you can trust the weights I've cooked
.
.
How the Weights for my Training experiments were formed I made weights that are based on the Chinese Machine Learning FM Study. It's probably the most quality test on attributes we might ever see for some time due to sample size and the process that is AI machine learning. You can find the weights from the experiment called ykykyk balanced here from @Mark's post on it.
Here is an example of the Full Backs for ykykyk balanced.
There was a problem with just copying these weights 1 for 1 for me.
I feel that Physicals were not weighted important enough. When I was playing FM using these weights a lot of players I was being recommended(as a lower league club) were over the hill and were really declining the league in the 2 most important attributes Pace and Acceleration, but had great Technical/Mental skill to "make up" for it.
Then I used the weights that @ZaZ created all ratings under 50 minus 25 weights which @Mark balanced here. These weights fixed this problem for me by decreasing weights on the less important attributes which coincidentally really fixed the importance for the Physical Attributes.
Here is an example of the Full Backs for all ratings under 50 minus 25.
However, the problem with these weights is that I'm using the top tactic in the Patch 24.2.0 (v2.0) - Tactic Testing Table. The Machine Learning Weights are based off of this tactic from FM22. The meta tactics for FM24 are different from 2 years ago with the biggest difference being the role of the Fullback. The IWB played more like a recycle the ball DM which shows in these weights. Now the Fullback has the responsibilities of a Winger, and needs more weights to represent this.
Here is an example of changes I made to my weights to attempt to accomplish this which you can find here
For example, I raised some weights in things like Passing, Crossing, and Technique to what I felt best emulated the role of a FB in that tactic.
I made tweaks like this to these weights to fit the top tactic for FM 24 for not just Fullback but the other roles as well(but should work other 4231 tactics).
TL;DR The weights are my attempt at mixing Chinese Machine Learning with prioritizing Physical Attributes to fit FM 24.
Because of that, almost all of the weights above 70 in my weights are trained with any sort of Physical training. I have screenshots here underlining these attributes in green.
Looking at these weights, Physical Attributes are weighted as importantly as the center of the universe. Because of that I try to maximize their growth as much as reasonably possible instead of being efficient with just 1 or 2 sessions.
If the results from physical training showed that more is better then I would do more, but it seems the maximum training sessions that grow Physical sessions in a week is 3. So that's why in the foundational 11 training schedule there are 3 physical sessions in it.
Congratulations for the work!
There's one thing I've never understood: why in Chinese Machine Learning FM Study on @ZaZ tactic, stamina is indicated as an important attribute for all roles, but in Player Attributes Testing Table it isn't?
Stuaret said: Congratulations for the work!
There's one thing I've never understood: why in Chinese Machine Learning FM Study on @ZaZ tactic, stamina is indicated as an important attribute for all roles, but in Player Attributes Testing Table it isn't?
I do still think that the ratings still view Stamina as important, but my mindset is that I don't think that Stamina is at the level of importance of say Pace and Acceleration for all positions.
I based the Stamina ratings on movement data from the tactic, which involved the stats Miles Traveled per 90, High Intenstity Sprints per 90 and Presses per 90 relative to Fullbacks. I also used my own experiences using the tactic in my sunday-league no badge pentagon journeyman save. I have been able to manage the health of the squad by using rotation and by using a lower intensity version of the tactic when I'm dominating the match. Usually, the only position have to sub in my matches are the fullbacks, whose Stamina weight is already at the max in my ratings. I personally think if they aren't in Red by minute 80 or so, the weights have gotten me the players that fit this goal while also maximizing pace/acceleration as much as possible.
Of course not all players might not agree with every number from the weights for every position, however as long as you find a system of analyzing players that works that's all that matters. I do hope that the ratings are close enough to people's priorities so that they still learn something regardless. For Stamina in particular, I don't think changing the weights to their original would've affected the physical training rankings from part 1.
Han106 said: But cheese pizza is boring and we still have 4-7 sessions that we can fill.
Max said that we don't need to fill the empty spaces and he is right. I had worse training results when I used his training schedule and filled wednesday. That's why I'm looking forward to your results. Match practice performed quite well in my tests.
Meriten said: Max said that we don't need to fill the empty spaces and he is right. I had worse training results when I used his training schedule and filled wednesday. That's why I'm looking forward to your results. Match practice performed quite well in my tests.
2 points I'll make regarding this.
Point 1: Convenience Not Maximization
EBFM's training schedule is made to be plug and play, convenient and efficient within its restrictions, not to maximize training growth. I will touch on this in my 2nd point.
The convenience is the reason why he doesn't fill in more trainings. There are 21 slots available in a training week. 11 are used for trainings. 6 are taken up by Match days, and it doesn't you actually have a match that day. For example, I play in a small league in India where I only have 1 match a week. 4 are used for if you have an away match and have to travel. It doesn't consider whether or not you have a extra slot available because you have a home match that week, or you live in England and Away matches don't require 2 training slots worth of travel.
Point 2: More Is Better
Then the question arises, is more trainings actually better? From this video from EBFM himself(Although it's from FM 22), he shows this table.
It shows growth when you have X amount of sessions in a training week and the results are clear that more is better. How much more I don't know, but you can see that 18 sessions a week does have a measurable increase compared to 10 and very likely 11 sessions. This is why it is worth it to try and fill up slots where available.
Hmm, I did see someone's post on running tests on it and saying that more might not be better so I decided to try and test it myself before moving forward through these tests.
The Test file is attached linked here.
From the test file, I have these players found in FM Genie Scout using the best.grf. I will simply run the tests 5 times for 4 teams and I will report hopefully soon.
I'm using EBFM's testing league with these changes made to it.
Set 1 MC to DM
Name Set to DM
Set focus for position in Tactic
Change tactic to Katana
Week 1 set to Rest
Import EBFM Schedule
Set all training weeks to EBFM
Updated file as I didn't set position focus for Team D
EDIT: Version 3 18:59 PM PT:
Team A/B AMC position focus was listed as Shadow Striker instead of Attacking Mid
One of Team B's names was DM instead of DMC
Stay on Feet was enabled for all team
This is the results from the first of 5 tests. Also noted injuries down for medical context, this spreadsheet will be filled out within the next hour.
Ok the spreadsheet is complete for the baseline tests of EBFM. It has stats like difference between Starting% and End%, Total %, Potential Rating, difference between End % and Potential %, and injury stats.
This took 2 hours.
Thanks for sharing the results of the testing. What was the answer? Does more training sessions promote more growth?
something popped into my mind while watching EBFM's series that I guess might be relevant here. A lot of these tests look to maximise CA growth per time from training. Given that:
-Attributes 'cost' different amounts of CA depending on position.
-Players grow faster the larger the difference between their CA and PA.
-We can estimate the quality of a player with the attribute weightings.
I think it might be better to maximise the value (weighted attribute) while minimising CA growth.
Another way of looking at is that we want to 'spend' our CA->PA growth in the most efficient way to maximise the weighted attribute values. My feeling is that broad/general training programs like EBFMs will tend to grow most attributes evenly. This is probably the fastest way to grow CA overall, but I think you will get higher quality players if you can focus the attribute growth on highly weighted attributes.
@roman Well, I would need to add more sessions and do the experiments again, i just made a baseline of just doing EBFM's so that we can compare later to not just mine, but test other training schedules in the community as well. I'm a bit tired so I will do so more tests tomorrow. But yes I agree with your thoughts on efficiency of training.
Looking at the data point, Diff from Baseline, I feel that the growth of GKs and FBs are low comparatively to the rest of the positions. Looking at Room to grow, which took the Current % after a year compared to FM Genie Scout's Pot Rating(which may or may not be reliable), DCs and DMs are all lagging behind compared to their attacking position counterparts.
Attacking unit players seem to be doing very well in all metrics. On the other hand, I believe Defending/Goalkeeping players might need more. I'll try adding 2 sessions on the Wednesday designed to try and improve these metrics for those 2 groups. But would moving Defensive players to the Attacking unit work as well?
I will try to test both tomorrow.
roman said: This is probably the fastest way to grow CA overall, but I think you will get higher quality players if you can focus the attribute growth on highly weighted attributes.
There is a big problem. One example: Anticipation is good but Decision is bad in context of CA efficiency according to the test results on fm-arena. But there is no training session which trains more Anticipation but less Decision.
Excerpt from MaxEBFM results:
Another example is Acceleration/Pace (both are good) and Agility (bad). I tried to beat the system by using a good and a bad coach but that doesn't work either. Anticipation/Decision is trained by the same coach and Acceleration/Pace/Agility is trained by the same coach too.
Another problem is that we don't know all good attributes. Stamina e.g. is bad according to fm-arena test results but good according to ykykyk.
Meriten said: There is a big problem. One example: Anticipation is good but Decision is bad in context of CA efficiency according to the test results on fm-arena. But there is no training session which trains more Anticipation but less Decision.
Excerpt from MaxEBFM results:
Another example is Acceleration/Pace (both are good) and Agility (bad). I tried to beat the system by using a good and a bad coach but that doesn't work either. Anticipation/Decision is trained by the same coach and Acceleration/Pace/Agility is trained by the same coach too.
Another problem is that we don't know all good attributes. Stamina e.g. is bad according to fm-arena test results but good according to ykykyk.
This is only a thought and totally unproven, but the machine learning was using overall, that is combinations whereas fm arena used single attributes.
Now I know that the machine learning was done with just one tactic over one year, but to me it seems more reliable.
The changes by year for attributes seem marginal. As Rampaging Roy Slavin would say, make of that what you will.
Han106 said: On the topic of physical training, everything has to do with the my personal weights and how it effects the results I get. You can read how I made the weights for my experiments below or just read the TL;DR if you can trust the weights I've cooked
.
.
How the Weights for my Training experiments were formed I made weights that are based on the Chinese Machine Learning FM Study. It's probably the most quality test on attributes we might ever see for some time due to sample size and the process that is AI machine learning. You can find the weights from the experiment called ykykyk balanced here from @Mark's post on it.
Here is an example of the Full Backs for ykykyk balanced.
There was a problem with just copying these weights 1 for 1 for me.
I feel that Physicals were not weighted important enough. When I was playing FM using these weights a lot of players I was being recommended(as a lower league club) were over the hill and were really declining the league in the 2 most important attributes Pace and Acceleration, but had great Technical/Mental skill to "make up" for it.
Then I used the weights that @ZaZ created all ratings under 50 minus 25 weights which @Mark balanced here. These weights fixed this problem for me by decreasing weights on the less important attributes which coincidentally really fixed the importance for the Physical Attributes.
Here is an example of the Full Backs for all ratings under 50 minus 25.
However, the problem with these weights is that I'm using the top tactic in the Patch 24.2.0 (v2.0) - Tactic Testing Table. The Machine Learning Weights are based off of this tactic from FM22. The meta tactics for FM24 are different from 2 years ago with the biggest difference being the role of the Fullback. The IWB played more like a recycle the ball DM which shows in these weights. Now the Fullback has the responsibilities of a Winger, and needs more weights to represent this.
Here is an example of changes I made to my weights to attempt to accomplish this which you can find here
For example, I raised some weights in things like Passing, Crossing, and Technique to what I felt best emulated the role of a FB in that tactic.
I made tweaks like this to these weights to fit the top tactic for FM 24 for not just Fullback but the other roles as well(but should work other 4231 tactics).
TL;DR The weights are my attempt at mixing Chinese Machine Learning with prioritizing Physical Attributes to fit FM 24.
Because of that, almost all of the weights above 70 in my weights are trained with any sort of Physical training. I have screenshots here underlining these attributes in green.
Looking at these weights, Physical Attributes are weighted as importantly as the center of the universe. Because of that I try to maximize their growth as much as reasonably possible instead of being efficient with just 1 or 2 sessions.
If the results from physical training showed that more is better then I would do more, but it seems the maximum training sessions that grow Physical sessions in a week is 3. So that's why in the foundational 11 training schedule there are 3 physical sessions in it.
I have adjusted these and tested them for a few seasons, seems that you do find the right players for the Katana v3.1 tactic on the tests.
Feel free to look and please provide a feedback
https://www.mediafire.com/file/0pbx7cv1hsdtnta/attributes+doru.grf/file
doru228 said: I have adjusted these and tested them for a few seasons, seems that you do find the right players for the Katana v3.1 tactic on the tests.
Feel free to look and please provide a feedback
https://www.mediafire.com/file/0pbx7cv1hsdtnta/attributes+doru.grf/file
Well I'm glad that my weight balance has some merit to it! Thanks!
Han106 said: Well I'm glad that my weight balance has some merit to it! Thanks!
well you done 75% of the work, i just tweaked it a little bit.
@Han106 there's something that i am wondering tho, if i should include consistency in the ratings, since a consistent player does get better ratings ingame (tested with same player, just consistency was modified over one season. difference was like 1.5 points between 1 consistency and 20, and about 0.75 between 10 and 20.)
results were like this:
1 consistency 6.50
10 consistency 7.12
20 consistency 7.76
all of the above over 40 games simulated, frozen attributes and no injuries on
doru228 said: well you done 75% of the work, i just tweaked it a little bit.
@Han106 there's something that i am wondering tho, if i should include consistency in the ratings, since a consistent player does get better ratings ingame (tested with same player, just consistency was modified over one season. difference was like 1.5 points between 1 consistency and 20, and about 0.75 between 10 and 20.)
results were like this:
1 consistency 6.50
10 consistency 7.12
20 consistency 7.76
all of the above over 40 games simulated, frozen attributes and no injuries on
You can include it in yours. However I don't include weights that aren't visible to the player. Seems great otherwise
Meriten said: There is a big problem. One example: Anticipation is good but Decision is bad in context of CA efficiency according to the test results on fm-arena. But there is no training session which trains more Anticipation but less Decision.
Excerpt from MaxEBFM results:
Another example is Acceleration/Pace (both are good) and Agility (bad). I tried to beat the system by using a good and a bad coach but that doesn't work either. Anticipation/Decision is trained by the same coach and Acceleration/Pace/Agility is trained by the same coach too.
Another problem is that we don't know all good attributes. Stamina e.g. is bad according to fm-arena test results but good according to ykykyk.
Just as a reminder, the fact that one attribute did not get impressive results in the table of FM-Arena does not mean that attribute is useless. For example, stamina might be impactful only for two or three players, like DM and WBs, which represents less than 30% of the team. In that case, even if Stamina is as impactful as Pace and Acceleration for those positions, it will still give worse results than attributes that have impact across all roles.
ZaZ said: Just as a reminder, the fact that one attribute did not get impressive results in the table of FM-Arena does not mean that attribute is useless. For example, stamina might be impactful only for two or three players, like DM and WBs, which represents less than 30% of the team. In that case, even if Stamina is as impactful as Pace and Acceleration for those positions, it will still give worse results than attributes that have impact across all roles.
No way that Stamina can be impactful as Acceleration and Pace for any positions.
In the FM-Arena test Stamina attribute was increased by 5 points for every positions including the DMs and WBs and that improved the result just by 2 points from 60 pts to 62 pts, which honestly can be just the RNG but when Acceleration or Pace attributes were increased by 5 then it improved the result by 21 points from 60 pts to 81 pts.
So no way Stamina attribute or any other attribute can be as impactful as Pace and Acceleration for any position.
Pace - https://fm-arena.com/tactic/7528-5-points-to-pace-for-all-positions/
Acceleration - https://fm-arena.com/tactic/7527-5-points-to-acceleration-for-all-positions/
Stamina - https://fm-arena.com/tactic/7557-5-points-to-stamina-for-all-positions/
Here's another test when Acceleration and Pace was tested, specifically, for Central Midfielder position - https://fm-arena.com/thread/5351-should-you-follow-the-highlighted-attributes-of-the-roles/
ZaZ said: Just as a reminder, the fact that one attribute did not get impressive results in the table of FM-Arena does not mean that attribute is useless. For example, stamina might be impactful only for two or three players, like DM and WBs, which represents less than 30% of the team. In that case, even if Stamina is as impactful as Pace and Acceleration for those positions, it will still give worse results than attributes that have impact across all roles.
Decision is for every position more expensive than Anticipation (except SC) but the results on the tables are worse in comparision to Anticipation (FM22/FM23/FM24). Maybe there is one position (or even two positions) where Decision has a better cost-benefit ratio than Anticipation. But that doesn't matter. Why? Because you can't train Decision just for one position. You always must train Decision for all positions:
So a high value for Decision in a training session is always a bad thing.
Meriten said: Decision is for every position more expensive than Anticipation (except SC) but the results on the tables are worse in comparision to Anticipation (FM22/FM23/FM24). Maybe there is one position (or even two positions) where Decision has a better cost-benefit ratio than Anticipation. But that doesn't matter. Why? Because you can't train Decision just for one position. You always must train Decision for all positions:
So a high value for Decision in a training session is always a bad thing.
If you give a player Additional Focus of Final Third they will train both Composure and Decisions.
Lapidus said: No way that Stamina can be impactful as Acceleration and Pace for any positions.
In the FM-Arena test Stamina attribute was increased by 5 points for every positions including the DMs and WBs and that improved the result just by 2 points from 60 pts to 62 pts, which honestly can be just the RNG but when Acceleration or Pace attributes were increased by 5 then it improved the result by 21 points from 60 pts to 81 pts.
So no way Stamina attribute or any other attribute can be as impactful as Pace and Acceleration for any position.
Pace - https://fm-arena.com/tactic/7528-5-points-to-pace-for-all-positions/
Acceleration - https://fm-arena.com/tactic/7527-5-points-to-acceleration-for-all-positions/
Stamina - https://fm-arena.com/tactic/7557-5-points-to-stamina-for-all-positions/
Here's another test when Acceleration and Pace was tested, specifically, for Central Midfielder position - https://fm-arena.com/thread/5351-should-you-follow-the-highlighted-attributes-of-the-roles/
It was just an example.
By the way, I am testing the schedule below right now. Keep in mind that it requires training with half intensity, else it will lead to a high number of injuries. I also discovered a glitch on training that allows you to "super rest" players (almost 40% more recovery of condition and fatigue than just full rest schedule), but I am still investigating it to understand exactly how it works.
ZaZ said: It was just an example.
By the way, I am testing the schedule below right now. Keep in mind that it requires training with half intensity, else it will lead to a high number of injuries. I also discovered a glitch on training that allows you to "super rest" players (almost 40% more recovery of condition and fatigue than just full rest schedule), but I am still investigating it to understand exactly how it works.
What do you think about individual training? Are they worth using?