Some mechanism detailed tests (it took me 1-2 months)

1.Position Proficiency位置熟练度





This has little to do with the main theme, so I present it independently
这个和主题关系不大，所以我独立的将他摆出来


(Test Condition setting and Preparation Section)
（测试条件设置和准备部分）
↓

2.The primary purpose of the test is to solve a previously hard-to-explain problem:
Why does an increase in attributes (under some given conditions) lead to a decrease in goal difference/winning rate instead?
Why does the reduction of attributes (under some given conditions) lead to an increase in goal difference/winning rate instead?

测试的最主要目的是解决一个之前难以解释的问题，为什么（特定情况下）属性的增加，反而导致净胜球/胜率的下降？
为什么（给定条件下）属性的减少，反而导致净胜球/胜率的增加？


3.According to the principles of science. I hope to be able to control variables, facilitate observation measurement and statistics, be reproducible in the same environment, and have a sufficient sample size to eliminate random errors.
However, I have to sacrifice a certain degree of universality (that is, testing under more different preset conditions), and because I lack software tools, knowledge related to automation testing and high-performance computers to do so.

But, even if not entirely rigorous, this is still more reliable than the vast majority of "empiricism".
And, due to time constraints, I am also unable to produce a very large number of samples.
The lack of sufficient samples will weaken the credibility to a certain extent. However, if the same situation occurs multiple times and the distinction is obvious, it can prove that this is credible.

根据科学的原则。我希望能够控制变量、方便观测测量统计、同环境可重复再现、有足够的样本量来排除随机误差。
而我必须牺牲一定程度的普适性（即在更多不同的预设条件下测试），因为我缺乏软件工具、自动化相关的知识和高性能的电脑来这样做。

但是，即使不完全严谨，这也比绝大部分“经验主义”可靠。
而由于时间关系，我也没法做到非常大量的样本。
缺少足够样本会一定程度削弱可信程度，但是如果同样的情况多次出现并且区分很明显，可以证明这个是可信的。

4.One of the leagues being tested comes from "DZEK" on the forum. One of the league archives I modified, the download link is:
https://pixeldrain.com/u/7vaWxyAC

测试的其中一个联赛来自论坛的“DZEK”。我更改后的联赛的存档中的其中一个，下载链接是：https://pixeldrain.com/u/7vaWxyAC

Part of the test tactics used is:https://pixeldrain.com/u/Smeh7B9e
Part of the test data is:https://pixeldrain.com/u/KuemeJ9R

使用到的测试战术的一部分是：https://pixeldrain.com/u/Smeh7B9e
测试数据的一部分是：https://pixeldrain.com/u/KuemeJ9R

5.Thank you to the players of the "Bao Peng" forum for reverse engineering the game engine (have pictures in previous post).

According to his achievements,he crack the attribute is "propensity (number of attempts)" + "success rate".———— And if this is true, because "trying" or increasing "attempts" is not always a good thing, in some cases it is bound to have much more negative effects.
And if this logic holds true, this phenomenon will be clearly visible in many situations. In different circumstances, different attributes will be increased, which will have a negative effect on the team's performance.

感谢“爆棚”论坛的玩家对游戏引擎的逆向工程（我在之前的帖子里有图片）。
根据他的成果，属性是“倾向性（尝试次数）”+“成功率”。————而如果这个成果是真的，因为“尝试”或者增加“次数”并不总是好事，在一些情况下它必然会导致负面效果。而如果这个逻辑成立，这个现象会在很多情况中清晰可见。在不同的情况下，将有不同的属性被增加后，对球队的表现会是负面效果。


For example, a high "passing" attribute = "prefer passing more attempts" (and to some extent ignore your tactical instructions) + "Higher passing success rate"

比如“传球”属性高=”更喜欢传球“（并且一定程度无视你的战术指令）+”传球成功率更高“

It is possible that high attributes led the player to go against the correct tactical choices and do what he likes. Although his success rate in doing this was higher and his post-game rating might have been higher as well, it hurt the team. (And vice versa)

可能：属性高导致了球员违背了正确的战术选择而去做他喜欢做的事情。尽管他做这个事情的成功率更高了，他的赛后评分可能也更高，但是这伤害了球队。（反之亦然）


(Test Condition setting and Preparation Section)
（测试条件设置和准备部分）
↓

6.Early test
Using FMRTE to fix various variables and create a test condition, I first try to see if I can directly observe some results that conform to the theory
使用FMRTE固定各种变量，并创造一个测试条件，我首先尝试能不能直接观察到一些符合理论的结果


Playing the same opponent 120 times under the same conditions, the player's team is much stronger than AI under the set conditions
与同一个对手反复比赛60场在相同的条件下，在设置的条件下玩家更强

Test result
↓

Then, all the the player on player's team simultaneously added 10 attributes such as passing, Finishing and tackling, and played another 60 games.
It was found that in the group with higher shots, the ratio of "actual goals" to "expected goals" increased by nearly 25%, making "expected goals" almost equal to "actual goals".
However, the number of shots and the number of shots on target did not increase significantly, nor did the number and success rate of passes and tackles increase significantly
I guess this set of data might be due to the fact that the player's team is too strong while the AI's team is too weak, so that the player's data has reached the upper limit of some kind of "soft limit", making it impossible to distinguish.
然后，全队球员同时增加了10的传球、射门、抢断等属性，再比赛60场。
发现，射门高的一组，“实际进球”/“预计进球”增加了接近25%，让“预计进球”几乎等于“实际进球”
但是，射门次数和射正次数并没有明显增加、传球和抢断的次数和成功率也没有明显增加
我猜想是这组数据可能是玩家球队太强而AI的球队太弱，以至于玩家的数据已经到达了某种“软限制”的上限，所以无法区分。


7.I changed the test conditions to make the attributes of both sides basically the same.
我更改了测试条件使双方属性基本相同，




This time, taking the addition of "Shots Finishing" as an example, the added player had more "Shots Attempted" and "Shots On Target".
But he will "snatch" some shooting opportunities from other players

这次结果，以增加“射门Finishing”为例，被增加的球员有更多的“Shots Attempted”和“Shots On Target”。
但是他会“抢走”一部分其他球员的射门机会

This result fits with the above theory, attribute is "propensity (number of attempts)" + "success rate".
这个结果符合上面的理论，属性是“倾向性（尝试次数）”+“成功率”


(Test Condition setting and Preparation Section)
（测试条件设置和准备部分）
↓

8.I began to modify the test league to make it closer to the actual tactics of the AI head coach in the actual game.
The reason is that I found that all the  AI head coach 's "tendencies" in the original save have been completely removed,
and many of the "Formations" selected are tactics that AI coaches would never choose in actual games.

我开始更改测试联赛，让它更接近实际游戏里AI主教练的实际战术。
原因是我发现原来的测试联赛中的设置战术中的““tendencies””被全部清除了，而选用的“Formations”很多都是实际游戏里AI教练完全不会选择的战术。



Open FMRTE and you can read the panel of each AI head coach, which is divided into two parts. Among them, the "Formations" and "tendencies" on the second page are the main parts that have a significant impact.
打开FMRTE，可以读取每个AI主教练的面板，分成2个部分，其中第二页的“Formations”和“tendencies”是起主要影响的部分。

Tests have found that if these two parts of the two AI head coaches are swapped, under the control variable, their overall performance will also be swapped. This indicates that this information of the AI head coaches dominates all their behaviors in the game
测试发现，如果将两个AI主教练的这2个部分调换，在控制变量下，他们的整体成绩也会调换，这说明AI主教练的这些信息就是主导了他们在游戏里的一切行为

I selected 15 head coaches (the head coaches of 15 major teams at the beginning of the game) as the subjects of the test. On the player's side, they are free to choose the tactics I have selected. The attributes of both sides are exactly the same, and the attributes of each player are also the same. Moreover, the attributes of players are not distinguished based on the responsibilities of their positions and roles, thus avoiding the influence of bias and also preventing the impact of some "overly strong attributes"
我选择了15个主教练（15个主要球队在游戏开始时的主教练）作为被测试的对象。而玩家这边则是自由选择我选定的战术。双方的属性是完全一样的，每个球员的属性也是一样的，并且不按照位置角色职责来区分球员属性，从而避免倾向性的影响，也避免一些“过度强势属性”带来影响

These 15 head coaches (and their corresponding teams) are respectively:
这15个主教练（和他们对应的队伍）分别是：

XAVI哈维
Barcelona巴塞罗那

CARLO ANCELOTTI安切洛蒂
Real Madrid 皇家马德里

DIEGO SIMEONE西蒙尼
Atletico Madrid马德里竞技

LUIS ENRIQUE恩里克
Paris Saint-Germain Football Club巴黎圣日耳曼

MARCO ROSE罗斯
RasenBallsport Leipzig e.V莱比锡红牛

THOMAS TUCHEL图赫尔
Bayern Munich拜仁慕尼黑

EDIN TERZIC特尔季奇
Borussia Dortmund多特蒙德

STEFANO PIOLI皮奥利
Associazione Calcio Milan AC米兰

SIMONE INZAGHI因扎吉
Inter Milan 国际米兰

MASSIMILIANO ALLEGRI阿莱格里
Juventus尤文图斯

PEP GUARDIOLA瓜迪奥拉
Manchester City曼城

ERIK TEN HAG泰恩哈格
Manchester United曼联

MIKEL ARTETA阿尔特塔
Arsenal阿森纳

KLOPP克洛普
Liverpool Football Club利物浦

MAURICIO POCHETTINO波切蒂诺
Chelsea切尔西


每组测试若干赛季。得到这样一个表格。
如果你需要这些战术的下载，链接是：https://pixeldrain.com/u/Smeh7B9e
Each group is tested for several seasons. Obtain such a table.
If you need to download these tactics, the link is:https://pixeldrain.com/u/Smeh7B9e

Test result
↓


Simply looking at it, if the table is in blue, it indicates that this tactic is an advantage for this AI head coach.
If it is red, it represents a disadvantage.
Therefore, if a row of AI head coaches is all red, it indicates that he is very difficult to deal with. His tactics and "tendencies" are very much in line with the "META" of game engines.
简单的看，如果表格里是蓝色，那说明这个战术对这个AI主教练是优势，
如果是红色，则代表是劣势。
因此如果一个AI主教练的一列都是红色，那说明他很不好对付，他的战术和“tendencies”非常符合游戏引擎的“META”


You will notice that the intensity of the tactics of several AI head coaches is even on par with the player's "best of the best" tactics (indicated in red in the table). It even has an advantage over some of these AI head coaches.
While some other head coaches can find that almost all their tactics can defeat him by a large margin.
This indicates that some of these AI head coaches are actually using the "Meta" strategy

你会注意到，几个AI主教练的战术的强度甚至是和玩家“最好中的最好”的战术打成平手（在表格里是红色表示）。甚至对于其中的一些AI主教练有优势。
而另一些主教练，可以发现几乎所有战术都能以大优势赢他。
这说明其中一些AI主教练实际在用“Meta”战术


Among them, the best-performing head coach is:

MARCO ROSE
From RasenBallsport Leipzig E.V. RB Leipzig

DIEGO SIMEONE
From Atletico Madrid

STEFANO PIOLI
From Associazione Calcio Milan AC

其中，表现最好的主教练是：
MARCO ROSE 罗斯 来自 RasenBallsport Leipzig e.V 莱比锡红牛
DIEGO SIMEONE 西蒙尼 来自 Atletico Madrid 马德里竞技
STEFANO PIOLI 皮奥利 来自 Associazione Calcio Milan AC米兰


You will also notice that two of the tactics are the "Presets" tactics
They can only draw with some of the "worst" AI head coaches, such as "PEP GUARDIOLA"
你还会注意到，其中有2个战术是“Presets 预设”战术
他们仅能和一些“最差”的AI主教练，比如“PEP GUARDIOLA”打平手

9.The vast majority of the "Presets" tactics can be clearly seen in the table.
In fact, they are all at a disadvantage or an absolute disadvantage.
Because "Preset GegenPress 4231" in the table is already the best one among the "Presets" tactics.
而绝大部分的“Presets 预设”战术，在表格里可以明显看到，
实际上都处于劣势或者绝对的劣势，
因为表格里的“Preset GegenPress 4231”已经是“Presets 预设”战术里最好的一个了，


The performance of other "Presets" tactics was even worse
The table is another independent and earlier test used to compare the strengths and weaknesses of different "Presets" tactics
其他“Presets 预设”战术的表现更差
表格是另一个独立的更早期的测试，用于比较不同“Presets 预设”战术的强和弱

If you employ the "Presets" tactic or some less powerful tactics, you will naturally be at a significant disadvantage when facing the previous "meta" head coaches
如果你使用“Presets 预设”战术，或者一些没那么强的战术，那么你对上一部分“meta”主教练的时候就天然处于非常大的劣势

Part of the test tactics used is:https://pixeldrain.com/u/Smeh7B9e
使用到的测试战术的一部分是：https://pixeldrain.com/u/Smeh7B9e



I also tested a lot of other tactics, and the results showed that if you don 't do the' extreme Meta ', say, I cancelled 'tackling -- Get stuck in',
These meta tactics, against opponents such as MARCO ROSE, DIEGO SIMEONE, and STEFANO PIOLI, have a "loss rate" that is 10-20% higher than the "win rate", that is, they are overwhelmed

我还测试了更多的其他战术，结果表面，如果你不做到“极致Meta”，比如说，我取消掉了"凶狠逼抢"，
这些meta战术，对手比如MARCO ROSE，DIEGO SIMEONE，STEFANO PIOLI，“败率”会比“胜率”高10-20%，即被压倒了


(Test Condition setting and Preparation Section)
（测试条件设置和准备部分）
↓

10.Main part
主要部分

I have fixed to using one of the tactics, 3421.Change one attribute each time and test with these 15 head coaches as opponents
我固定了使用其中一个战术3421，每次改变一个属性，用这15个主教练做对手来测试

Part of the test data is:https://pixeldrain.com/u/KuemeJ9R[/b]
测试数据的一部分是：https://pixeldrain.com/u/KuemeJ9R


Test result
↓



The red part indicates:
Although the attributes have increased, the team's performance has not improved significantly (neutral), or has even worsened.
红色的部分表示：虽然属性增加，但是球队表现没有明显变好（中立），或者是变更差了。

The appearance of red is common:
This is in line with our initial inference ,The attribute is "tendency (number of attempts)" + "success rate"
红色的出现普遍：这符合了我们最初的推断,属性是“倾向性（尝试次数）”+“成功率”



Much earlier, I focused on testing the attributes of different roles/positions/responsibilities, but the results were very chaotic. The occasional "negative" attributes simply couldn't be explained properly.
Because an "ineffective" attribute can be understood as the engine not referencing it, but a "negative" attribute cannot be explained

Using the original theories (some widely circulated strategies and videos all repeat these theories, but have their own opinions, but all cannot do without the "highlighting attribute of character responsibilities" provided by the official)
"Position XX, because YYY is a highlighted attribute, the YYY attribute is even more needed." This doesn't explain why, theoretically, everyone knows that players need the YYY attribute, but why adding YYY would lead to a decline in the team's performance.

在更早以前，我对不同角色/位置/职责的属性的侧重测试，但是结果非常混乱，偶尔出现的“负面”属性根本无法正常解释，因为一个“无效果”的属性可以理解为引擎没有引用它，但是一个“负面”的属性，就无法解释

用原来的理论（一些广为流传的攻略和视频都是重复这些理论，只是有各自的见解，但是都离不开官方提供的“角色职责的高亮属性”）“XX位置，因为YYY是高亮属性，更需要YYY属性”，这解释不通为什么理论上，人人都知道球员需要YYY属性，但是为什么增加YYY会导致球队表现下降。The result of the test can conform to the theory.
Your tactics, the individual Settings within the tactics, are what you expect the players to do in the game (but without changing the "success rate".

The attributes of the player are "What will this player actively do?" plus "If his attributes are high, it will increase his success rate." This kind of enthusiasm or attempts is not necessarily beneficial (it attempts too much or an untimely attempt); sometimes it is harmful.

测试的结果能够符合理论。
你的战术、战术中的个人设置，是希望球员在比赛里怎么做（但是不改变“成功率”）
而球员的属性，则是”这个球员会积极的干什么“+“他的属性如果高会给他增加成功率”。这种积极性不一定是有益的(过多的尝试或者不合时宜的尝试)，有的时候是有害的。

Is this the reason why, in my previous multiple tests, some attributes showed "negative"
这就是为什么，在我之前的多种测试中，一部分属性出现“负面”的原因


This table is the initial attribute from 10 or 12 in this test. Change to 18.
The increase in the resulting winning rate

It can roughly be used to compare the importance of attributes (some of which may be inaccurate. For example, the Work Rate is very, very important at the stage of 1-10).
这个表格是这次测试里从10或者12的初始属性，
改变到18，
导致的胜率的增加值，

大致可以用来比较一下属性的重要性（其中会有一部分不准确，比如说Work Rate在1-10的阶段非常非常重要）


11.In the “Champions League Final”, test "Important Matches"
And the difference between "Go on holiday" and "commanded by the player himself"

在“欧冠决赛”，测试“Important Matches”
以及"度假"和“玩家亲自指挥”的区别





Change the hidden attribute “Important Matches = 18” to “Important Matches = 1”
Then compare the grades.
There are a total of 3 sets of data, with 60 matches in each set

把隐藏属性“大赛18”改成了“大赛1”
然后比较成绩，
一共有3组数据，每组60场比赛

The results show that major competitions do have an impact on important events like the UEFA Champions League.
And the most influential factor is the ratio of "expected goals" to "actual goals"
结果说明：大赛确实影响“欧冠”这种重要比赛。
并且影响最大的是“预计进球”和“实际进球”的比

“
（预计进球，实际进球）是4.07，3.03（手打）和4.03，3.20（度假）
（预计进球，实际进球）是0.74，0.67（手打，皇家马德里）和0.79，0.67（度假，皇家马德里）
……
传球成功率，玩家86.8%比皇家马德里85.0%
抢断成功率，玩家79.4%比皇家马德里77.7%
（手打）
传球成功率，玩家86.4%比皇家马德里84.9%
抢断成功率，玩家82.8%比皇家马德里72.8%
（度假）
……
玩家射门，射正，射正率（都是平均）
31.7，13.4，42.2%
（手打）
玩家射门，射正，射正率
31.9，13.7，42.0%
（度假）
……
3.38，2.73（手打），0.77，0.77（手打，皇家马德里）
3.30，2.43（度假），0.88，0.77（度假，皇家马德里）
……
（预计进球，实际进球）是4.65，4.53（手打）和4.45，4.50（度假）
（预计进球，实际进球）是0.57，0.43（手打，皇家马德里）和0.68，0.53（度假，皇家马德里）
”

Here is a series of data that is rather troublesome to translate. Moreover, I didn't make the table well, resulting in a very messy table, so I ignored it
这里有一连串翻译起来比较麻烦的数据，而且我没有很好的制作表格导致表格很混乱，就忽略了

Then, an earlier test has already proved that "consistency" is an effective and well-performing attribute
然后，更早的一个测试，已经证明了，“稳定consistency”是一个有效的效果不错的属性

“Important Matches”is same , an effective and well-performing attribute
大赛类似，是一个有效的效果不错的属性

Then comes "Go on holiday" and "commanded by the player himself"
然后比较“度假”和“手打”
Players do not use "Shouting" (I don't know how it is translated into English) in the competition.
Those with yellow cards and injuries in the midfield were substituted
玩家在比赛中不使用“喊话”（我不知道翻译成英语是怎样）。
中场黄牌和受伤的换下

Simple conclusion:
简单总结

"Player Control" does indeed present you with more shooting animations, but the final expected goal and the actual goal will be exactly the same as "Go on holiday"
For example, in "Go on holiday", your team takes 20 shots and scores 2 goals. However, in "Player Control", it takes 30 shots but still scores 2 goals
“玩家控制”确确实实的给你呈现了更多的射门动画，但是最后预计进球和实际进球，会和度假完全相同
比如说，“度假”统计你的球队射门20次，进2球，但是到“玩家控制”就会射门30次，但是还是进2球

The condition of our test was "no Shouting".
And "Shouting" can boost the morale of players. So, without changing the tactics at all, the greatest significance of the "Player Control" is to boost the morale of the team and thereby increase the winning rate (because morale is very important).
我们测试的条件是“不喊话”。
而“喊话”会提高玩家的士气。所以，在完全不改变战术的情况下，玩家操作的最大意义就是提高了球队的士气从而增加了胜率（因为士气很重要）

Outstanding work on this extensive testing! Here are the key findings for those, who struggle to read this 😉

Core Discovery: Attribute Mechanics

- Attributes = "Propensity (attempts)" + "Success Rate"
- Higher attributes make players attempt actions MORE OFTEN, not just better
- This can HARM team performance when players ignore tactical instructions to do what they prefer

Why Higher Attributes Can Be Negative
- Player with high Passing will attempt more passes (even when inappropriate)
- Player with high Finishing will "steal" shooting opportunities from better-positioned teammates
- Success rate improves, but overall team performance may decline
- Explains the "red results" where attribute increases hurt win rates

AI Manager META Hierarchy
Strongest (META-level AI):
- Marco Rose (RB Leipzig)
- Diego Simeone (Atletico Madrid)
- Stefano Pioli (AC Milan)

Weakest:
- Pep Guardiola
- Most other "famous" managers

Critical Finding: Some AI managers are using actual META tactics and are extremely difficult to beat

Preset Tactics Performance
- ALL preset tactics perform poorly against strong AI managers
- Even the best preset (GegenPress 4231) only draws with the WORST AI managers
- Using presets puts you at massive disadvantage against META AI managers

Extreme META Requirement
- Without "extreme META" settings (like "Get Stuck In", you lose 10-20% more than you win against top AI
- Small tactical compromises lead to being "overwhelmed" by strong opponents
- Every instruction matters at the highest level

Match Management Reality

- "Player Control" vs "Holiday" shows identical xG and actual goals
- More shot animations in player control, but same statistical outcomes
- Only real difference: team morale boost from "Shouting"
- Without shouting, manual control provides zero statistical advantage

Hidden Attributes Impact

- "Important Matches" significantly affects xG to actual goals conversion
- "Consistency" confirmed as highly effective attribute
- Position proficiency differences (even 18 vs 20) meaningfully impact performance

Revolutionary Implications

- Attribute stacking isn't always beneficial - can create tactical disobedience
- META tactics are mandatory against strong AI, not optional
- AI difficulty varies drastically based on manager tactical setup
- Manual control is largely cosmetic without morale management
- Every tactical detail matters at competitive levels

This research explains why "logical" attribute increases sometimes hurt performance and why certain AI teams feel impossibly difficult while others are pushovers. Exceptional scientific approach to testing FM24 mechanics!

harvestgreen22 said: I also tested a lot of other tactics, and the results showed that if you don 't do the' extreme Meta ', say, I cancelled 'tackling -- Get stuck in',
These meta tactics, against opponents such as MARCO ROSE, DIEGO SIMEONE, and STEFANO PIOLI, have a "loss rate" that is 10-20% higher than the "win rate", that is, they are overwhelmed Expand

I know it's not always THAT simple but can we make a rule of thumb that many 'semi-decent' tactics (or just self made) will improve significantly just by turning on 'Get stuck in' and probably using for most of the team Player's Instruction 'Tackle Harder' - since most meta tactics use both of these features?

So if your tactic is not total garbage that doesn't make sens, just tick press more, get stuck in, tackle harder and you're ready to overperform.