I have recently started to wrap my head around all the info I could find from amazing testers.
First, the concept of "meta" tactics, for which Gerrard feels like the unequaled master. For FM24, most of the best tactics are a variation of some 424 Gegenpress. Others can perform, but those seem to have the best results. Every FM iteration has their "meta" startegies that give statistically better results.
Second, the concept of "meta" attibutes, for which we all agree that Pace/Acceleration dominate for some time now, with some tweaks in the weighting in different opuses. Many have tested this and the quality of data and analysis amazes me.
Where I find myself left with a sense of misunderstanding is the why. So there is my question:
Is the match engine: 1 Complex and as advertised, but with inherent (and unavoidable... this is a simulation) imbalances that make some tactics/attributes stand out more in the "results" department 2 An illusion of complexity, but actually very simplified, with barely any real-time micro-calculations that are attributes-specific
I personnally feel like all the data we have does NOT actually point toward option 2, unlike many comments and analysis I find in FM-Arena. However, I have never done rigorous testing, data mining, code testing, or anything. So I just want to share my hypothesis so I can better understand FM's match engine and, mostly, how much fun us tacticians can have with it
Hypothesis: The match engine is ACTUALLY complex and taking into accounts live micro-computations that use all attributes and tactic tweaks, even if attribute testing shows that many attributes have no impact on results.
How does that hypothesis still fits the concept of meta tactic and meta attributes: Let's say that you want to play positionnal play, based on many passes in tight spaces. If the match engine is made in such a way that no player can find a real good space, you will never achieve good positional play / tiki-taka style. Even a player with 20 Off-the-ball and 1 Pace, could find the best "pocket of space" ACCORDING to the ME, if the ME does not suggest the best spot, his "amazing skill" won't help at all.
On the other hand, if the player is the opposite (20 Pace, 1 Off the Ball), even if he doesn't find a good pocket of space, he'll be the first on the ball anyways. So as a "result", fast players would be better even in "positional play" possession tactics. Not because he finds better pockets, but because the ME can't find good pockets for any of both those players... so the fast one, at least, can still be useful.
Another example: If a player receives the ball and has 3 options. 1 good, 2 bad. Let's say the Engine looks at "decision" for a fast computation of the next play. If the ME is made so a 1-Decision player has 80% chances of taking the right decision while a 20-Decision player has a 95% chance, then, very clearly, the "Decision" attribute ha little-to-no value. Not because it is not used in the computation, but because it does not have a realistic impact ingame.
Many have hypothesised that a player with high Pace makes better decisions (leaning toward the idea that the match engine is actually over-simplified with a "general" rating for each actions). But just looking at the results does not necessarily points toward that. A fast player might provide better results because the ME is not good at providing intelligent options anyways, so the fast player will just be generally way off his marker when he makes his move.
Where my hypothesis falls short Vision, decision, off the ball, positionning are attributes that are very tough to understand. If the ME forces players to find some specific pockets of space depending on their roles/attributes and those pockets are not the "smartest ones" (because you know... football IS complicated), a player with high vision might still do weird stuff ingame. Just because the ME might think it's good.
But when I see someone with 3 Long shots pull a top-corner screamer in, then I feel a little bit weirded out. So this is why I'm still very much wondering how the ME actually works, and how micro-computations within a game are actually made.
How are attributes testing working with different tactics If attributes testing are done with a "meta tactic" (which is wide/fast/gegenpress-heavy), then maybe we are actually skewing our perception of attributes importance toward physical a lot. Let's say we look at a hyper-conservative possession-heavy tactic, how would attributes scale up? I would guess that whatever the attributes, those tactics will score lower. This has been thoroughly seen. But would Pace/Accel/Dribb still be that high up? Would Vision/Off-the-Ball show their head slightly above the water of "no impact"?
That might provide insight on how the ME actually works.
******************************
I conclude this by saying that I see many amazing analystst and data crunchers in these threads. My question really leads to this:
- As a FM tactician who would LOVE to create club DNAs and "chose" a style, adjusting games and using the complexity of our favorite sport, can the ME actually evolve the right way or is it too simplified and there is no hope in actually finding nuances in performance and style of individuals outside of meta attributes?
If my hypothesis is right, then the problem is simply that the ME is not yet providing players with "smarts", or giving enough weight of some attributes (IMO, 1 vision means the player looks at his feet 99% of the time, maybe the ME just says: let's drop his options by 5%), or just not attributing "success probability" to actions the right way, making a 20 decision player still making decisions that won't help the team. Meaning that possession tactics don't work because smart players don't lead to smart plays. But fast players lead to easy plays.
If my hypothesis is wrong, then yeah... the ME won't realistically evolve anytime soon and we'll have to keep playing pace monsters to reproduce possession-heavy tactics that can't really work anyways.
P.S. Training is not as advertised, that much seems certain. But my questionning here is solely on the ME. Training is it's own subject.
You guys (or gals) have tested things WAY more than the average Joe. Your inputs on this might help MANY players better understand their favorite games. As I also saw many people getting disenchanted with the ME, while I feel all of your findings are not as damning as it seems.
I think you are approaching the question from the wrong direction. Every single game has a meta = strategy or decisions that produce the best result. Doesn't matter if it's tic tac toe or something more complex like chess or FM. So for any given set of conditions there is an optimal solution, whether we find it or not and how close we get is kind of irrelevant.
The way (computer) games tend to solve this issue is by making slight changes, buffs and nerfs if you will, that force players to adapt and switch the strategy. Where SI fell short is that they simply didn't or couldn't adjust this complex ME enough for it to make a significant difference - or maybe they did at a secondary attributes level, but physicals have always been king for more than 2 decades now. If you can't tackle you can just outsprint and outmuscle someone for example. Another thing in favor of physicals is that they are not affected by consistency unlike the other categories. I don't expect any more huge changes in attribute strength this FM, they already did more than I expected but we can hope.
Fairly sure attributes have been tested in other tactical systems and produced similar results.
Yarema said: I think you are approaching the question from the wrong direction. Every single game has a meta = strategy or decisions that produce the best result. Doesn't matter if it's tic tac toe or something more complex like chess or FM. So for any given set of conditions there is an optimal solution, whether we find it or not and how close we get is kind of irrelevant.
The way (computer) games tend to solve this issue is by making slight changes, buffs and nerfs if you will, that force players to adapt and switch the strategy. Where SI fell short is that they simply didn't or couldn't adjust this complex ME enough for it to make a significant difference - or maybe they did at a secondary attributes level, but physicals have always been king for more than 2 decades now. If you can't tackle you can just outsprint and outmuscle someone for example. Another thing in favor of physicals is that they are not affected by consistency unlike the other categories. I don't expect any more huge changes in attribute strength this FM, they already did more than I expected but we can hope.
Fairly sure attributes have been tested in other tactical systems and produced similar results. Expand
Just some additional input, problems with multiple variables and objectives often have multiple solutions that are not comparable to each other. In case of Football Manager, you could have three tactics in which A > B > C > A. Some tactic could also have higher influence of player attributes than another, or some tactic could be better suited for a match where you just need to hold a score, without worrying about building advantage.
I might not have been clear. I am fully aware that attributes have their strenghts and that this is all a multi-variable problem. The question is whether the ME makes micro-computations "on the spot" with true complex interactions between attributes or if it is only a render of more macro-computations that only ponderates "success" with main attributes.
In other words, is it true that "high pace = best decisions" or is the huge success of pace more an indicator that ME does not make "decisions" and "vision" shine enough during the play-by-play computations.
It is certainly complex rather than simple. It is mainly the training & weighting system that is broken.
The two most important player attributes are speed & work rate. This reflects reality accurately. The problem is when you can use HarvestGreen22's training regimes to turn literally any player into an elite player (pace/acc 8 > 20), and when you can give a full back 20 pace/acc/jump/nat/long/fin/drib/det/agg/off the ball/work rate and he'll be 1 CA with a handful of points left to spare.
I think you are correct in your hypothesis about how attributes are processed by the ME. Of course until we can see the game's code, we have to make our best judgement based on results and what SI staff have said over the years.
So this is what I understand to be the case:
The ME uses typically 2-3 attributes (sometimes up to 5) for each game event (i.e. passing + work rate = pass attempt)
An ATTEMPT ('choice' ) is made FIRST A COMPLETION ('execution' ) is made SECOND
i.e. passing + work rate = pass attempt passing + decisions = pass completed
As you hypothesize, for many/most players, the only way to win is not to play at all. Instead of going for a slim chance of a slim chance of a pass leading to a slim chance of a goal, if you go with passing '1', then decisions can also be '1' since it almost never comes into play now, and you can now instead rely on dribbling '20' with pace/acc '20' to run up the park and score in ~17% of instances with [dribbling (20) + balance (7) = 67%(?) evasion chance] x [finishing (7) + composure (7) + decisions (1) = 15%(?) scoring chance].
Something I'm observing is that '20' pace/acc will win games where even '17' pace/acc would see you get demolished. I suspect this is because you simply run past/around the opposition's players, so the intercept & block chances are left out of the calculations. This is showing up in my 1 CA player testing, but I think it would still hold partially true for normal gameplay where you have weaker technicals/mentals/other physicals than your opponent.
Another thing I have observed is that while changing full backs to have drib/long/finish 20 (all players still only 1 CA) resulted in a lot more premier league wins, the full backs themselves rarely scored goals, or even assists. And looking at the match highlights, it doesn't appear to me to be that say fullbacks are making big forward runs up the side leading to later goals. This is still a mystery to me right now.
I do suspect that tactics give favor to certain attributes, and that meta tactics favor/exploit high pace/acc/drib, but I also think that only meta tactics tend to work, especially since your AI opponents (at least, the better ones) are also going to be utilizing these same meta tactics against you. I have noticed however that you can use a low pace/acc target man with no ill effect.
I think there may be something in thinking about the fact that attributes aren't denoting actual ability (i.e. you can even see in the 3d engine, that a dribbling '1' player can still dribble), but rather the only factors in decision outcomes. So a dribbling '1' player from Bedford Town dribbles just as good as Messi would in the game; it's only when Messi tries to tackle him that there is a difference. I think this partly explains why pace is so important, but technicals could be '1' and still do ok. Pace is active all the time, a slow player is always obvious and hampered in the 3D engine. But often a '1' dribbler looks just like a '20' dribbler (particularly when he has high pace to lengthen it out). Therefore the whole 'a player is only as good as his worst minimum' logic shouldn't be applied.
A similar potential misconception is that an attribute is simultaneously tendency and success rate. It seems true, but only to an extent: passing '20' gives you both more attempts & success, however a player with '1' passing still makes passes anyway, so probably '1' passing & '20' pace does more actual passes than '20' passing and '1' pace. So really '20' pass you could look at as say +5% tendency, +20% success, rather than +95% tendency +95% success.
Here's two other quirks I feel are worth mentioning:
If you go by what the staff-written guides say, strength alone can be used in lieu of tackling + marking + aggression. Strength is pretty highly weighted, but then again, it isn't subject to the consistency CA decrease like tackling/marking is, and it also says that strength reduces fouls as they can muscle the player off the ball instead of sliding in. For ST, tackling/marking is 1 weight, strength is 6, and a foul probably won't result in a penalty against you, so low strength probably makes sense here. But for DC, marking is 8 weight, tackling 5, and strength 6, and sliding in probably results in more conceded penalties. I'm yet to test this though.
If you think about it, you can set consistency to 1 for 1 CA players and it will make no difference, as consistency reduces CA for technicals to use. My early testing of this finds it does indeed make no difference. I suppose you could utilize this in a very low league team.
You guys (or gals) have tested things WAY more than the average Joe. Your inputs on this might help MANY players better understand their favorite games. As I also saw many people getting disenchanted with the ME, while I feel all of your findings are not as damning as it seems. Expand
“If attributes testing are done with a "meta tactic" (which is wide/fast/gegenpress-heavy), then maybe we are actually skewing our perception of attributes importance toward physical a lot”
——In fact, I have also conducted tests in other tactics .Not Meta tactics , It's the preset tactics .
The conclusion is basically the same (some minor attributes have slightly change , but the overall situation remains unchanged).
Since my native language is not English,
It's too troublesome for me to organize these things and translate them into English.
So I only posted the result under the condition of the Gegenpress Meta tactic (which is also the type of tactic used by most casual players).
Panneton0 said: I might not have been clear. I am fully aware that attributes have their strenghts and that this is all a multi-variable problem. The question is whether the ME makes micro-computations "on the spot" with true complex interactions between attributes or if it is only a render of more macro-computations that only ponderates "success" with main attributes.
In other words, is it true that "high pace = best decisions" or is the huge success of pace more an indicator that ME does not make "decisions" and "vision" shine enough during the play-by-play computations. Expand
High pace is not about decisions, it is about performing well even when players make bad decisions. A pass that goes too far can still be recovered simply because the player moves faster than the opponent, or a character that takes too long to anticipate can still intercept the ball simply because he moves like an arrow. That's what makes mobility attributes so strong, because they make bad decisions into good decisions.
Panneton0 said: I have recently started to wrap my head around all the info I could find from amazing testers.
First, the concept of "meta" tactics, for which Gerrard feels like the unequaled master. For FM24, most of the best tactics are a variation of some 424 Gegenpress. Others can perform, but those seem to have the best results. Every FM iteration has their "meta" startegies that give statistically better results.
Second, the concept of "meta" attibutes, for which we all agree that Pace/Acceleration dominate for some time now, with some tweaks in the weighting in different opuses. Many have tested this and the quality of data and analysis amazes me.
Where I find myself left with a sense of misunderstanding is the why. So there is my question:
Is the match engine: 1 Complex and as advertised, but with inherent (and unavoidable... this is a simulation) imbalances that make some tactics/attributes stand out more in the "results" department 2 An illusion of complexity, but actually very simplified, with barely any real-time micro-calculations that are attributes-specific
I personnally feel like all the data we have does NOT actually point toward option 2, unlike many comments and analysis I find in FM-Arena. However, I have never done rigorous testing, data mining, code testing, or anything. So I just want to share my hypothesis so I can better understand FM's match engine and, mostly, how much fun us tacticians can have with it
Hypothesis: The match engine is ACTUALLY complex and taking into accounts live micro-computations that use all attributes and tactic tweaks, even if attribute testing shows that many attributes have no impact on results.
How does that hypothesis still fits the concept of meta tactic and meta attributes: Let's say that you want to play positionnal play, based on many passes in tight spaces. If the match engine is made in such a way that no player can find a real good space, you will never achieve good positional play / tiki-taka style. Even a player with 20 Off-the-ball and 1 Pace, could find the best "pocket of space" ACCORDING to the ME, if the ME does not suggest the best spot, his "amazing skill" won't help at all.
On the other hand, if the player is the opposite (20 Pace, 1 Off the Ball), even if he doesn't find a good pocket of space, he'll be the first on the ball anyways. So as a "result", fast players would be better even in "positional play" possession tactics. Not because he finds better pockets, but because the ME can't find good pockets for any of both those players... so the fast one, at least, can still be useful.
Another example: If a player receives the ball and has 3 options. 1 good, 2 bad. Let's say the Engine looks at "decision" for a fast computation of the next play. If the ME is made so a 1-Decision player has 80% chances of taking the right decision while a 20-Decision player has a 95% chance, then, very clearly, the "Decision" attribute ha little-to-no value. Not because it is not used in the computation, but because it does not have a realistic impact ingame.
Many have hypothesised that a player with high Pace makes better decisions (leaning toward the idea that the match engine is actually over-simplified with a "general" rating for each actions). But just looking at the results does not necessarily points toward that. A fast player might provide better results because the ME is not good at providing intelligent options anyways, so the fast player will just be generally way off his marker when he makes his move.
Where my hypothesis falls short Vision, decision, off the ball, positionning are attributes that are very tough to understand. If the ME forces players to find some specific pockets of space depending on their roles/attributes and those pockets are not the "smartest ones" (because you know... football IS complicated), a player with high vision might still do weird stuff ingame. Just because the ME might think it's good.
But when I see someone with 3 Long shots pull a top-corner screamer in, then I feel a little bit weirded out. So this is why I'm still very much wondering how the ME actually works, and how micro-computations within a game are actually made.
How are attributes testing working with different tactics If attributes testing are done with a "meta tactic" (which is wide/fast/gegenpress-heavy), then maybe we are actually skewing our perception of attributes importance toward physical a lot. Let's say we look at a hyper-conservative possession-heavy tactic, how would attributes scale up? I would guess that whatever the attributes, those tactics will score lower. This has been thoroughly seen. But would Pace/Accel/Dribb still be that high up? Would Vision/Off-the-Ball show their head slightly above the water of "no impact"?
That might provide insight on how the ME actually works.
******************************
I conclude this by saying that I see many amazing analystst and data crunchers in these threads. My question really leads to this:
- As a FM tactician who would LOVE to create club DNAs and "chose" a style, adjusting games and using the complexity of our favorite sport, can the ME actually evolve the right way or is it too simplified and there is no hope in actually finding nuances in performance and style of individuals outside of meta attributes?
If my hypothesis is right, then the problem is simply that the ME is not yet providing players with "smarts", or giving enough weight of some attributes (IMO, 1 vision means the player looks at his feet 99% of the time, maybe the ME just says: let's drop his options by 5%), or just not attributing "success probability" to actions the right way, making a 20 decision player still making decisions that won't help the team. Meaning that possession tactics don't work because smart players don't lead to smart plays. But fast players lead to easy plays.
If my hypothesis is wrong, then yeah... the ME won't realistically evolve anytime soon and we'll have to keep playing pace monsters to reproduce possession-heavy tactics that can't really work anyways.
P.S. Training is not as advertised, that much seems certain. But my questionning here is solely on the ME. Training is it's own subject. Expand
Whendiscussing this, we need to take the following into account:
Players with high Pace and Acceleration make perfect through balls (even though their passing and decisions are low). They finish brilliantly in one on ones with the keeper (even though their finishing is low). They defend extremely well (even though their marking and tackling are low). This has been the case since CM 03/04 (at least as far as I know. I never played the earlier CM games, but it's probably the same in those as well).
You notice this even more clearly when watching matches on comprehensive mode. A fast player doesn’t just run quickly, he plays way above his potential and even scores goals.
This pushes me more toward the second option (illusion of complexity).
None of us have any concrete evidence, we can only make guesses. But I also think this, if it were the first option, this issue wouldn’t have stayed the same for 20 years despite all the resources and the team of coders they have. It’s not that SI doesn’t know how to fix it, fixing it would require a new match engine, and they either won’t or can’t do it.
By the way, a bug from the FM 2005 era just came to my mind and I wanted to mention it because i think it's related to the topic.
In FM 2005, when the match kicked off, if you clicked on the “overview” panel and checked the fixture, you could actually see the final score of the match before it ended. And when the match finished, it always ended with exactly the score shown there. This shows that the score is determined the moment the match kicks off.
They fixed it quickly with a patch, but since Steam wasn’t really popular back then, updates didn’t reach everyone instantly like they do today. So I remember playing with that bug for quite a long time.
I’m not even sure much has changed since then. Not being sure is the scary part. We should be sure about this after 20 years, but SI just doesn't work that way. jesus christ man, SI should’ve gone bankrupt decades ago. They’ve been putting us on this shitshow for 20 years. Thankfully, with FM 26 this whole show is finally over, and at this point waiting for moves from competitor studios makes more sense than waiting for FM 27 tbh.
Sanfierro said: By the way, a bug from the FM 2005 era just came to my mind and I wanted to mention it because i think it's related to the topic.
In FM 2005, when the match kicked off, if you clicked on the “overview” panel and checked the fixture, you could actually see the final score of the match before it ended. And when the match finished, it always ended with exactly the score shown there. This shows that the score is determined the moment the match kicks off.
They fixed it quickly with a patch, but since Steam wasn’t really popular back then, updates didn’t reach everyone instantly like they do today. So I remember playing with that bug for quite a long time.
I’m not even sure much has changed since then. Not being sure is the scary part. We should be sure about this after 20 years, but SI just doesn't work that way. jesus christ man, SI should’ve gone bankrupt decades ago. They’ve been putting us on this shitshow for 20 years. Thankfully, with FM 26 this whole show is finally over, and at this point waiting for moves from competitor studios makes more sense than waiting for FM 27 tbh. Expand
Results are not determined at the start of the match. That is pretty easy to check, you can simply start with a very bad tactic, then change to a very good one after 5 or 10 minutes, and result will be consistent with the good tactic.
ZaZ said: Results are not determined at the start of the match. That is pretty easy to check, you can simply start with a very bad tactic, then change to a very good one after 5 or 10 minutes, and result will be consistent with the good tactic. Expand There has been a theory going around for a while now that basically the result is predetermined when you enter the match and that highlights are created to suit the final result and not the other way around. However any changes made during the match (tactical, subs...) will reset the calculated end result and thus highlights. I think it has been largely confirmed by SI staff at some point but that was a while ago and there may have been changes since.
So I guess that the "precomputed" result (if this is the case, I know nothing is yet confirmed) is not only the result, but also pass maps / positions / etc. Because ingame highlights still "kinda" feel like they fit with the tactic (giving the potentially better illusion of micro-computations). Thanks for all your inputs, this is all very interesting
I have recently started to wrap my head around all the info I could find from amazing testers.
First, the concept of "meta" tactics, for which Gerrard feels like the unequaled master. For FM24, most of the best tactics are a variation of some 424 Gegenpress. Others can perform, but those seem to have the best results. Every FM iteration has their "meta" startegies that give statistically better results.
Second, the concept of "meta" attibutes, for which we all agree that Pace/Acceleration dominate for some time now, with some tweaks in the weighting in different opuses. Many have tested this and the quality of data and analysis amazes me.
It is very obvious to me, after reading all of the rigorous works like
https://fm-arena.com/thread/14009-attribute-testing-football-manager-24/page-1/
https://fm-arena.com/thread/15934-summary-of-recent-findings-for-optimal-play-in-fm24-amp-fm26/page-1/
Plus all the tactics hall of fame that there is indeed a way to play the game to optimize the RESULT.
Where I find myself left with a sense of misunderstanding is the why. So there is my question:
Is the match engine:
1 Complex and as advertised, but with inherent (and unavoidable... this is a simulation) imbalances that make some tactics/attributes stand out more in the "results" department
2 An illusion of complexity, but actually very simplified, with barely any real-time micro-calculations that are attributes-specific
I personnally feel like all the data we have does NOT actually point toward option 2, unlike many comments and analysis I find in FM-Arena. However, I have never done rigorous testing, data mining, code testing, or anything. So I just want to share my hypothesis so I can better understand FM's match engine and, mostly, how much fun us tacticians can have with it
Hypothesis: The match engine is ACTUALLY complex and taking into accounts live micro-computations that use all attributes and tactic tweaks, even if attribute testing shows that many attributes have no impact on results.
How does that hypothesis still fits the concept of meta tactic and meta attributes:
Let's say that you want to play positionnal play, based on many passes in tight spaces. If the match engine is made in such a way that no player can find a real good space, you will never achieve good positional play / tiki-taka style. Even a player with 20 Off-the-ball and 1 Pace, could find the best "pocket of space" ACCORDING to the ME, if the ME does not suggest the best spot, his "amazing skill" won't help at all.
On the other hand, if the player is the opposite (20 Pace, 1 Off the Ball), even if he doesn't find a good pocket of space, he'll be the first on the ball anyways. So as a "result", fast players would be better even in "positional play" possession tactics. Not because he finds better pockets, but because the ME can't find good pockets for any of both those players... so the fast one, at least, can still be useful.
Another example: If a player receives the ball and has 3 options. 1 good, 2 bad. Let's say the Engine looks at "decision" for a fast computation of the next play. If the ME is made so a 1-Decision player has 80% chances of taking the right decision while a 20-Decision player has a 95% chance, then, very clearly, the "Decision" attribute ha little-to-no value. Not because it is not used in the computation, but because it does not have a realistic impact ingame.
Many have hypothesised that a player with high Pace makes better decisions (leaning toward the idea that the match engine is actually over-simplified with a "general" rating for each actions). But just looking at the results does not necessarily points toward that. A fast player might provide better results because the ME is not good at providing intelligent options anyways, so the fast player will just be generally way off his marker when he makes his move.
Where my hypothesis falls short
Vision, decision, off the ball, positionning are attributes that are very tough to understand. If the ME forces players to find some specific pockets of space depending on their roles/attributes and those pockets are not the "smartest ones" (because you know... football IS complicated), a player with high vision might still do weird stuff ingame. Just because the ME might think it's good.
But when I see someone with 3 Long shots pull a top-corner screamer in, then I feel a little bit weirded out. So this is why I'm still very much wondering how the ME actually works, and how micro-computations within a game are actually made.
How are attributes testing working with different tactics
If attributes testing are done with a "meta tactic" (which is wide/fast/gegenpress-heavy), then maybe we are actually skewing our perception of attributes importance toward physical a lot. Let's say we look at a hyper-conservative possession-heavy tactic, how would attributes scale up? I would guess that whatever the attributes, those tactics will score lower. This has been thoroughly seen. But would Pace/Accel/Dribb still be that high up? Would Vision/Off-the-Ball show their head slightly above the water of "no impact"?
That might provide insight on how the ME actually works.
******************************
I conclude this by saying that I see many amazing analystst and data crunchers in these threads. My question really leads to this:
- As a FM tactician who would LOVE to create club DNAs and "chose" a style, adjusting games and using the complexity of our favorite sport, can the ME actually evolve the right way or is it too simplified and there is no hope in actually finding nuances in performance and style of individuals outside of meta attributes?
If my hypothesis is right, then the problem is simply that the ME is not yet providing players with "smarts", or giving enough weight of some attributes (IMO, 1 vision means the player looks at his feet 99% of the time, maybe the ME just says: let's drop his options by 5%), or just not attributing "success probability" to actions the right way, making a 20 decision player still making decisions that won't help the team. Meaning that possession tactics don't work because smart players don't lead to smart plays. But fast players lead to easy plays.
If my hypothesis is wrong, then yeah... the ME won't realistically evolve anytime soon and we'll have to keep playing pace monsters to reproduce possession-heavy tactics that can't really work anyways.
P.S. Training is not as advertised, that much seems certain. But my questionning here is solely on the ME. Training is it's own subject.
@Zippo
@harvestgreen22
@Gerrard
You guys (or gals) have tested things WAY more than the average Joe. Your inputs on this might help MANY players better understand their favorite games. As I also saw many people getting disenchanted with the ME, while I feel all of your findings are not as damning as it seems.
I think you are approaching the question from the wrong direction. Every single game has a meta = strategy or decisions that produce the best result. Doesn't matter if it's tic tac toe or something more complex like chess or FM. So for any given set of conditions there is an optimal solution, whether we find it or not and how close we get is kind of irrelevant.
The way (computer) games tend to solve this issue is by making slight changes, buffs and nerfs if you will, that force players to adapt and switch the strategy. Where SI fell short is that they simply didn't or couldn't adjust this complex ME enough for it to make a significant difference - or maybe they did at a secondary attributes level, but physicals have always been king for more than 2 decades now. If you can't tackle you can just outsprint and outmuscle someone for example. Another thing in favor of physicals is that they are not affected by consistency unlike the other categories.
I don't expect any more huge changes in attribute strength this FM, they already did more than I expected but we can hope.
Fairly sure attributes have been tested in other tactical systems and produced similar results.
Yarema said: I think you are approaching the question from the wrong direction. Every single game has a meta = strategy or decisions that produce the best result. Doesn't matter if it's tic tac toe or something more complex like chess or FM. So for any given set of conditions there is an optimal solution, whether we find it or not and how close we get is kind of irrelevant.
The way (computer) games tend to solve this issue is by making slight changes, buffs and nerfs if you will, that force players to adapt and switch the strategy. Where SI fell short is that they simply didn't or couldn't adjust this complex ME enough for it to make a significant difference - or maybe they did at a secondary attributes level, but physicals have always been king for more than 2 decades now. If you can't tackle you can just outsprint and outmuscle someone for example. Another thing in favor of physicals is that they are not affected by consistency unlike the other categories.
I don't expect any more huge changes in attribute strength this FM, they already did more than I expected but we can hope.
Fairly sure attributes have been tested in other tactical systems and produced similar results.
Just some additional input, problems with multiple variables and objectives often have multiple solutions that are not comparable to each other. In case of Football Manager, you could have three tactics in which A > B > C > A. Some tactic could also have higher influence of player attributes than another, or some tactic could be better suited for a match where you just need to hold a score, without worrying about building advantage.
I might not have been clear. I am fully aware that attributes have their strenghts and that this is all a multi-variable problem. The question is whether the ME makes micro-computations "on the spot" with true complex interactions between attributes or if it is only a render of more macro-computations that only ponderates "success" with main attributes.
In other words, is it true that "high pace = best decisions" or is the huge success of pace more an indicator that ME does not make "decisions" and "vision" shine enough during the play-by-play computations.
It is certainly complex rather than simple. It is mainly the training & weighting system that is broken.
The two most important player attributes are speed & work rate. This reflects reality accurately. The problem is when you can use HarvestGreen22's training regimes to turn literally any player into an elite player (pace/acc 8 > 20), and when you can give a full back 20 pace/acc/jump/nat/long/fin/drib/det/agg/off the ball/work rate and he'll be 1 CA with a handful of points left to spare.
I think you are correct in your hypothesis about how attributes are processed by the ME. Of course until we can see the game's code, we have to make our best judgement based on results and what SI staff have said over the years.
So this is what I understand to be the case:
The ME uses typically 2-3 attributes (sometimes up to 5) for each game event (i.e. passing + work rate = pass attempt)
An ATTEMPT ('choice' ) is made FIRST
A COMPLETION ('execution' ) is made SECOND
i.e.
passing + work rate = pass attempt
passing + decisions = pass completed
As you hypothesize, for many/most players, the only way to win is not to play at all. Instead of going for a slim chance of a slim chance of a pass leading to a slim chance of a goal, if you go with passing '1', then decisions can also be '1' since it almost never comes into play now, and you can now instead rely on dribbling '20' with pace/acc '20' to run up the park and score in ~17% of instances with [dribbling (20) + balance (7) = 67%(?) evasion chance] x [finishing (7) + composure (7) + decisions (1) = 15%(?) scoring chance].
Something I'm observing is that '20' pace/acc will win games where even '17' pace/acc would see you get demolished. I suspect this is because you simply run past/around the opposition's players, so the intercept & block chances are left out of the calculations. This is showing up in my 1 CA player testing, but I think it would still hold partially true for normal gameplay where you have weaker technicals/mentals/other physicals than your opponent.
Another thing I have observed is that while changing full backs to have drib/long/finish 20 (all players still only 1 CA) resulted in a lot more premier league wins, the full backs themselves rarely scored goals, or even assists. And looking at the match highlights, it doesn't appear to me to be that say fullbacks are making big forward runs up the side leading to later goals. This is still a mystery to me right now.
I do suspect that tactics give favor to certain attributes, and that meta tactics favor/exploit high pace/acc/drib, but I also think that only meta tactics tend to work, especially since your AI opponents (at least, the better ones) are also going to be utilizing these same meta tactics against you. I have noticed however that you can use a low pace/acc target man with no ill effect.
I think there may be something in thinking about the fact that attributes aren't denoting actual ability (i.e. you can even see in the 3d engine, that a dribbling '1' player can still dribble), but rather the only factors in decision outcomes. So a dribbling '1' player from Bedford Town dribbles just as good as Messi would in the game; it's only when Messi tries to tackle him that there is a difference. I think this partly explains why pace is so important, but technicals could be '1' and still do ok. Pace is active all the time, a slow player is always obvious and hampered in the 3D engine. But often a '1' dribbler looks just like a '20' dribbler (particularly when he has high pace to lengthen it out). Therefore the whole 'a player is only as good as his worst minimum' logic shouldn't be applied.
A similar potential misconception is that an attribute is simultaneously tendency and success rate. It seems true, but only to an extent: passing '20' gives you both more attempts & success, however a player with '1' passing still makes passes anyway, so probably '1' passing & '20' pace does more actual passes than '20' passing and '1' pace. So really '20' pass you could look at as say +5% tendency, +20% success, rather than +95% tendency +95% success.
Here's two other quirks I feel are worth mentioning:
If you go by what the staff-written guides say, strength alone can be used in lieu of tackling + marking + aggression. Strength is pretty highly weighted, but then again, it isn't subject to the consistency CA decrease like tackling/marking is, and it also says that strength reduces fouls as they can muscle the player off the ball instead of sliding in. For ST, tackling/marking is 1 weight, strength is 6, and a foul probably won't result in a penalty against you, so low strength probably makes sense here. But for DC, marking is 8 weight, tackling 5, and strength 6, and sliding in probably results in more conceded penalties. I'm yet to test this though.
If you think about it, you can set consistency to 1 for 1 CA players and it will make no difference, as consistency reduces CA for technicals to use. My early testing of this finds it does indeed make no difference. I suppose you could utilize this in a very low league team.
Panneton0 said: @Zippo
@harvestgreen22
@Gerrard
You guys (or gals) have tested things WAY more than the average Joe. Your inputs on this might help MANY players better understand their favorite games. As I also saw many people getting disenchanted with the ME, while I feel all of your findings are not as damning as it seems.
“If attributes testing are done with a "meta tactic" (which is wide/fast/gegenpress-heavy), then maybe we are actually skewing our perception of attributes importance toward physical a lot”
——In fact, I have also conducted tests in other tactics .Not Meta tactics , It's the preset tactics .
The conclusion is basically the same (some minor attributes have slightly change , but the overall situation remains unchanged).
Since my native language is not English,
It's too troublesome for me to organize these things and translate them into English.
So I only posted the result under the condition of the Gegenpress Meta tactic (which is also the type of tactic used by most casual players).
Panneton0 said: I might not have been clear. I am fully aware that attributes have their strenghts and that this is all a multi-variable problem. The question is whether the ME makes micro-computations "on the spot" with true complex interactions between attributes or if it is only a render of more macro-computations that only ponderates "success" with main attributes.
In other words, is it true that "high pace = best decisions" or is the huge success of pace more an indicator that ME does not make "decisions" and "vision" shine enough during the play-by-play computations.
High pace is not about decisions, it is about performing well even when players make bad decisions. A pass that goes too far can still be recovered simply because the player moves faster than the opponent, or a character that takes too long to anticipate can still intercept the ball simply because he moves like an arrow. That's what makes mobility attributes so strong, because they make bad decisions into good decisions.
Panneton0 said: I have recently started to wrap my head around all the info I could find from amazing testers.
First, the concept of "meta" tactics, for which Gerrard feels like the unequaled master. For FM24, most of the best tactics are a variation of some 424 Gegenpress. Others can perform, but those seem to have the best results. Every FM iteration has their "meta" startegies that give statistically better results.
Second, the concept of "meta" attibutes, for which we all agree that Pace/Acceleration dominate for some time now, with some tweaks in the weighting in different opuses. Many have tested this and the quality of data and analysis amazes me.
It is very obvious to me, after reading all of the rigorous works like
https://fm-arena.com/thread/14009-attribute-testing-football-manager-24/page-1/
https://fm-arena.com/thread/15934-summary-of-recent-findings-for-optimal-play-in-fm24-amp-fm26/page-1/
Plus all the tactics hall of fame that there is indeed a way to play the game to optimize the RESULT.
Where I find myself left with a sense of misunderstanding is the why. So there is my question:
Is the match engine:
1 Complex and as advertised, but with inherent (and unavoidable... this is a simulation) imbalances that make some tactics/attributes stand out more in the "results" department
2 An illusion of complexity, but actually very simplified, with barely any real-time micro-calculations that are attributes-specific
I personnally feel like all the data we have does NOT actually point toward option 2, unlike many comments and analysis I find in FM-Arena. However, I have never done rigorous testing, data mining, code testing, or anything. So I just want to share my hypothesis so I can better understand FM's match engine and, mostly, how much fun us tacticians can have with it
Hypothesis: The match engine is ACTUALLY complex and taking into accounts live micro-computations that use all attributes and tactic tweaks, even if attribute testing shows that many attributes have no impact on results.
How does that hypothesis still fits the concept of meta tactic and meta attributes:
Let's say that you want to play positionnal play, based on many passes in tight spaces. If the match engine is made in such a way that no player can find a real good space, you will never achieve good positional play / tiki-taka style. Even a player with 20 Off-the-ball and 1 Pace, could find the best "pocket of space" ACCORDING to the ME, if the ME does not suggest the best spot, his "amazing skill" won't help at all.
On the other hand, if the player is the opposite (20 Pace, 1 Off the Ball), even if he doesn't find a good pocket of space, he'll be the first on the ball anyways. So as a "result", fast players would be better even in "positional play" possession tactics. Not because he finds better pockets, but because the ME can't find good pockets for any of both those players... so the fast one, at least, can still be useful.
Another example: If a player receives the ball and has 3 options. 1 good, 2 bad. Let's say the Engine looks at "decision" for a fast computation of the next play. If the ME is made so a 1-Decision player has 80% chances of taking the right decision while a 20-Decision player has a 95% chance, then, very clearly, the "Decision" attribute ha little-to-no value. Not because it is not used in the computation, but because it does not have a realistic impact ingame.
Many have hypothesised that a player with high Pace makes better decisions (leaning toward the idea that the match engine is actually over-simplified with a "general" rating for each actions). But just looking at the results does not necessarily points toward that. A fast player might provide better results because the ME is not good at providing intelligent options anyways, so the fast player will just be generally way off his marker when he makes his move.
Where my hypothesis falls short
Vision, decision, off the ball, positionning are attributes that are very tough to understand. If the ME forces players to find some specific pockets of space depending on their roles/attributes and those pockets are not the "smartest ones" (because you know... football IS complicated), a player with high vision might still do weird stuff ingame. Just because the ME might think it's good.
But when I see someone with 3 Long shots pull a top-corner screamer in, then I feel a little bit weirded out. So this is why I'm still very much wondering how the ME actually works, and how micro-computations within a game are actually made.
How are attributes testing working with different tactics
If attributes testing are done with a "meta tactic" (which is wide/fast/gegenpress-heavy), then maybe we are actually skewing our perception of attributes importance toward physical a lot. Let's say we look at a hyper-conservative possession-heavy tactic, how would attributes scale up? I would guess that whatever the attributes, those tactics will score lower. This has been thoroughly seen. But would Pace/Accel/Dribb still be that high up? Would Vision/Off-the-Ball show their head slightly above the water of "no impact"?
That might provide insight on how the ME actually works.
******************************
I conclude this by saying that I see many amazing analystst and data crunchers in these threads. My question really leads to this:
- As a FM tactician who would LOVE to create club DNAs and "chose" a style, adjusting games and using the complexity of our favorite sport, can the ME actually evolve the right way or is it too simplified and there is no hope in actually finding nuances in performance and style of individuals outside of meta attributes?
If my hypothesis is right, then the problem is simply that the ME is not yet providing players with "smarts", or giving enough weight of some attributes (IMO, 1 vision means the player looks at his feet 99% of the time, maybe the ME just says: let's drop his options by 5%), or just not attributing "success probability" to actions the right way, making a 20 decision player still making decisions that won't help the team. Meaning that possession tactics don't work because smart players don't lead to smart plays. But fast players lead to easy plays.
If my hypothesis is wrong, then yeah... the ME won't realistically evolve anytime soon and we'll have to keep playing pace monsters to reproduce possession-heavy tactics that can't really work anyways.
P.S. Training is not as advertised, that much seems certain. But my questionning here is solely on the ME. Training is it's own subject.
Whendiscussing this, we need to take the following into account:
Players with high Pace and Acceleration make perfect through balls (even though their passing and decisions are low). They finish brilliantly in one on ones with the keeper (even though their finishing is low). They defend extremely well (even though their marking and tackling are low). This has been the case since CM 03/04 (at least as far as I know. I never played the earlier CM games, but it's probably the same in those as well).
You notice this even more clearly when watching matches on comprehensive mode. A fast player doesn’t just run quickly, he plays way above his potential and even scores goals.
This pushes me more toward the second option (illusion of complexity).
None of us have any concrete evidence, we can only make guesses. But I also think this, if it were the first option, this issue wouldn’t have stayed the same for 20 years despite all the resources and the team of coders they have. It’s not that SI doesn’t know how to fix it, fixing it would require a new match engine, and they either won’t or can’t do it.
By the way, a bug from the FM 2005 era just came to my mind and I wanted to mention it because i think it's related to the topic.
In FM 2005, when the match kicked off, if you clicked on the “overview” panel and checked the fixture, you could actually see the final score of the match before it ended. And when the match finished, it always ended with exactly the score shown there. This shows that the score is determined the moment the match kicks off.
They fixed it quickly with a patch, but since Steam wasn’t really popular back then, updates didn’t reach everyone instantly like they do today. So I remember playing with that bug for quite a long time.
I’m not even sure much has changed since then. Not being sure is the scary part. We should be sure about this after 20 years, but SI just doesn't work that way.
jesus christ man, SI should’ve gone bankrupt decades ago. They’ve been putting us on this shitshow for 20 years. Thankfully, with FM 26 this whole show is finally over, and at this point waiting for moves from competitor studios makes more sense than waiting for FM 27 tbh.
Sanfierro said: By the way, a bug from the FM 2005 era just came to my mind and I wanted to mention it because i think it's related to the topic.
In FM 2005, when the match kicked off, if you clicked on the “overview” panel and checked the fixture, you could actually see the final score of the match before it ended. And when the match finished, it always ended with exactly the score shown there. This shows that the score is determined the moment the match kicks off.
They fixed it quickly with a patch, but since Steam wasn’t really popular back then, updates didn’t reach everyone instantly like they do today. So I remember playing with that bug for quite a long time.
I’m not even sure much has changed since then. Not being sure is the scary part. We should be sure about this after 20 years, but SI just doesn't work that way.
jesus christ man, SI should’ve gone bankrupt decades ago. They’ve been putting us on this shitshow for 20 years. Thankfully, with FM 26 this whole show is finally over, and at this point waiting for moves from competitor studios makes more sense than waiting for FM 27 tbh.
Results are not determined at the start of the match. That is pretty easy to check, you can simply start with a very bad tactic, then change to a very good one after 5 or 10 minutes, and result will be consistent with the good tactic.
ZaZ said: Results are not determined at the start of the match. That is pretty easy to check, you can simply start with a very bad tactic, then change to a very good one after 5 or 10 minutes, and result will be consistent with the good tactic.
There has been a theory going around for a while now that basically the result is predetermined when you enter the match and that highlights are created to suit the final result and not the other way around. However any changes made during the match (tactical, subs...) will reset the calculated end result and thus highlights. I think it has been largely confirmed by SI staff at some point but that was a while ago and there may have been changes since.
So I guess that the "precomputed" result (if this is the case, I know nothing is yet confirmed) is not only the result, but also pass maps / positions / etc. Because ingame highlights still "kinda" feel like they fit with the tactic (giving the potentially better illusion of micro-computations). Thanks for all your inputs, this is all very interesting