I want to let you know that soon, there'll be some improvements to FM-Arena Tactic Testing League.
We've managed to adjust the league setting in such way that we forced AI Controlled Teams to play only vs Human Controlled Teams and Human Controlled Teams to play only vs AI controlled Teams, that adjustments prevent wasting computing power on calculation matches where AI Controlled Teams play vs AI Controlled Teams matches and that speeds up the testing process about 2-3 times , so we probably will be able to test the tactics in the Hall of Fame for 7,000 - 8,000 matches instead 2,000 - 4,000 marches as it happens now.
So the updated tactic testing league would look something like this:
AI Controlled Teams play only vs Human Controlled Teams:
Human Controlled Teams play only vs AI Controlled Teams:
So no more spending computing power on calculating matches where AI Controlled Teams plays vs AI Controlled Teams matches, which speeds up the testing process 2-3 times.
Also, there'll be other improvements, for example, at the moment we've gathered huge amount of data about which AI tactical settings work the best and we're gonna implement them so no more spending computing power on testing vs weak AI managers.
We've been doing internal tests on the updated testing league and the current top tactic "4123 Pressure Cooker" dropped about 9 points vs AI "Elite" Managers, the score decreased from 66 to 57.
When I say AI "Elite" Managers I mean the most successful AI tactical settings that we've tested so far.
Probably, the first wave of test results on the new tactic testing league should pop up in the next couple of hours.
The delay occurred because we couldn't apply a fix that protects the AI managers from being sacked, it didn't happen often but still happened and added a negative impact on the accuracy of the testing but now it's been sorted out.
Zippo said: Probably, the first wave of test results on the new tactic testing league should pop up in the next couple of hours.
The delay occurred because we couldn't apply a fix that protects the AI managers from being sacked, it didn't happen often but still happened and added a negative impact on the accuracy of the testing but now it's been sorted out. Expand
I've looked at the tests with Patch 22.1.1 (N) and Patch 22.1.1. The tactics change very radically in terms of success.It's hard to say anything for 456 matches, but it's still very different from 304 matches.(sorry for translate)
Zippo said: We've been doing internal tests on the updated testing league and the current top tactic "4123 Pressure Cooker" dropped about 9 points vs AI "Elite" Managers, the score decreased from 66 to 57.
When I say AI "Elite" Managers I mean the most successful AI tactical settings that we've tested so far. Expand
A suggestion :
Can we add to the queque like the top 2-3 for each Hall of Fame tactic rank (just giving a suggestion), so we can understand more in depth how the things are working, imho is useless to just to spam again the "try-outs". Ofc it can happen once all of the Hall of fame are being tested.
Indeed it will take "time" but imho is the best way to avoid more "brute" force testing. And i would personally even give a different tune to the threshold,i mean before was like a coin-flip (960 games),if you were lucky you could get more test otherwise nop, and imho like that maybe we missed some good setup as well.
There are 10 teams in the league, but 5 teams use the same tactic. It just didn't seem quite right to me. I also think it would make more sense to have an arrangement suitable for real-life popular leagues. A calendar of 76 matches with 20 teams would have been better.
In addition, how does a team in a 10-team league complete the season with 80 matches? Shouldn't it be 81 if they met 9 times?
Lapidus said: It isn't irrelevant that 5 teams use the same tactic because they don't play against each other. Expand
Oh right. I had to read the descriptions. I was focused on the screenshot. The explanations also answered my second question. Teams appear to be completing 80 matches, meeting 16 times against AI teams.
pixar said: There are 10 teams in the league, but 5 teams use the same tactic. It just didn't seem quite right to me. I also think it would make more sense to have an arrangement suitable for real-life popular leagues. A calendar of 76 matches with 20 teams would have been better.
In addition, how does a team in a 10-team league complete the season with 80 matches? Shouldn't it be 81 if they met 9 times? Expand
The human teams (Team A, Team B and so on...) plays only against AI teams.
When I checked the save file from last thread, it seems like all Human teams have the same CA, isnt it better to have each team being a 10CA less than the top one so it goes from like 120CA to 160CA, so we can actually tell which tactics are better for underdogs and which are suited for top dogs, as for now, most of the top tactics here are made mostly for top teams.
Wail said: When I checked the save file from last thread, it seems like all Human teams have the same CA, isnt it better to have each team being a 10CA less than the top one so it goes from like 120CA to 160CA, so we can actually tell which tactics are better for underdogs and which are suited for top dogs, as for now, most of the top tactics here are made mostly for top teams. Expand
Wail said: When I checked the save file from last thread, it seems like all Human teams have the same CA, isnt it better to have each team being a 10CA less than the top one so it goes from like 120CA to 160CA, so we can actually tell which tactics are better for underdogs and which are suited for top dogs, as for now, most of the top tactics here are made mostly for top teams. Expand
they only played with big team bro (Man city, Liverpool)
Hey guys i love your job, all of you. I just have some questions. In this new table what would be consider a good pts average and and a good GD? Or is too early to know? The table delivers more accurate evaluation of the tactics?
Zippo said: We've been doing internal tests on the updated testing league and the current top tactic "4123 Pressure Cooker" dropped about 9 points vs AI "Elite" Managers, the score decreased from 66 to 57.
When I say AI "Elite" Managers I mean the most successful AI tactical settings that we've tested so far. Expand
will you share the attributes of this "elite" AI manager with us as well? I like creating AI managers in m save and just observe their career. Would be really fun to see how well they perform with the elte attributes.
dzek said: @Zippo Can we have a saved game of the new test league? Expand I can only provide you with a game save at the end of a test - Game Save
Floppyaams said: will you share the attributes of this "elite" AI manager with us as well? I like creating AI managers in m save and just observe their career. Would be really fun to see how well they perform with the elte attributes. Expand
the bellow 2 the most OP formation for AI
But bear in mind, to be successful in real game the "Non Tactical" attributes of an AI manager should also be high but for a tactic testing league it isn't that important some we ignored many of them like, I'm talking about attributes like "Buying Players", "Negotiating", "Business" and some other.
Excuse me , I know this isnt the correct forum , but I want to ask what is the difference of the 23.4.0 (N) patch compared to the 23.4.0 . I have searched on the web but I cant seem to find anything regarding this new update. Furthermore , the game itself hasnt downloaded any new patch . Once again, knowing that this is the wrong forum , but it seems to be the most active. Apologies for any incoveniences.
Zippo said: I can only provide you with a game save at the end of a test - Game Save
the bellow 2 the most OP formation for AI
But bear in mind, to be successful in real game the "Non Tactical" attributes of an AI manager should also be high but for a tactic testing league it isn't that important some we ignored many of them like, I'm talking about attributes like "Buying Players", "Negotiating", "Business" and some other.
Hey everybody,
, so we probably will be able to test the tactics in the Hall of Fame for 7,000 - 8,000 matches instead 2,000 - 4,000 marches as it happens now.
I want to let you know that soon, there'll be some improvements to FM-Arena Tactic Testing League.
We've managed to adjust the league setting in such way that we forced AI Controlled Teams to play only vs Human Controlled Teams and Human Controlled Teams to play only vs AI controlled Teams, that adjustments prevent wasting computing power on calculation matches where AI Controlled Teams play vs AI Controlled Teams matches and that speeds up the testing process about 2-3 times
So the updated tactic testing league would look something like this:
AI Controlled Teams play only vs Human Controlled Teams:
Human Controlled Teams play only vs AI Controlled Teams:
So no more spending computing power on calculating matches where AI Controlled Teams plays vs AI Controlled Teams matches, which speeds up the testing process 2-3 times.
Also, there'll be other improvements, for example, at the moment we've gathered huge amount of data about which AI tactical settings work the best and we're gonna implement them so no more spending computing power on testing vs weak AI managers.
I hope you like it!
Cheers.
Incredible work! 🔥🔥
That is awesome.I want it to start as soon as possible.
Are you sure you people are not from another planet?
Totaly crazy. GOOD JOB!!
We've been doing internal tests on the updated testing league and the current top tactic "4123 Pressure Cooker" dropped about 9 points vs AI "Elite" Managers, the score decreased from 66 to 57.
When I say AI "Elite" Managers I mean the most successful AI tactical settings that we've tested so far.
Wow.. this is a total game changer, congrats man
@Zippo do you have any estimation when the new testing league will take place on new test runs?
dzek said: @Zippo do you have any estimation when the new testing league will take place on new test runs?
probably, it'll occur in 1-2 days
sweet
Probably, the first wave of test results on the new tactic testing league should pop up in the next couple of hours.
The delay occurred because we couldn't apply a fix that protects the AI managers from being sacked, it didn't happen often but still happened and added a negative impact on the accuracy of the testing but now it's been sorted out.
it would look like this
thanks for doing this
Zippo said: Probably, the first wave of test results on the new tactic testing league should pop up in the next couple of hours.
The delay occurred because we couldn't apply a fix that protects the AI managers from being sacked, it didn't happen often but still happened and added a negative impact on the accuracy of the testing but now it's been sorted out.
Awesome
@Zippo Can we have a saved game of the new test league?
I've looked at the tests with Patch 22.1.1 (N) and Patch 22.1.1. The tactics change very radically in terms of success.It's hard to say anything for 456 matches, but it's still very different from 304 matches.(sorry for translate)
Zippo said: We've been doing internal tests on the updated testing league and the current top tactic "4123 Pressure Cooker" dropped about 9 points vs AI "Elite" Managers, the score decreased from 66 to 57.
When I say AI "Elite" Managers I mean the most successful AI tactical settings that we've tested so far.
A suggestion :
Can we add to the queque like the top 2-3 for each Hall of Fame tactic rank (just giving a suggestion), so we can understand more in depth how the things are working, imho is useless to just to spam again the "try-outs". Ofc it can happen once all of the Hall of fame are being tested.
Indeed it will take "time" but imho is the best way to avoid more "brute" force testing. And i would personally even give a different tune to the threshold,i mean before was like a coin-flip (960 games),if you were lucky you could get more test otherwise nop, and imho like that maybe we missed some good setup as well.
There are 10 teams in the league, but 5 teams use the same tactic. It just didn't seem quite right to me.
I also think it would make more sense to have an arrangement suitable for real-life popular leagues. A calendar of 76 matches with 20 teams would have been better.
In addition, how does a team in a 10-team league complete the season with 80 matches? Shouldn't it be 81 if they met 9 times?
pixar said: There are 10 teams in the league, but 5 teams use the same tactic. It just didn't seem quite right to me.
It isn't irrelevant that 5 teams use the same tactic because they don't play against each other.
Lapidus said: It isn't irrelevant that 5 teams use the same tactic because they don't play against each other.
Oh right. I had to read the descriptions. I was focused on the screenshot. The explanations also answered my second question. Teams appear to be completing 80 matches, meeting 16 times against AI teams.
pixar said: There are 10 teams in the league, but 5 teams use the same tactic. It just didn't seem quite right to me.
I also think it would make more sense to have an arrangement suitable for real-life popular leagues. A calendar of 76 matches with 20 teams would have been better.
In addition, how does a team in a 10-team league complete the season with 80 matches? Shouldn't it be 81 if they met 9 times?
The human teams (Team A, Team B and so on...) plays only against AI teams.
5 Human teams
5 AI teams
They play 16 times against each other
5 x 16 = 80
When I checked the save file from last thread, it seems like all Human teams have the same CA, isnt it better to have each team being a 10CA less than the top one so it goes from like 120CA to 160CA, so we can actually tell which tactics are better for underdogs and which are suited for top dogs, as for now, most of the top tactics here are made mostly for top teams.
Wail said: When I checked the save file from last thread, it seems like all Human teams have the same CA, isnt it better to have each team being a 10CA less than the top one so it goes from like 120CA to 160CA, so we can actually tell which tactics are better for underdogs and which are suited for top dogs, as for now, most of the top tactics here are made mostly for top teams.
CA doesn't tell anything about the quality of players - https://fm-arena.com/find-comment/21601/
Chewbacca said: CA doesn't tell anything about the quality of players - https://fm-arena.com/find-comment/21601/
its the closest thing to value the 5 human teams, the 5 teams are pretty close in terms of level even if you ignore CA.
Wail said: When I checked the save file from last thread, it seems like all Human teams have the same CA, isnt it better to have each team being a 10CA less than the top one so it goes from like 120CA to 160CA, so we can actually tell which tactics are better for underdogs and which are suited for top dogs, as for now, most of the top tactics here are made mostly for top teams.
they only played with big team bro (Man city, Liverpool)
irapbot said: they only played with big team bro (Man city, Liverpool)
Nah the Test league is with teams of around 150CA i think, all 5 teams are on the same level.
Hey guys i love your job, all of you. I just have some questions. In this new table what would be consider a good pts average and and a good GD? Or is too early to know? The table delivers more accurate evaluation of the tactics?
Zippo said: We've been doing internal tests on the updated testing league and the current top tactic "4123 Pressure Cooker" dropped about 9 points vs AI "Elite" Managers, the score decreased from 66 to 57.
When I say AI "Elite" Managers I mean the most successful AI tactical settings that we've tested so far.
will you share the attributes of this "elite" AI manager with us as well? I like creating AI managers in m save and just observe their career. Would be really fun to see how well they perform with the elte attributes.
dzek said: @Zippo Can we have a saved game of the new test league?
I can only provide you with a game save at the end of a test - Game Save
Floppyaams said: will you share the attributes of this "elite" AI manager with us as well? I like creating AI managers in m save and just observe their career. Would be really fun to see how well they perform with the elte attributes.
the bellow 2 the most OP formation for AI
But bear in mind, to be successful in real game the "Non Tactical" attributes of an AI manager should also be high but for a tactic testing league it isn't that important some we ignored many of them like, I'm talking about attributes like "Buying Players", "Negotiating", "Business" and some other.
Excuse me , I know this isnt the correct forum , but I want to ask what is the difference of the 23.4.0 (N) patch compared to the 23.4.0 . I have searched on the web but I cant seem to find anything regarding this new update. Furthermore , the game itself hasnt downloaded any new patch . Once again, knowing that this is the wrong forum , but it seems to be the most active. Apologies for any incoveniences.
Zippo said: I can only provide you with a game save at the end of a test - Game Save
the bellow 2 the most OP formation for AI
But bear in mind, to be successful in real game the "Non Tactical" attributes of an AI manager should also be high but for a tactic testing league it isn't that important some we ignored many of them like, I'm talking about attributes like "Buying Players", "Negotiating", "Business" and some other.
How did you determine the best formations for AI?