Pikabo said: Ok, I had a small mistake in my calculations (mixed the multipliers for teamwork and leadership). Thanks for checking it on your side!
The only question thats left is that if it is accurate to use TRCA for other tactics as well? Does it matter what roles were used? Or as long as the tactic uses GK/CD/FB/DM/W/AM/ST and avoids WB/MC/MRL it is fine to use the TRCA to calculate players usefulness for that Position excluding factors like Players Instructions and Roles. I think that in previous converstations you have already touched the topic but I also think it was based on the General Rating from Genie Scout and not the TRCA. Any idea on that or is it something that was not tested so we can't be sure of the outcome? Expand
We definitely cant be as confident using a formation outside of Blue, however I think you can be reasonably confident. I have used the positions around and related to the ones tested to derive the ratings for other positions. I am confident it works.
I will give you my thoughts on the remaining questions. I have not seen anything that indicates roles are important. I have no faith in the general ratings, I only use positional ratings. Given the experience with the the ykykyky ratings I would be surprised if TRCA doesn't work. They claimed to have tested it and their work all stacks up. I will now have a bit of a look at it.
Pikabo said: Sure, maybe I can show you some examples for better understanding Let's forget about the part with dividing by the perfect player and only stop at the Tactical True Current Ability. Following only your caluclations quoted in my first post I've made an excel file using ykyky balanced table for FST.
I took 2 players and did the calculations:
Then I looked them up in the Genie scout (don't the have the G version so I downloaded FM22 again and used the older version) and their General Rating shows this: ^I know it says TS for Mavididi but its the exact the same as for the FST
And now like I mentioned before, I'm not a fan of Genie Scout cause it shows too much information (CA/PA, other hidden attributes etc.), but the Tactictal True Ability favors Mavididi by a margin and the General Rating in Genie shows that Abel Ruiz would be a better option by 0,5%. I just want to make sure that the excel file can help me pick up the player who probably should perform better based on the weighted attribute table I'm using. Expand
Ok, now I think I understand what you are after. I haven't really played with the Tactical Real CA (TRCA) before. When I do the calculations based on the info you provided I get different scores - 133.17 for Ruiz and 139.55 for Mavididi. When compared with their current CA of 130 (133.17) and 129 (139.55).
The machine learning conclusion would be pick Mavididi, because as they said "if the tactical real CA is higher than the player’s CA, the player is very suitable for ZaZ-Blue DM tactics. The higher the tactical real CA, the more suitable it is for player selection".
Pikabo said: Hello,@Mark I have a question that u have probably already answered in this thread but I'm a little bit lost. Same as others I would like to use excel for the calculations as I don't want to see CA/PA etc.
Based on the calculations provided I did some test to see how it works for Lewandowski as FST. After multiplying each player attribute by the atribute weighting for the FST from the ykyky balanced table, then divided the total of it by the total value of all the weightings (which is 1172 cause im not taking weakfoot) and it gave me a number which then I used to multiply by 20 and subtract 121. This resulted in 204,2 Tactical True Cuerrent Ability for Lewandowski. I also did the calculation for the "perfect player" and the maximum number based on the calculation shows 294,3. I divided Lewandowski's score by the "perfect player" score and % value is 72,7.
And here it is where I'm lost. What other calculation needs to be done to get the General Rating for the position I'm looking for. As I have seen some screenshots posted by You it shows that Lewandowski is somewhere around 92% and its way off from the 72,7 I'm getting.
Sorry in advance if something is unclear, but I hope it all makes sense Expand
I am not sure what some of that means. You dont need to divide by the perfect player score as genie scout will do the calculation.
Please tell me what you are trying to achieve and I will see if I can help you
My current testing table. As a recap I am using the bottom ranked side from all the 7 English competitions. So that is 314 matches for each run. The results differ slightly from the FM Arena tests where all teams are equal. So my tests favour how the tactics are for underdogs.
Any tactic in the top 7 will get you great results in my opinion, as there really isn't much between them. For instance, the earlier version of Wind 2.0 (Wind 1.1) had one bad division result that was quite costly. An average result there would have seen it 3rd or 4th, and I believe there were improvements to the tactic since I tested.
My recommendation would be to use 3 of the top 7 tactics in your 3 training slots making sure they are easy to move between. For example: 1 World of the Best Tactic (4-2-0-2W-2), 2 The Amazing Hindle (4-1-2-2W-1), 3 Amater v2 (4-2-1-2W-1). Ensuring you have (or train) a DM/MC to move from 1 to 2, and a ST/MC to move from 1 to 3 would make it easy to move around the 3 tactics during the game. You could do just as well using 1 Corsair, 2 ZaZ Wind and 3 WestHammer.
@ZaZ I would be happy to test. I have set up an environment that appears to be relative stable. All injury proneness and dirtiness set to 1 for all players, consistency and important matches set to 20, all injuries and bans removed prior to starting, morale and fitness etc set to 20, most contracts frozen as all players have 10 year contracts. I have only loaded the game for English Leagues but a couple of European players have come in for major teams. I have run most of the top tactics and the results are very close to proportionally to here. I have the lowest ranked side in each of the 7 English comps ie 314 matches per run, EPL thru VNS. The Training Rest has been set as per your advice, and the tests run on Holiday with must use tactics. The tests run for about 90 minutes but you get 314 match results for the run. I am just running the second Corsair one at the moment.
Please let me know if you want to be my next guinea pig.
I have taken my old Sponge Bob tactic from FM21 as the base. It was my most successful tactic in FM21. Following NTR proving that AM wingers are better than MRLs and that strikerless dont work I moved the wingers and strikers up a slot each. this moves the formation from 4-2-2W-2-0 to 4-2-0-2W-2. I didn't alter the default positional instructions for the wingers and strikers when moved forward and left what was already there for the other roles. I have not adjusted the set pieces.
I have tested the tactics 6 times with St Albans and 3 times with Everton. The results are impressive. With St Albans the worst position was 3rd with 87 points and still being promoted. The best is the picture below. These are comparable with the best tactics on the FM Arena Table that I tested in this league during Beta.
I also tested with Everton and got a 3rd place finish. I think there is room for improvement but I do think this tactic is competitive.
It still seems to be a good base rating file. I used ykykyky balanced last year for a fair while but I didn't find a big overall difference in team selections or scouting so I am back to using this for FM23.
I should have done more testing. The next test I got sacked. Oh well, worth trying a new formation. I might dust off my FM21 Sponge Bob tactic which is the same shape as the current leader and give it a try
I decided to update my Winx tactic looking at the structures that have been successful so far. Moved a defender up to DM and went with 3 DCs at the back so pretty much still 5 defend and 5 attack. I looked at the defensive player settings for DMs and DCs in tactics SIU, Roll the Dice and Westhammer, and then changed my position settings. I changed the F9 to an AF.
I used St Albans, predicted 16th. The first run scored 110 goals and conceded 85. Tweaked some of the settings and ended with version 2 doing sensationally well. The defence is still leaky but the scoring potential is sensational, 131 goals in the test season. I cant recall seeing a formation like this in FM23. There were 8 red cards and 123 yellows, so could probably have done a fair bit better if I was playing instead of holiday mode.
I still need to play with the set pieces but think it is on the right track.
Anyhow, give it a try and let me know what you think.
Guys, lighten up. We are really here to collaborate not to fight. Fighting will stop our progress. Collaboration will lead to advancement. I appreciate all progress and accept that we all need to see failures to get there. Please appreciate everyone is trying as the starting point.
Take accusations off line through DMs the way described earlier. Please don't bring them all to the thread for open discussion.
The screenshot of the tests results look like this:
The average score after 10,944 matches will be referred to the "true score" and "Random Number Generation" will be referred to "RNG"
Please notice, for everyone's convenience the data is translated into "a typical 38 matches season".
After analyzing the data here's what we found:
If you test a tactic for 1 season ( 1 season x 38 matches = total 38 matches ): - The highest RNG = 25.8 points away from the "true score"
If you test a tactic for 2 seasons ( 2 seasons x 38 matches = total 76 matches ): - The highest RNG = 23.8 points away from the "true score"
If you test a tactic for 4 seasons ( 4 seasons x 38 matches = total 152 matches ): - The highest RNG = 11.6 points away from the "true score"
If you test a tactic for 8 seasons ( 8 seasons x 38 matches = total 304 matches ): - The highest RNG = 8.4 points away from the "true score"
If you test a tactic for 12 seasons ( 12 seasons x 38 matches = total 456 matches ): - The highest RNG = 6.6 points away from the "true score"
If you test a tactic for 24 seasons ( 24 seasons x 38 matches = total 912 matches ): - The highest RNG = 4.3 points away from the "true score"
If you test a tactic for 48 seasons ( 48 seasons x 38 matches = 1,824 matches ): - The highest RNG = 3.2 points away from the "true score"
If you test a tactic for 84 seasons ( 84 seasons x 38 matches = 3,192 matches ): - The highest RNG = 1.9 point away from the "true score"
If you test a tactic for 168 seasons ( 168 seasons x 38 matches = 6,384 matches ): - The highest RNG = 1 point away from the "true score"
As you can see if you test a tactic only for 1 season ( 38 matches ) and you hit the highest RNG then your result will be about 25.8 points away from the "true score" and if you want to test with an accuracy of 1 point then you need to test a tactic 168 seasons or 6,384 matches.
Of course, you can test a tactic only 1 season ( 38 matches ) and hit "the true result" but the probability of doing so is about 3% or so.
Please not, that our testing methodology eliminates many factors that increase the RNG:
- We set the morale and conditions of all players in the testing league to 100% and "freeze" it so the morale and conditions don't change at all. That's true for the AI controlled teams and human controlled teams.
- All the teams in the league use the same starting eleven all the time. No rotation happens.
- No transfers happen. AI managers can't be sacked.
If your testing methodology doesn't eliminate the factors above then the RNG in your tests will much higher than what we got in our 10,944 matches test.
Also, if you test tactics with a very strong team such as PSG in French league where your opponents are much weaker than your team then you won't have any RNG at all because almost with any tactic you'll be getting about 110-114 pts at the end of the season but don't be fooled by such low RNG, you need to understand when your team is much stronger than your opponents then you winning matches mostly due to having much better players than your opponents and the quality of your tactic doesn't make any significant difference and when your team is much weaker than your opponents then you loosing matches mostly due to having much weaker players than your opponents and in this case your tactic also doesn't make any significant difference. So testing tactics with the strongest or the weakest team in the league isn't a good idea at all and you need something in the middle. Expand
Wow, you guys are really getting through the testing now you are up and running. Thanks a lot. The community really appreciates all you guys do.
Tejash said: @Mark Do you think this tactic will top your tests that you did recently? Expand
I did 3 tests, not the full 12 and it was sitting above 91 points, so in the mix. I wont keep testing because I want to start a real game now. DBL and TheAmazingHIndle were both testing well in that test too.
I mean 100 for Pace, 70 for acceleration, and decrease with all the other ratings ?
The problem would be that you end up with the same rating for each position. Feel free to give it a try
Thanks
That's what the machine learning guys said. I have no knowledge of how they came to that conclusion
The only question thats left is that if it is accurate to use TRCA for other tactics as well? Does it matter what roles were used? Or as long as the tactic uses GK/CD/FB/DM/W/AM/ST and avoids WB/MC/MRL it is fine to use the TRCA to calculate players usefulness for that Position excluding factors like Players Instructions and Roles. I think that in previous converstations you have already touched the topic but I also think it was based on the General Rating from Genie Scout and not the TRCA. Any idea on that or is it something that was not tested so we can't be sure of the outcome?
We definitely cant be as confident using a formation outside of Blue, however I think you can be reasonably confident. I have used the positions around and related to the ones tested to derive the ratings for other positions. I am confident it works.
I will give you my thoughts on the remaining questions. I have not seen anything that indicates roles are important. I have no faith in the general ratings, I only use positional ratings. Given the experience with the the ykykyky ratings I would be surprised if TRCA doesn't work. They claimed to have tested it and their work all stacks up. I will now have a bit of a look at it.
Let's forget about the part with dividing by the perfect player and only stop at the Tactical True Current Ability. Following only your caluclations quoted in my first post I've made an excel file using ykyky balanced table for FST.
I took 2 players and did the calculations:
Then I looked them up in the Genie scout (don't the have the G version so I downloaded FM22 again and used the older version) and their General Rating shows this:
^I know it says TS for Mavididi but its the exact the same as for the FST
And now like I mentioned before, I'm not a fan of Genie Scout cause it shows too much information (CA/PA, other hidden attributes etc.), but the Tactictal True Ability favors Mavididi by a margin and the General Rating in Genie shows that Abel Ruiz would be a better option by 0,5%. I just want to make sure that the excel file can help me pick up the player who probably should perform better based on the weighted attribute table I'm using.
Ok, now I think I understand what you are after. I haven't really played with the Tactical Real CA (TRCA) before. When I do the calculations based on the info you provided I get different scores - 133.17 for Ruiz and 139.55 for Mavididi. When compared with their current CA of 130 (133.17) and 129 (139.55).
The machine learning conclusion would be pick Mavididi, because as they said "if the tactical real CA is higher than the player’s CA, the player is very suitable for ZaZ-Blue DM tactics. The higher the tactical real CA, the more suitable it is for player selection".
I hope this helps
Based on the calculations provided I did some test to see how it works for Lewandowski as FST. After multiplying each player attribute by the atribute weighting for the FST from the ykyky balanced table, then divided the total of it by the total value of all the weightings (which is 1172 cause im not taking weakfoot) and it gave me a number which then I used to multiply by 20 and subtract 121. This resulted in 204,2 Tactical True Cuerrent Ability for Lewandowski. I also did the calculation for the "perfect player" and the maximum number based on the calculation shows 294,3. I divided Lewandowski's score by the "perfect player" score and % value is 72,7.
And here it is where I'm lost. What other calculation needs to be done to get the General Rating for the position I'm looking for. As I have seen some screenshots posted by You it shows that Lewandowski is somewhere around 92% and its way off from the 72,7 I'm getting.
Sorry in advance if something is unclear, but I hope it all makes sense
I am not sure what some of that means. You dont need to divide by the perfect player score as genie scout will do the calculation.
Please tell me what you are trying to achieve and I will see if I can help you
Don't let them get to you mate, you are doing a great job
Any tactic in the top 7 will get you great results in my opinion, as there really isn't much between them. For instance, the earlier version of Wind 2.0 (Wind 1.1) had one bad division result that was quite costly. An average result there would have seen it 3rd or 4th, and I believe there were improvements to the tactic since I tested.
My recommendation would be to use 3 of the top 7 tactics in your 3 training slots making sure they are easy to move between. For example: 1 World of the Best Tactic (4-2-0-2W-2), 2 The Amazing Hindle (4-1-2-2W-1), 3 Amater v2 (4-2-1-2W-1). Ensuring you have (or train) a DM/MC to move from 1 to 2, and a ST/MC to move from 1 to 3 would make it easy to move around the 3 tactics during the game. You could do just as well using 1 Corsair, 2 ZaZ Wind and 3 WestHammer.
They appear to - yes
Please let me know if you want to be my next guinea pig.
I have tested the tactics 6 times with St Albans and 3 times with Everton. The results are impressive. With St Albans the worst position was 3rd with 87 points and still being promoted. The best is the picture below. These are comparable with the best tactics on the FM Arena Table that I tested in this league during Beta.
I also tested with Everton and got a 3rd place finish. I think there is room for improvement but I do think this tactic is competitive.
It still seems to be a good base rating file. I used ykykyky balanced last year for a fair while but I didn't find a big overall difference in team selections or scouting so I am back to using this for FM23.
Have you tried it any "professional" league?
I'm afraid you'll get butchered with it in any top league... you should give it a go in EPL
I might give it a run with Everton later tonight. My save is up to the first competitive game so I want to get a few games in there first.
I used St Albans, predicted 16th. The first run scored 110 goals and conceded 85. Tweaked some of the settings and ended with version 2 doing sensationally well. The defence is still leaky but the scoring potential is sensational, 131 goals in the test season. I cant recall seeing a formation like this in FM23. There were 8 red cards and 123 yellows, so could probably have done a fair bit better if I was playing instead of holiday mode.
I still need to play with the set pieces but think it is on the right track.
Anyhow, give it a try and let me know what you think.
Take accusations off line through DMs the way described earlier. Please don't bring them all to the thread for open discussion.
I'm sure that many of you wonder how many matches it requites testing a tactic to get an accurate result? So we've tested that and found an answer.
We tested one tactic 10,944 matches to measure the RNG on different distances.
Here's you can download the screenshots of the tests results - 10,944 Matches Test - Download
The screenshot of the tests results look like this:
The average score after 10,944 matches will be referred to the "true score" and "Random Number Generation" will be referred to "RNG"
Please notice, for everyone's convenience the data is translated into "a typical 38 matches season".
After analyzing the data here's what we found:
If you test a tactic for 1 season ( 1 season x 38 matches = total 38 matches ):
- The highest RNG = 25.8 points away from the "true score"
If you test a tactic for 2 seasons ( 2 seasons x 38 matches = total 76 matches ):
- The highest RNG = 23.8 points away from the "true score"
If you test a tactic for 4 seasons ( 4 seasons x 38 matches = total 152 matches ):
- The highest RNG = 11.6 points away from the "true score"
If you test a tactic for 8 seasons ( 8 seasons x 38 matches = total 304 matches ):
- The highest RNG = 8.4 points away from the "true score"
If you test a tactic for 12 seasons ( 12 seasons x 38 matches = total 456 matches ):
- The highest RNG = 6.6 points away from the "true score"
If you test a tactic for 24 seasons ( 24 seasons x 38 matches = total 912 matches ):
- The highest RNG = 4.3 points away from the "true score"
If you test a tactic for 48 seasons ( 48 seasons x 38 matches = 1,824 matches ):
- The highest RNG = 3.2 points away from the "true score"
If you test a tactic for 84 seasons ( 84 seasons x 38 matches = 3,192 matches ):
- The highest RNG = 1.9 point away from the "true score"
If you test a tactic for 168 seasons ( 168 seasons x 38 matches = 6,384 matches ):
- The highest RNG = 1 point away from the "true score"
As you can see if you test a tactic only for 1 season ( 38 matches ) and you hit the highest RNG then your result will be about 25.8 points away from the "true score" and if you want to test with an accuracy of 1 point then you need to test a tactic 168 seasons or 6,384 matches.
Of course, you can test a tactic only 1 season ( 38 matches ) and hit "the true result" but the probability of doing so is about 3% or so.
Please not, that our testing methodology eliminates many factors that increase the RNG:
- We set the morale and conditions of all players in the testing league to 100% and "freeze" it so the morale and conditions don't change at all. That's true for the AI controlled teams and human controlled teams.
- All the teams in the league use the same starting eleven all the time. No rotation happens.
- No transfers happen. AI managers can't be sacked.
If your testing methodology doesn't eliminate the factors above then the RNG in your tests will much higher than what we got in our 10,944 matches test.
Also, if you test tactics with a very strong team such as PSG in French league where your opponents are much weaker than your team then you won't have any RNG at all because almost with any tactic you'll be getting about 110-114 pts at the end of the season but don't be fooled by such low RNG, you need to understand when your team is much stronger than your opponents then you winning matches mostly due to having much better players than your opponents and the quality of your tactic doesn't make any significant difference and when your team is much weaker than your opponents then you loosing matches mostly due to having much weaker players than your opponents and in this case your tactic also doesn't make any significant difference. So testing tactics with the strongest or the weakest team in the league isn't a good idea at all and you need something in the middle.
Wow, you guys are really getting through the testing now you are up and running. Thanks a lot. The community really appreciates all you guys do.
I did 3 tests, not the full 12 and it was sitting above 91 points, so in the mix. I wont keep testing because I want to start a real game now. DBL and TheAmazingHIndle were both testing well in that test too.
Thanks