johnconnerson said: Have you (or anyone else) done research into how wage demand is calculated? I'm curious if there is a way to estimate how much a player "should" make (based on age, CA, reputation, or whatever else affects it). This would be helpful for quickly understanding whether or not I can afford a player/if a player is undervalued by the market. Expand
No. Contract negotiations have not really been tested, mainly because noone has made an easily scalable and reproduceable testing framework.
Since the wage budget is near the cap, the player reduced his demands by nearly 75%.
Though, this may well be only for contract renewals (again, never tested).
There are some other comments in that post that could be tested (for instance, putting/keeping low salary players with the Star Player squad status, and see if it influences salary negotiations with actual players that could be Star Players for that team.
LightningFlik said: Now I'm not saying that the game uses the same aggregation/scaling in the match engine--the only place that I've found it make a difference is the coach report (see below; this is me hacking the return value from the function so it thinks groups 9, 8 and 3 are this player's best)--but it's worth bearing in mind. I'm going to keep this line of enquiry up and I'll report back anything I learn. Expand
That is really interesting. It could well be simply the calculations on which message to output for the scout's report.
I do remember when the original attribute testing was done, back in FM23, I asked @Zippo if testing attribute combinations - as opposed to individual attributes - would show that there would be attribute combinations that would synergize with each other, making them far more effective in combination with specific other attributes, then by themselves, and Zippo said there was no difference in how they scored, either being tested separately or together, which I took to presume that testing, for instance, 16 Heading + 16 Jumping Reach together, and seeing the impact of Points, Goals Scored and Goals Against, would not be different from the average of the sum of the result of the testing 16 Heading (with 10 Jumping Reach), with the result of the test of 16 Jumping Reach (with 10 Heading).
Looking at these equations, and how they are aggregating attributes together, I do wonder if that's actually (still?) the case.
I would presume it would be some sort of equation between:
1. Current Club reputation vs target club reputation 2. Current League reputation vs target League reputation (and future, taking promotions/relegations into account) 3. Player reputation 4. Player happiness 5. Some attributes (Loyalty, Ambition, Professionalism, Adaptability, Pressure - This interacts with most other scenarios involving player happiness, if they aren't getting playtime, club was relegated, contract expiring, wants a new challenge, club isn't matching ambitions, pressure from interest demonstrated by other clubs etc.) 6. Current city attractiveness vs target city attractiveness (I presume this is calculated based on which city the main stadium is located) 7. Player's transfer status 8. Club's average CA rating vs player's CA rating 9. Player's age 10. Current club's current continental competition vs target club's current continental competition 11. Player's favorite club and personnel 12. Current player's salary vs target club's salary budget (or average salary for Star Player squad status?) 13. Current player's squad role vs target club's hypothetical squad role 14. If leaving due to contract expiring, then target club's whole package vs other hypothetical clubs future interest when user is free (Player wants to wait and see if someone other club will show an interest, rather than immediately signing for target club)
EDIT: Because very few of these things are directly influenceable by the player (as opposed to which tactics, training or players are better), testing this would not be useful for the vast majority of players.
It would be useful to create a custom league where after each game, there is a penalty shootout, and test the attributes (including hidden ones like Pressure which should play a part) to see which ones are better for penalties, so if the player gets to a shootout, he can choose the actual best players.
GeorgeFloydOverdosed said: I presume it got at least most of them right, but now I can't be certain some weren't miscalculated by ChatGPT. Expand
That is easy enough: You run the calculation through multiple AIs simultaneously to sanity crosscheck calculations. There are tools online for that. That way you remove potential errors from the most critical part of your analysis.
GeorgeFloydOverdosed said: I was skeptical, but this does seem to be the case. In fact I tested this and after a three-year loan to inactive England Div 6, the 140 CA player remained at 140 CA even though the distribution of his CA changed and his PA was 200.
A 88 CA player (148 PA) I put in inactive Div 6, grew to 96 CA by contrast. An identical player I put in active Div 3, grew to 94 CA but only had ~1/4 the games. An identical player I put in active Div 5, grew to 117 CA.
All players were 20 years old starting.
I will be doing more testing of this, but so far it seems like CA growth is hardcapped by league quality or reputation. And given the active Div 5 player did a lot better than inactive Div 6, maybe they changed the inactive league growth in FM24.. or at least, whatever's causing it, active Vanarama National is simply far better to loan to than inactive Vanarama National North.
This also has some implications beyond loans. It would affect training in general, and I guess suggests optimizing for low CA is even more important. Expand
There is potentially a simpler solution: Check the (player with the) maximum CA of the league, the average and the mean CA of the league.
See if the improvement stops when those thresholds are reached.
GeorgeFloydOverdosed said: Win rate/Relative strength of opposition = No difference Player morale Perfect > Okay > Abysmal = 17 > 17.04 > 17.29 CA
Since win rate and morale make no difference, I think it's reasonable to assume match rating in and of itself makes no difference. Expand
Exactly what I presumed. I remember EBFM testing win rate and seeing it had no difference. But I didn't remember the morale.
Either way, I think what is happening is someone along the chain of information confused "match experience" which is the term used by EBFM for matches played in a season, to Match Rating. I'm just not sure where people are getting ideal ratings from.
Panneton0 said: Optimal training + U18/U23 appearances vs Loaner club's training + First team appearances Expand
Since you can min-max players attribute development in your club, and you can't while they are on loan, you should generally only loan someone once playing time is crucial to keep developing (and match/league reputation is crucial to develop the player) to offset your optimized training.
match practice doesn't decrease condition 1 x match practice recovers condition same rate as all rest 1 x match practice doesn't boost match sharpness (identical to all rest)
2 x match practice + quickness on one day, is identical to full rest in regards to condition recovery & match sharpness
What you might not know is that condition recovery doesn't have a randomness factor, unlike many other FM mechanics, so there shouldn't be any need for a ton of tests.
Seems EBFM found this back in 2022, and additionally found that match practice actually negatively impacts match performances, even though it has no condition impact.
This is strange. Maybe the 'injury risk' and 'fatigue' labels are also falsehoods and the whole training meta needs a rethink if one is not looking solely for attribute gain. Expand
If corroborated, that's great to know.
I would presume (complete speculation though), that these heavier training sessions have a higher percentage RNG injury trigger roll than lighter ones, and that must be multiplied by the player's condition percentage when it's lower, increasing the risk of rolling a positive for injury for the player.
Hypothetically, instead of every training resolution there are injury rolls for every single player, a more simplistic system would be: First roll: Team-wide injury roll, that if triggered positively: Second roll: The game rolls who the injury will land on, with added weights for players with lower condition, higher injury proneness, and repeated injuries, and once it has selected a player. Third roll: Rolls the chance of which injury it's going to be, based on the training session being resolved, and if the user already is susceptible to a specific injury. Four roll: Rolls severity of injury, within injury type range. Then the game rolls again for another team-wide injury roll, as (if I remember correctly) on rare ocasions, it is possible to get two injuries while resolving a single training session.
In any case, if: 1. Lower individual player condition is indeed an added individual risk factor for that player getting an injury, and 2. The game rolls injuries on an individual basis, as opposed to a team basis (and AFAIK, there's no way of testing whether this is the case)
Basically we should always put rest sessions after the game, and put training sessions as far away from the previous game as possible, as long as it is within the same week, to ensure players have as high condition as possible before they start going through the training sessions.
EDIT: An extremely useful data point that can be tested is: 1. From an injury risk perspective, does stacking Attacking, Match Practice and Quickness training on the same training day (immediately before the game), leads to more, less, or equal amount of injuries during a season, as opposed to spreading one of each on different days. 2. From a player development perspective, is there a measurable difference in attribute growth between stacking all three on the same day, or spreading them out on different days?
The answers to this could basically lead to training sessions being stacking all three trainings on the day immediately before a match, using the rest of the days of the week for Rest and Recovery sessions.
harvestgreen22 said: All of these training schedules have a very low risk of injury, Even if you suffer three Match a week, don't worry about the impact of the lack of training, set all week rest, you lose nothing Expand
A couple of quick questions to harvestgreen22 or someone that knows.
The training sessions generally show lowering "Condition"
But from my observations, condition always improves after each day with just training, and only lowerswhen playing matches (or getting injured, but that's beside the point). So my question, has it ever been tested whether these trainings actually: 1. Lower condition? OR 2. Training sessions with heavier condition reduction don't allow the players to recover condition as much as training sessions with lighter condition reduction?
The reason why I'm asking is simple: I'm unsure whether I should schedule my training sessions like this:
or like this:
If trainings don't lower condition, then putting a rest day in between sessions makes no sense, and it is always ideal to put all rest sessions immediately after a game, and the training sessions as late as possible (like in the second case).
If they do lower condition, then adding a rest day in between training sessions to allow the players to recover, can make sense to lower risk of injuries.
Hopefully my question is understandable.
如果你看不懂我的意思，我也会中文，所以如果我必须写汉字，我可以。(This says that I can also write in Chinese in case harvestgreen22 doesn't understand).
Orion said: Keep in mind that this is for difference between actual attribute value and the average for the league. So for example if average finishing was 10, the higher than that would result in -0.000456 per every finishing point. So if the average was 10 and players value was 20 it still results in lowering average rating by 0,00456. So for features with coefficients this small you can basically say that they just don't make any difference. Finishing having negative coefficient this low could be possibly just 0. Simply for such numerical models it's very unlikely to have coefficient equal exactly 0. Tl;dr coefficients this low are basically noise and they don't contribute either way into target variable. Expand
Yes, the underlying assessment that for strikers (the one position where the attribute would be essential to successfully accomplish the main task of the position), having above average finishing is does not make any meaningful difference in average rating (knowing that scoring goals is practically the one certain and main way of substantially increasing the average rating for strikers) is mind boggling.
This one experiment was also conducted with the same underlying parameters (A very significant amount of leagues and divisions, running for 11 seasons)?
Orion said: This is the result for linear model for STC in FM23 that uses all the features. Where would you put the line which features are still important and which are not? Expand
This can't be right...
So according to your results, STC having higher Finishing equals having worse average rating?
FREVKY said: I believe there is a typo in DC coefficient:
It should be 0,015882, right?
Also you said:
But I can see that crossing is still a valid coefficient in your lists on a number of positions. Did you mean corners by any chance? Expand
Both correct. He has sorted attributes by highest to lowest, and the only one that is a massive outlier and is not sorted correctly. In regards to Crossing, several positions have Crossing in their most important attributes, so it's definitely Corners.
Jolt said: That is extremely surprising. So every combination tried merely increases cumulatively with the increase of the combined chosen attributes, with no statistical difference from the sum of the chosen attributes?
EDIT: So what does testing of a 20 Pace and 20 Acceleration look like, in terms of points, goals for and goals against, as opposed to a 20 Pace 10 Acceleration, or a 10 Pace 20 Acceleration?
EDIT 2: Or perhaps even a more clear example of synergy: "Goals for" with both Crossing and Heading in 20. For an attacking team to fully exploit high heading attributes, accurate crossings with high Crossing attributes generally means higher opportunities of goal scoring, as opposed great crosses to mediocre heading players or mediocre crosses that don't frequently reach great heading players. I wonder what's the testing difference specifically goals for, between 20 Crossing 20 Heading, and 20 Crossing 10 Heading, and 10 Crossing 20 Heading. Expand
What's the practical difference in points, Goals For, Goals Against, between the current tests where only one attribute gets raised, and when two with theorerical synergies get raised (Pace+Acceleration; Crossing+Heading; or others such as Heading+Jumping Reach). I'm curious to understand in practical terms what "there's no difference" means.
Zippo said: We've tired testing attributes in combinations, assuming that one attribute might somehow amplify another, but we found out that wasn't the case so there's no point testing attribute combinations. Expand
That is extremely surprising. So every combination tried merely increases cumulatively with the increase of the combined chosen attributes, with no statistical difference from the sum of the chosen attributes?
EDIT: So what does testing of a 20 Pace and 20 Acceleration look like, in terms of points, goals for and goals against, as opposed to a 20 Pace 10 Acceleration, or a 10 Pace 20 Acceleration?
EDIT 2: Or perhaps even a more clear example of synergy: "Goals for" with both Crossing and Heading in 20. For an attacking team to fully exploit high heading attributes, accurate crossings with high Crossing attributes generally means higher opportunities of goal scoring, as opposed great crosses to mediocre heading players or mediocre crosses that don't frequently reach great heading players. I wonder what's the testing difference specifically goals for, between 20 Crossing 20 Heading, and 20 Crossing 10 Heading, and 10 Crossing 20 Heading.
Ultimately the major thing that matters from a developmental point of view is professionalism. Everything else is of secondary importance.
1. We know that the ideal numerical setup is one mentor to two mentees. 2. We know (roughly) the factors that make someone have more or less influence in the group. 3. We know which personalities have professionalism thresholds.
As such, this is a game of pairing up comparatively high professionalism established (playtime and age wise) players, with comparatively lower professionalism young players.
Because of the random nature of what goes up and down and the many attributes that can go up or down (or favoured moves gained or lost), depending on the differences between the mentor and the two mentees, mentoring shouldn't really be used for anything else, on the risks of backfiring and making the youngsters worse.
No. Contract negotiations have not really been tested, mainly because noone has made an easily scalable and reproduceable testing framework.
What has been found years ago was wage demands were tied to the club's wage budget, and lowering the wage budget in the Board screen, before starting contract negotiations would significantly lower wage demands. It is something you can do if you renew all your players contracts in bulk: https://www.reddit.com/r/footballmanagergames/comments/s259l2/how_to_lower_player_contract_demands_this_game_is/
Since the wage budget is near the cap, the player reduced his demands by nearly 75%.
Though, this may well be only for contract renewals (again, never tested).
There are some other comments in that post that could be tested (for instance, putting/keeping low salary players with the Star Player squad status, and see if it influences salary negotiations with actual players that could be Star Players for that team.
That is really interesting. It could well be simply the calculations on which message to output for the scout's report.
I do remember when the original attribute testing was done, back in FM23, I asked @Zippo if testing attribute combinations - as opposed to individual attributes - would show that there would be attribute combinations that would synergize with each other, making them far more effective in combination with specific other attributes, then by themselves, and Zippo said there was no difference in how they scored, either being tested separately or together, which I took to presume that testing, for instance, 16 Heading + 16 Jumping Reach together, and seeing the impact of Points, Goals Scored and Goals Against, would not be different from the average of the sum of the result of the testing 16 Heading (with 10 Jumping Reach), with the result of the test of 16 Jumping Reach (with 10 Heading).
Looking at these equations, and how they are aggregating attributes together, I do wonder if that's actually (still?) the case.
1. Current Club reputation vs target club reputation
2. Current League reputation vs target League reputation (and future, taking promotions/relegations into account)
3. Player reputation
4. Player happiness
5. Some attributes (Loyalty, Ambition, Professionalism, Adaptability, Pressure - This interacts with most other scenarios involving player happiness, if they aren't getting playtime, club was relegated, contract expiring, wants a new challenge, club isn't matching ambitions, pressure from interest demonstrated by other clubs etc.)
6. Current city attractiveness vs target city attractiveness (I presume this is calculated based on which city the main stadium is located)
7. Player's transfer status
8. Club's average CA rating vs player's CA rating
9. Player's age
10. Current club's current continental competition vs target club's current continental competition
11. Player's favorite club and personnel
12. Current player's salary vs target club's salary budget (or average salary for Star Player squad status?)
13. Current player's squad role vs target club's hypothetical squad role
14. If leaving due to contract expiring, then target club's whole package vs other hypothetical clubs future interest when user is free (Player wants to wait and see if someone other club will show an interest, rather than immediately signing for target club)
EDIT: Because very few of these things are directly influenceable by the player (as opposed to which tactics, training or players are better), testing this would not be useful for the vast majority of players.
That is easy enough: You run the calculation through multiple AIs simultaneously to sanity crosscheck calculations. There are tools online for that. That way you remove potential errors from the most critical part of your analysis.
A 88 CA player (148 PA) I put in inactive Div 6, grew to 96 CA by contrast.
An identical player I put in active Div 3, grew to 94 CA but only had ~1/4 the games.
An identical player I put in active Div 5, grew to 117 CA.
All players were 20 years old starting.
I will be doing more testing of this, but so far it seems like CA growth is hardcapped by league quality or reputation. And given the active Div 5 player did a lot better than inactive Div 6, maybe they changed the inactive league growth in FM24.. or at least, whatever's causing it, active Vanarama National is simply far better to loan to than inactive Vanarama National North.
This also has some implications beyond loans. It would affect training in general, and I guess suggests optimizing for low CA is even more important.
There is potentially a simpler solution: Check the (player with the) maximum CA of the league, the average and the mean CA of the league.
See if the improvement stops when those thresholds are reached.
Player morale Perfect > Okay > Abysmal = 17 > 17.04 > 17.29 CA
Since win rate and morale make no difference, I think it's reasonable to assume match rating in and of itself makes no difference.
Exactly what I presumed. I remember EBFM testing win rate and seeing it had no difference. But I didn't remember the morale.
Either way, I think what is happening is someone along the chain of information confused "match experience" which is the term used by EBFM for matches played in a season, to Match Rating. I'm just not sure where people are getting ideal ratings from.
只是想确认一下：
Just to confirm something:
是否进行过测试，看看与将所有训练集中在一天完成相比，将训练分散在一周内进行是否会导致整个赛季中出现更多伤病？
Has it been tested if spreading out the training:
Causes more injuries throughout the season than this:
vs
Loaner club's training + First team appearances
Since you can min-max players attribute development in your club, and you can't while they are on loan, you should generally only loan someone once playing time is crucial to keep developing (and match/league reputation is crucial to develop the player) to offset your optimized training.
And that is 24+ aged players, IIRC.
match practice doesn't decrease condition
1 x match practice recovers condition same rate as all rest
1 x match practice doesn't boost match sharpness (identical to all rest)
2 x match practice + quickness on one day, is identical to full rest in regards to condition recovery & match sharpness
What you might not know is that condition recovery doesn't have a randomness factor, unlike many other FM mechanics, so there shouldn't be any need for a ton of tests.
Seems EBFM found this back in 2022, and additionally found that match practice actually negatively impacts match performances, even though it has no condition impact.
This is strange. Maybe the 'injury risk' and 'fatigue' labels are also falsehoods and the whole training meta needs a rethink if one is not looking solely for attribute gain.
If corroborated, that's great to know.
I would presume (complete speculation though), that these heavier training sessions have a higher percentage RNG injury trigger roll than lighter ones, and that must be multiplied by the player's condition percentage when it's lower, increasing the risk of rolling a positive for injury for the player.
Hypothetically, instead of every training resolution there are injury rolls for every single player, a more simplistic system would be:
First roll: Team-wide injury roll, that if triggered positively:
Second roll: The game rolls who the injury will land on, with added weights for players with lower condition, higher injury proneness, and repeated injuries, and once it has selected a player.
Third roll: Rolls the chance of which injury it's going to be, based on the training session being resolved, and if the user already is susceptible to a specific injury.
Four roll: Rolls severity of injury, within injury type range.
Then the game rolls again for another team-wide injury roll, as (if I remember correctly) on rare ocasions, it is possible to get two injuries while resolving a single training session.
In any case, if:
1. Lower individual player condition is indeed an added individual risk factor for that player getting an injury, and
2. The game rolls injuries on an individual basis, as opposed to a team basis (and AFAIK, there's no way of testing whether this is the case)
Basically we should always put rest sessions after the game, and put training sessions as far away from the previous game as possible, as long as it is within the same week, to ensure players have as high condition as possible before they start going through the training sessions.
EDIT: An extremely useful data point that can be tested is:
1. From an injury risk perspective, does stacking Attacking, Match Practice and Quickness training on the same training day (immediately before the game), leads to more, less, or equal amount of injuries during a season, as opposed to spreading one of each on different days.
2. From a player development perspective, is there a measurable difference in attribute growth between stacking all three on the same day, or spreading them out on different days?
The answers to this could basically lead to training sessions being stacking all three trainings on the day immediately before a match, using the rest of the days of the week for Rest and Recovery sessions.
A couple of quick questions to harvestgreen22 or someone that knows.
The training sessions generally show lowering "Condition"
But from my observations, condition always improves after each day with just training, and only lowerswhen playing matches (or getting injured, but that's beside the point). So my question, has it ever been tested whether these trainings actually: 1. Lower condition? OR 2. Training sessions with heavier condition reduction don't allow the players to recover condition as much as training sessions with lighter condition reduction?
The reason why I'm asking is simple: I'm unsure whether I should schedule my training sessions like this:
or like this:
If trainings don't lower condition, then putting a rest day in between sessions makes no sense, and it is always ideal to put all rest sessions immediately after a game, and the training sessions as late as possible (like in the second case).
If they do lower condition, then adding a rest day in between training sessions to allow the players to recover, can make sense to lower risk of injuries.
Hopefully my question is understandable.
如果你看不懂我的意思，我也会中文，所以如果我必须写汉字，我可以。(This says that I can also write in Chinese in case harvestgreen22 doesn't understand).
Tl;dr coefficients this low are basically noise and they don't contribute either way into target variable.
Yes, the underlying assessment that for strikers (the one position where the attribute would be essential to successfully accomplish the main task of the position), having above average finishing is does not make any meaningful difference in average rating (knowing that scoring goals is practically the one certain and main way of substantially increasing the average rating for strikers) is mind boggling.
This one experiment was also conducted with the same underlying parameters (A very significant amount of leagues and divisions, running for 11 seasons)?
This can't be right...
So according to your results, STC having higher Finishing equals having worse average rating?
It should be 0,015882, right?
Also you said:
But I can see that crossing is still a valid coefficient in your lists on a number of positions. Did you mean corners by any chance?
Both correct. He has sorted attributes by highest to lowest, and the only one that is a massive outlier and is not sorted correctly. In regards to Crossing, several positions have Crossing in their most important attributes, so it's definitely Corners.
EDIT: So what does testing of a 20 Pace and 20 Acceleration look like, in terms of points, goals for and goals against, as opposed to a 20 Pace 10 Acceleration, or a 10 Pace 20 Acceleration?
EDIT 2: Or perhaps even a more clear example of synergy: "Goals for" with both Crossing and Heading in 20. For an attacking team to fully exploit high heading attributes, accurate crossings with high Crossing attributes generally means higher opportunities of goal scoring, as opposed great crosses to mediocre heading players or mediocre crosses that don't frequently reach great heading players. I wonder what's the testing difference specifically goals for, between 20 Crossing 20 Heading, and 20 Crossing 10 Heading, and 10 Crossing 20 Heading.
@Zippo Could you clarify this?
What's the practical difference in points, Goals For, Goals Against, between the current tests where only one attribute gets raised, and when two with theorerical synergies get raised (Pace+Acceleration; Crossing+Heading; or others such as Heading+Jumping Reach). I'm curious to understand in practical terms what "there's no difference" means.
That is extremely surprising. So every combination tried merely increases cumulatively with the increase of the combined chosen attributes, with no statistical difference from the sum of the chosen attributes?
EDIT: So what does testing of a 20 Pace and 20 Acceleration look like, in terms of points, goals for and goals against, as opposed to a 20 Pace 10 Acceleration, or a 10 Pace 20 Acceleration?
EDIT 2: Or perhaps even a more clear example of synergy: "Goals for" with both Crossing and Heading in 20. For an attacking team to fully exploit high heading attributes, accurate crossings with high Crossing attributes generally means higher opportunities of goal scoring, as opposed great crosses to mediocre heading players or mediocre crosses that don't frequently reach great heading players. I wonder what's the testing difference specifically goals for, between 20 Crossing 20 Heading, and 20 Crossing 10 Heading, and 10 Crossing 20 Heading.
1. We know that the ideal numerical setup is one mentor to two mentees.
2. We know (roughly) the factors that make someone have more or less influence in the group.
3. We know which personalities have professionalism thresholds.
As such, this is a game of pairing up comparatively high professionalism established (playtime and age wise) players, with comparatively lower professionalism young players.
Because of the random nature of what goes up and down and the many attributes that can go up or down (or favoured moves gained or lost), depending on the differences between the mentor and the two mentees, mentoring shouldn't really be used for anything else, on the risks of backfiring and making the youngsters worse.