Mrjoser said: Just want to say how much I appreciate the work you put into testing and letting us see behind the curtain. I like the influx of new research regarding attributes and training and whatnot.
I just find it hard to wrap my mind around the fact that this is currently the best Left back I can have (playing as Malmö 3rd season) Expand
Looks really amazing to be honest. He will be able to bully any winger just by raw physicality. I was saying for quite long time that for almost any case the attributes priority is Physical > Mental > Technical.
Orion said: Just to keep you guys up to date I've decided to make a small adjustment for the features that use two attributes - so the last one for MLR and the last one for Striker. I've decided to take square of a product from multiplication those features (SQR(Attribute1 * Attribute2) * Coefficient). This way you can keep it proportional to the coefficient. I don't know how it works with your files. I assume that for spreadsheet it's not an issue. What about GenieScout? Does it support other mathematical operations? If not I'd suggest just to get rid of those 2 features (since they are last important for given positions) and adjust others to sum up to 100% just like you did for the other positions. Expand
Updated the spreadsheet with your corrected rating evaluation for positions you mentioned. So now for striker it stacks up jumping reach, as it is taken into equation twice (most and least important coefficient - for the latter sqrt((jumping reach)^2) equals just jumping reach.
FREVKY said: Updated the spreadsheet with your corrected rating evaluation for positions you mentioned. So now for striker it stacks up jumping reach, as it is taken into equation twice (most and least important coefficient - for the latter sqrt((jumping reach)^2) equals just jumping reach. Expand I would like to contribute to the forum with an opinion. I apologize for my English. I'm using AI to translate. I want to thank the creator of the test and the improvements that were posted here on how to use it in Excel. I would like to confirm that using the table can lead to success in choosing players for our teams. I'm playing the second season with Chester. Managing to be champion in the first place. To assemble the cast for the second season, decide to use the Excel spreadsheet posted here with the following method. I first analyzed my team, and found the coefficients for each position. I searched in sky league 2, league one, efl, and premier league, for players who had higher coefficients than mine. If a player of mine performed with an average of 7.20, if I find a player with a better coefficient than my player, playing in a better league, everything indicates that he also performs better. Below are photos from the season so far. I am the candidate for relegation
Orion said: Just to keep you guys up to date I've decided to make a small adjustment for the features that use two attributes - so the last one for MLR and the last one for Striker. I've decided to take square of a product from multiplication those features (SQR(Attribute1 * Attribute2) * Coefficient). This way you can keep it proportional to the coefficient. I don't know how it works with your files. I assume that for spreadsheet it's not an issue. What about GenieScout? Does it support other mathematical operations? If not I'd suggest just to get rid of those 2 features (since they are last important for given positions) and adjust others to sum up to 100% just like you did for the other positions. Expand
Genie doesn't allow such mathematical operations unfortunately. Spreadsheet would be 1:1 replica of the coefficients. If you have time and patience i would suggest using the spreadsheet. You can still use the Genie ratings they are very close to spreadsheet.
FREVKY said: Updated the spreadsheet with your corrected rating evaluation for positions you mentioned. So now for striker it stacks up jumping reach, as it is taken into equation twice (most and least important coefficient - for the latter sqrt((jumping reach)^2) equals just jumping reach. Expand What do you mean? The effect of jumping reach should be much smaller than it was before.
Although honestly I think for now it would be best to just remove the double attributes for the time being. Too much confusion and desperately trying to make it work for very minimal effect.
Yarema said: What do you mean? The effect of jumping reach should be much smaller than it was before.
Although honestly I think for now it would be best to just remove the double attributes for the time being. Too much confusion and desperately trying to make it work for very minimal effect. Expand
It certailny is smaller than before, what I meant is that ST formula still covers jumping reach twice. I just wanted to point out that square root of "X to the power of 2" is "X", so there's no need for making it more complex than it should be from mathematical point of view.
FREVKY said: It certailny is smaller than before, what I meant is that ST formula still covers jumping reach twice. I just wanted to point out that square root of "X to the power of 2" is "X", so there's no need for making it more complex than it should be from mathematical point of view. Expand Yes but it's a general solution. MRL also has "Acc Agi" combo, interestingly without Agi itself as a solo attribute.
I rerun the model for those 2 positions to use only linear regressions so no we don't get issues with double attributes or anything like that. Results:
STC Attribute Coefficient Jum 0,020557 Pac 0,020096 Acc 0,014496 Cnt 0,013675 Bal 0,012043 Dri 0,010739 Vis 0,009392 Cmp 0,009161
MLR Attribute Coefficient Pac 0,020497 Dri 0,014018 Acc 0,012883 Cmp 0,012072 Vis 0,011542 Jum 0,011150 Cro 0,010598 Sta 0,009658
I'll update first post with the new values. I think this will save us a lot of confusion.
Orion said: I rerun the model for those 2 positions to use only linear regressions so no we don't get issues with double attributes or anything like that. Results:
STC Attribute Coefficient Jum 0,020557 Pac 0,020096 Acc 0,014496 Cnt 0,013675 Bal 0,012043 Dri 0,010739 Vis 0,009392 Cmp 0,009161
MLR Attribute Coefficient Pac 0,020497 Dri 0,014018 Acc 0,012883 Cmp 0,012072 Vis 0,011542 Jum 0,011150 Cro 0,010598 Sta 0,009658
I'll update first post with the new values. I think this will save us a lot of confusion. Expand
Are those coefficients for the ST a target striker or fast striker?
CBP87 said: Are those coefficients for the ST a target striker or fast striker? Expand
Yarema said: It doesn't differentiate by role Expand
This. It's just for position no matter the role. If I could somehow filter out players by their most played position and role I'd do it in a heartbeat.
Orion said: This. It's just for position no matter the role. If I could somehow filter out players by their most played position and role I'd do it in a heartbeat. Expand
So just to clarify, you're saying that jumping is the most important attribute for a striker?
I observed that for many positions, very similar type of player is the best. For example I have 3 pacy CDs with great jumping reach and well who would have guessed that, they would also rank as 1, 2 and 4 in my strikers lineup. So I think this weights give us great understanding what stats are most valuable, but only on very general scale. I for example don't believe that FB(att) will benefit from those 8 attributes the same way as IFB(Def). Or similar example wide playmaker VS inside forward.
Maybe the think that is missing in all these test is some atribute variability within the team. Like sure jumping reach is OP, but maybe it's enough to have 3-4 really good players in that and the team would benefit more if other players have better passing, or dribbling or whatever.
Mrjoser said: I observed that for many positions, very similar type of player is the best. For example I have 3 pacy CDs with great jumping reach and well who would have guessed that, they would also rank as 1, 2 and 4 in my strikers lineup. So I think this weights give us great understanding what stats are most valuable, but only on very general scale. I for example don't believe that FB(att) will benefit from those 8 attributes the same way as IFB(Def). Or similar example wide playmaker VS inside forward.
Maybe the think that is missing in all these test is some atribute variability within the team. Like sure jumping reach is OP, but maybe it's enough to have 3-4 really good players in that and the team would benefit more if other players have better passing, or dribbling or whatever. Expand
As said before, If I was able to filter players by their most played position and role I'd do it in a heartbeat. Sadly I don't know if there is even a possibility to extract such data in any useful way.
Kamas1 said: Vanarama south team, third season currently on sky bet 2 Expand
I see that, just like in my save, DMs tend to have lower ratings than the rest of the team. I thought that maybe my own DM is just not that good since his average rating was below 7 but on the other hand the others had worse average on that position.
Damn. 48 goals is really like a cheat code. Can you show this striker attributes?
I think consistently include momentum, taking into consideration the game. The full-backs are involved in aerial disputes, helping on a corner, or when the opposing goalkeeper covers a goal kick. Sometimes I just look at the value of the attribute found by the coefficients and compare it with the best player in the position in the league, or in the list of the best 11. or with the team that is favored to win the league.
robbr13 said: I think consistently include momentum, taking into consideration the game. The full-backs are involved in aerial disputes, helping on a corner, or when the opposing goalkeeper covers a goal kick. Sometimes I just look at the value of the attribute found by the coefficients and compare it with the best player in the position in the league, or in the list of the best 11. or with the team that is favored to win the league. Expand
That's exactly how the model works. If we wanted to use it 'properly' we should check the average value for given attributes in the league you're in. And then for the attribute value (the one you then multiply by coefficient) you should use the difference between players actual attribute value minus league average.
We use 'straightforwardly' just the attribute value to speed up things. Of course in general it induce the error but since we are comparing players that will play within the same league (all the players that will play for us) it's just much simpler and we're saying that we're ok with the possible error.
At the end of the day it's still better than rely on attributes that the game suggests.
I wonder if you could share the full lists of coefficients (not only the 8 most prominent ones). This way I can use them to make a squad analysis sheet using the coefficients to score your players for suitability on each of these positions.
DeFlow said: I wonder if you could share the full lists of coefficients (not only the 8 most prominent ones). This way I can use them to make a squad analysis sheet using the coefficients to score your players for suitability on each of these positions. Expand
Of course it's possible. I'd just have to generate model that have all the attributes. I choose to use only 8 to make it more usable. With 8 attributes you can even make yourself a simple spreadsheet to compare players. With 30-40-50 features a lot of them will literally have coefficients like 0,0001 so they won't bring much to the table. The question is what attributes you'll need. Only the visible ones? By visible I mean those from default player profile page. Because I extracted from the game around 50 features including attributes, hidden attributes, height, weight, player's league rank and foot (left/right) proficiency.
It is up to you how many coeffecients you want to extract. Those that are significantly lower than the ones above it could be excluded, but I am sure right now there are more coefficients that are almost as important as the 8 you have shared here?
Right now I am scoring the players on the coefficients for the positions you have extracted, but any coefficient can be incorporated in the spreadsheet.
DeFlow said: It is up to you how many coeffecients you want to extract. Those that are significantly lower than the ones above it could be excluded, but I am sure right now there are more coefficients that are almost as important as the 8 you have shared here?
Right now I am scoring the players on the coefficients for the positions you have extracted, but any coefficient can be incorporated in the spreadsheet. Expand
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?
Orion said: We use vastly different method. As far as I understand FM Arena attribute test has a testing league where they change certain attribute for every player in one team in that league and check the difference it makes. It's overall very good method. I think mine could be called closer to 'real' game environment - since I use data from 'real' leagues. So very simplified explanation is that players in certain position that have high Crossing attribute have high ratings. That's basically it. So the model pick that attribute as the one that correlates in those positions with player rating - so according to the model the higher the crossing the better player rating. If we give high crossing to every player it's kind of useless because a lot of position/roles do not utilize this. FM Arena attribute test looks for attributes that have highest impact on 'universal' level so they look for attributes that will benefit for the whole team, not just single player or single position. Expand
In that case shouldn't the comparisons be between players who play in that position in that league, rather than the average value across the entire league?
joshua said: In that case shouldn't the comparisons be between players who play in that position in that league, rather than the average value across the entire league? Expand
In 'general' model the average was calculated for every league for all the players. For models for each position average was calculated for every league for the players that play that certain position.
Mrjoser said: Just want to say how much I appreciate the work you put into testing and letting us see behind the curtain. I like the influx of new research regarding attributes and training and whatnot.
I just find it hard to wrap my mind around the fact that this is currently the best Left back I can have (playing as Malmö 3rd season)
Looks really amazing to be honest. He will be able to bully any winger just by raw physicality.
I was saying for quite long time that for almost any case the attributes priority is Physical > Mental > Technical.
Orion said: Just to keep you guys up to date I've decided to make a small adjustment for the features that use two attributes - so the last one for MLR and the last one for Striker. I've decided to take square of a product from multiplication those features (SQR(Attribute1 * Attribute2) * Coefficient). This way you can keep it proportional to the coefficient.
I don't know how it works with your files. I assume that for spreadsheet it's not an issue.
What about GenieScout? Does it support other mathematical operations? If not I'd suggest just to get rid of those 2 features (since they are last important for given positions) and adjust others to sum up to 100% just like you did for the other positions.
Updated the spreadsheet with your corrected rating evaluation for positions you mentioned. So now for striker it stacks up jumping reach, as it is taken into equation twice (most and least important coefficient - for the latter sqrt((jumping reach)^2) equals just jumping reach.
FREVKY said: Updated the spreadsheet with your corrected rating evaluation for positions you mentioned. So now for striker it stacks up jumping reach, as it is taken into equation twice (most and least important coefficient - for the latter sqrt((jumping reach)^2) equals just jumping reach.
I would like to contribute to the forum with an opinion. I apologize for my English. I'm using AI to translate.
I want to thank the creator of the test and the improvements that were posted here on how to use it in Excel. I would like to confirm that using the table can lead to success in choosing players for our teams.
I'm playing the second season with Chester. Managing to be champion in the first place. To assemble the cast for the second season, decide to use the Excel spreadsheet posted here with the following method. I first analyzed my team, and found the coefficients for each position. I searched in sky league 2, league one, efl, and premier league, for players who had higher coefficients than mine. If a player of mine performed with an average of 7.20, if I find a player with a better coefficient than my player, playing in a better league, everything indicates that he also performs better. Below are photos from the season so far. I am the candidate for relegation
https://gyazo.com/e2bb12f765bbf5775568145f31cac23c
All these players crossed out below were the ones I signed for this new season. and they are performing very well.
https://gyazo.com/39c6590b540c137c977eec2e2edb453c
After 16 rounds, this is the championship
https://gyazo.com/8a192eef70a3115b18495614e297281c
Thanks for making it public, and keep improving. A hug. Then I'll come back to tell you the end of the season.
Orion said: Just to keep you guys up to date I've decided to make a small adjustment for the features that use two attributes - so the last one for MLR and the last one for Striker. I've decided to take square of a product from multiplication those features (SQR(Attribute1 * Attribute2) * Coefficient). This way you can keep it proportional to the coefficient.
I don't know how it works with your files. I assume that for spreadsheet it's not an issue.
What about GenieScout? Does it support other mathematical operations? If not I'd suggest just to get rid of those 2 features (since they are last important for given positions) and adjust others to sum up to 100% just like you did for the other positions.
Genie doesn't allow such mathematical operations unfortunately.
Spreadsheet would be 1:1 replica of the coefficients. If you have time and patience i would suggest using the spreadsheet. You can still use the Genie ratings they are very close to spreadsheet.
FREVKY said: Updated the spreadsheet with your corrected rating evaluation for positions you mentioned. So now for striker it stacks up jumping reach, as it is taken into equation twice (most and least important coefficient - for the latter sqrt((jumping reach)^2) equals just jumping reach.
What do you mean? The effect of jumping reach should be much smaller than it was before.
Although honestly I think for now it would be best to just remove the double attributes for the time being. Too much confusion and desperately trying to make it work for very minimal effect.
Yarema said: What do you mean? The effect of jumping reach should be much smaller than it was before.
Although honestly I think for now it would be best to just remove the double attributes for the time being. Too much confusion and desperately trying to make it work for very minimal effect.
It certailny is smaller than before, what I meant is that ST formula still covers jumping reach twice. I just wanted to point out that square root of "X to the power of 2" is "X", so there's no need for making it more complex than it should be from mathematical point of view.
FREVKY said: It certailny is smaller than before, what I meant is that ST formula still covers jumping reach twice. I just wanted to point out that square root of "X to the power of 2" is "X", so there's no need for making it more complex than it should be from mathematical point of view.
Yes but it's a general solution. MRL also has "Acc Agi" combo, interestingly without Agi itself as a solo attribute.
I rerun the model for those 2 positions to use only linear regressions so no we don't get issues with double attributes or anything like that.
Results:
STC
Attribute Coefficient
Jum 0,020557
Pac 0,020096
Acc 0,014496
Cnt 0,013675
Bal 0,012043
Dri 0,010739
Vis 0,009392
Cmp 0,009161
MLR
Attribute Coefficient
Pac 0,020497
Dri 0,014018
Acc 0,012883
Cmp 0,012072
Vis 0,011542
Jum 0,011150
Cro 0,010598
Sta 0,009658
I'll update first post with the new values.
I think this will save us a lot of confusion.
Orion said: I rerun the model for those 2 positions to use only linear regressions so no we don't get issues with double attributes or anything like that.
Results:
STC
Attribute Coefficient
Jum 0,020557
Pac 0,020096
Acc 0,014496
Cnt 0,013675
Bal 0,012043
Dri 0,010739
Vis 0,009392
Cmp 0,009161
MLR
Attribute Coefficient
Pac 0,020497
Dri 0,014018
Acc 0,012883
Cmp 0,012072
Vis 0,011542
Jum 0,011150
Cro 0,010598
Sta 0,009658
I'll update first post with the new values.
I think this will save us a lot of confusion.
Are those coefficients for the ST a target striker or fast striker?
CBP87 said: Are those coefficients for the ST a target striker or fast striker?
It doesn't differentiate by role
CBP87 said: Are those coefficients for the ST a target striker or fast striker?
Yarema said: It doesn't differentiate by role
This. It's just for position no matter the role. If I could somehow filter out players by their most played position and role I'd do it in a heartbeat.
Orion said: This. It's just for position no matter the role. If I could somehow filter out players by their most played position and role I'd do it in a heartbeat.
So just to clarify, you're saying that jumping is the most important attribute for a striker?
CBP87 said: So just to clarify, you're saying that jumping is the most important attribute for a striker?
According to this model it's almost equally important as Pace but yes. These are 2 most important attributes for a striker.
Guys, how are your results with such attribute prioritization?
I observed that for many positions, very similar type of player is the best.
For example I have 3 pacy CDs with great jumping reach and well who would have guessed that, they would also rank as 1, 2 and 4 in my strikers lineup.
So I think this weights give us great understanding what stats are most valuable, but only on very general scale. I for example don't believe that FB(att) will benefit from those 8 attributes the same way as IFB(Def). Or similar example wide playmaker VS inside forward.
Maybe the think that is missing in all these test is some atribute variability within the team. Like sure jumping reach is OP, but maybe it's enough to have 3-4 really good players in that and the team would benefit more if other players have better passing, or dribbling or whatever.
Mrjoser said: I observed that for many positions, very similar type of player is the best.
For example I have 3 pacy CDs with great jumping reach and well who would have guessed that, they would also rank as 1, 2 and 4 in my strikers lineup.
So I think this weights give us great understanding what stats are most valuable, but only on very general scale. I for example don't believe that FB(att) will benefit from those 8 attributes the same way as IFB(Def). Or similar example wide playmaker VS inside forward.
Maybe the think that is missing in all these test is some atribute variability within the team. Like sure jumping reach is OP, but maybe it's enough to have 3-4 really good players in that and the team would benefit more if other players have better passing, or dribbling or whatever.
As said before, If I was able to filter players by their most played position and role I'd do it in a heartbeat. Sadly I don't know if there is even a possibility to extract such data in any useful way.
Vanarama south team, third season currently on sky bet 2
Kamas1 said: Vanarama south team, third season currently on sky bet 2
I see that, just like in my save, DMs tend to have lower ratings than the rest of the team.
I thought that maybe my own DM is just not that good since his average rating was below 7 but on the other hand the others had worse average on that position.
Damn. 48 goals is really like a cheat code. Can you show this striker attributes?
Generally the biggest problem is finding valuable DMs. No problem, this is striker
I think consistently include momentum, taking into consideration the game. The full-backs are involved in aerial disputes, helping on a corner, or when the opposing goalkeeper covers a goal kick. Sometimes I just look at the value of the attribute found by the coefficients and compare it with the best player in the position in the league, or in the list of the best 11. or with the team that is favored to win the league.
robbr13 said: I think consistently include momentum, taking into consideration the game. The full-backs are involved in aerial disputes, helping on a corner, or when the opposing goalkeeper covers a goal kick. Sometimes I just look at the value of the attribute found by the coefficients and compare it with the best player in the position in the league, or in the list of the best 11. or with the team that is favored to win the league.
That's exactly how the model works. If we wanted to use it 'properly' we should check the average value for given attributes in the league you're in. And then for the attribute value (the one you then multiply by coefficient) you should use the difference between players actual attribute value minus league average.
We use 'straightforwardly' just the attribute value to speed up things. Of course in general it induce the error but since we are comparing players that will play within the same league (all the players that will play for us) it's just much simpler and we're saying that we're ok with the possible error.
At the end of the day it's still better than rely on attributes that the game suggests.
Hi, i am new to this, i was trying to find strikers based on the "meta attributes". so this kid should be a machine.
I wonder if you could share the full lists of coefficients (not only the 8 most prominent ones). This way I can use them to make a squad analysis sheet using the coefficients to score your players for suitability on each of these positions.
DeFlow said: I wonder if you could share the full lists of coefficients (not only the 8 most prominent ones). This way I can use them to make a squad analysis sheet using the coefficients to score your players for suitability on each of these positions.
Of course it's possible. I'd just have to generate model that have all the attributes.
I choose to use only 8 to make it more usable. With 8 attributes you can even make yourself a simple spreadsheet to compare players.
With 30-40-50 features a lot of them will literally have coefficients like 0,0001 so they won't bring much to the table.
The question is what attributes you'll need. Only the visible ones? By visible I mean those from default player profile page. Because I extracted from the game around 50 features including attributes, hidden attributes, height, weight, player's league rank and foot (left/right) proficiency.
It is up to you how many coeffecients you want to extract. Those that are significantly lower than the ones above it could be excluded, but I am sure right now there are more coefficients that are almost as important as the 8 you have shared here?

Right now I am scoring the players on the coefficients for the positions you have extracted, but any coefficient can be incorporated in the spreadsheet.
DeFlow said: It is up to you how many coeffecients you want to extract. Those that are significantly lower than the ones above it could be excluded, but I am sure right now there are more coefficients that are almost as important as the 8 you have shared here?

Right now I am scoring the players on the coefficients for the positions you have extracted, but any coefficient can be incorporated in the spreadsheet.
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?
Orion said: We use vastly different method.
As far as I understand FM Arena attribute test has a testing league where they change certain attribute for every player in one team in that league and check the difference it makes.
It's overall very good method.
I think mine could be called closer to 'real' game environment - since I use data from 'real' leagues. So very simplified explanation is that players in certain position that have high Crossing attribute have high ratings. That's basically it.
So the model pick that attribute as the one that correlates in those positions with player rating - so according to the model the higher the crossing the better player rating.
If we give high crossing to every player it's kind of useless because a lot of position/roles do not utilize this.
FM Arena attribute test looks for attributes that have highest impact on 'universal' level so they look for attributes that will benefit for the whole team, not just single player or single position.
In that case shouldn't the comparisons be between players who play in that position in that league, rather than the average value across the entire league?
joshua said: In that case shouldn't the comparisons be between players who play in that position in that league, rather than the average value across the entire league?
In 'general' model the average was calculated for every league for all the players.
For models for each position average was calculated for every league for the players that play that certain position.
Hi, just wanna show half a season with a broke portuguese club. I´ve been the Excel to ajust my team to the best positions, and so far so good.