Normal result is ~5th-7th, so perhaps a hint of a benefit to crossing in there due to the 1st which is not unprecedented but quite rare. Not quite enough to conclude crossing has any benefit though. I think a few more samples with high consistency & technique would settle this one way or the other.
No revelations here. Again, perhaps a very marginal benefit outweighed by CA cost, similar to strength or vision.
I forgot to change Man City's media expectation to 20th, hence the high rate of sackings. Doesn't invalidate the results, just means some samples are semi-wasted (9th sacked means they wont end up first, but they could have ended up 4th say).
I am also doing this in FM24, not FM26, but I doubt that that is why he finds crossing matters and whatnot.
LightningFlik said: How do you feel about the premise that different positions require different attributes? I know that flies in the face of your theory of a generalised make-up (last I checked, anyway). I thought for a while that the positional requirements of attributes was overblown because FM always punishes me for expecting too much of it.
Now I'm unsure, but I've had great success promoting Warrington up the leagues using your old, old weights. Expand Before I played around with tactics, I would have made this critique of his findings, citing my attribute brute force testing, but now I am more open to the idea again of there being positional differences.
My impression now is that what is going on is that the Knap tactic favors a certain set of attributes, and this would extend to all the very similar top tactics in general (different roles or even formations, but all high pressing/intensity).
So if you play a defensive tactic, maybe then positioning would matter. Thing is, all defensive tactics lose. So effectively positioning never matters.
So it can be true at the same time that (1) only certain attributes matter for winning games, and (2) different tactics need different attributes.
Now I have tried positional variation in my attribute tests with the Knap tactic, and found there isn't any I can see (this is after being convinced for the longest time that concentration mattered for defenders but not forwards), but it's possible I am mistaken on one or two, or perhaps I didn't try the right combo (i.e. maybe I tested flair alone on ST, but not flair alone on MC).
I'll be doing some tests today based on Mustermann's findings, so we'll see if there are in fact position differences that actually increase/decrease win rate.
LightningFlik said: I just can't find where that code above is executed. I've decided to search for "weig" in hopes that I can find somewhere else in the code where that logic is performed. In doing so, I've found the nuttiest string references in the binary. What the actual hell is in this game?
Expand That's funny
I wonder what they've included that brought that stuff into it. I noticed there was some chinese characters, that links to a chinese website in the code for free template resources and whatnot - must be that they used an asset from it.
How did you get it to display the phrases as single lines like that? Maybe its cause I didn't let it fully finish processing I guess.
Maggh said: Thank you for all these discoveries, small question that i have, do staff have any relevant role in player development, i tend to try and have the best staff there is, but with all you all discovered im not sure if it really matters or not. Assitant managers, scouts, phisios, all that, does it really matter? Thank you Expand I would surmise overall that most staff do matter moderately, both for growth and performance. It can also matter for players being willing to sign for your club.
People are going to have different impressions, but for me I'd say staff are about 70% as impactful as I'd intuitively think they should be.
Gerrard said: Expand I came here today with the intention of making an assessment of this video lol
Very interesting findings
Flair is strongly negatively correlated with possession won for midfielders. Is this because of the attribute itself altering the role, or the high flair causing a low-possession-won role to be selected? And then, does more possession won lead to more or less wins?
I think overall this can actually be triangulated with the key attribute template and EF 424 IF HP V2 P101 AC tactic to determine what tactical measures are beneficial and which are detrimental, and then refine the key attributes - to per position specificity - based on that. Perhaps it might even clue us into how the tactic can be edged out too.
We see that flair and off-the-ball have good & bad results for different positions/measures. And he finds bravery & one-on-ones for GK were significant in relation to goals conceded. So I will do some testing to see.
There are some caveats to his conclusions I think most will miss. For instance, the GK goals conceded thing, isn't actually measuring team goals conceded, and finishing produces more goals on wingers, but that doesn't mean the team scores more goals as a whole (my own testing on finishing shows that it simply take off goals from ST and give it to the winger).
But there's a lot of valid and potentially fresh information in it.
LightningFlik said: Yes that's right. Everything in constants.h is right, that's from my app which scouts players in memory. I'm saying the names I gave to the attribute groups were made up by me.
Also I may have found the location in code where role/position suitability is calculated, but I can't get the code to fire in the course of the game, so testing it is proving difficult. It looks like that's what it's doing though. Expand So just to clarify, it is the case that 0x239 = acceleration, right?
What led you to 0x0144b40180 anyway? I've been looking through instances of 0x239 for example for something interesting, but maybe there's a better way of going about it
Would be interested in seeing the suspected role location
LightningFlik said: I'm sorry to say that all the names in my post were given by me. There's nothing in the code I shared that names attributes or the groups they belong to (although you could infer them by reading the exact scout report detail they lead to). Expand Hm.. isn't it the case though that 0x239 = acceleration?
Based on something my LLM said, it did occur to me that perhaps 0x239 just represents a generic offset (i.e. read ahead 0x239 amount starting at param1), but I saw interactions between attributes in lines of the code that made sense it was referring to specific attributes (i.e. I saw 0x244 (Aggression) with 0x240 (Dirtiness)).
I know your #defines list isn't the code itself, but you've nonetheless identified 0x239 effectively means acceleration, no?
I've been disassembling the "fm.exe" application for FM 24, which means I've turned the machine code back in to something resembling C so that I can look at it. I'm doing this because I want to find where shortlists/search results are stored in RAM so I can read them in my scouting tool but I might have found something else.
There's a function beginning at memory address 0x0144b40180 (at least in my Windows, Steam copy of FM 24.4.2+2081827), which appears to be grouping together player attributes and scoring them. I'll share the approximated C code here (feel free to ask Chat GPT what it means) as well as what I think it's doing. If you do feed this in to an LLM to decipher, give it this file too. That will map the offset codes (e.g. 0x23d, Pace) to the player's actual attributes.
From what I can see, this code is reading player attributes in groups, multiplying them together and creating a score for each group. It then returns the highest scoring group and the group itself.
For example, starting on line 65 of the code snippet:
- The lines beginning sVar1 = (short)((uint)(*(char *) read a different attribute from the Player entity (Acceleration and Pace in this example (0x239 and 0x23d)). - The lines uVar8 = (sVar1 >> 1) - (sVar1 >> 0xf); divide the attribute by 2, rounding down. - The next two if conditions ensure the number is between 1-20 inclusive. - The final if condition sums uVar11 (which is Pace / 2) and uVar6 (which is Acceleration / 2) and then multiplies the value by 5. If that result is > uVar2 (which is the highest rating it has seen so far), then it sets that to be the highest rating (*param_2 = uVar2) and sets the return value of the function to be the category ID (9 in this case).
Now here's the exciting part: the attributes it groups together and how it scores them.
Now I'm not saying that the game uses the same aggregation/scaling in the match engine--the only place that I've found it make a difference is the coach report (see below; this is me hacking the return value from the function so it thinks groups 9, 8 and 3 are this player's best)--but it's worth bearing in mind. I'm going to keep this line of enquiry up and I'll report back anything I learn.
Expand Although what you discovered isn't exactly the golden ticket we are looking for, it's nonetheless a significant insight for me.
Perhaps it's something obvious to a programmer, but as someone who doesn't know programming and couldn't make heads or tails of the code before, finding out what the attributes are called in the code gives me a solid foothold to work with.
I spent a while trying to find attribute combos in the code using this info. I came across a few intriguing ones, such as Reflexes interacting with CommandOfArea for some reason, but no clear formula I'm looking for. Strangely 'versatility' seems to interact with certain other attributes, but maybe I'm just misinterpreting what the code is about. Even without the formula, these attribute associations are perhaps useful clues for forming hypotheses to test about attributes.
Another thing I noticed is that there are references to four-letter names such as 'TrDp' (made-up example), after translating it into plain english with AI. If anyone else has a go at it, I know from previously trying to find the newgen mechanics in the code that these four-letter names are the shorthand they use for the actual mechanic. So a real example of this for instance is you can change the fitness heart icon back to fitness % by changing the mechanic it reads from 'PcOI' to 'PCRF' in the skin. If you go through the other game files, you can find a bunch of these four letter names, even ones that are meant to be hidden mechanics.
So I think I already found an example of say TrDp = anticipation + work rate, where if TrDp was what you were interested in finding out the formula for, you'd have your answer.
I reckon that the attribute combinations, and other things like newgen mechanics, are lurking in there waiting to be found.
GerardFM24 said: Interesting topic that I will keep following. Is there also such research about staff and staffroles, Like what are the most important roles and which attribute are the best. Expand EBFM has important findings about various staff. This video shows the impact of coach attributes, which I would surmise as somewhat impactful but probably a fair bit less important than you would think.
From memory he also found that you only need 1 physio to get the full benefit, and it doesn't matter if this 1 physio has '1' physiotherapy. Hopefully I'm recalling that correctly. I think the idea is that a physio's attributes only estimate length of injury better, they can't actually reduce injury rate/length aside from their presence as a staff member.
For newgen impact, staff are mostly irrelevant. The only effect is that HoYD will influence 1-2 players coming in each intake with their personality. This possibly extends to other 'youth coaches' who supposedly have half this effect.
IMO an aspect that is emerging as important is stuff to do with tactical & team selection, because these appear to significantly influence winning - even if it's just selecting between similar players you have in your squad. So tactical knowledge, judging player ability, motivating for whoever you set the relevant responsibility for. One thing this would mean in practice is getting an assistant manager who is a great manager, rather than great coach.
LightningFlik said: Are there CA-hungry attributes that people feel don't really contribute to actual results (e.g. dual-footedness)?
I have a theory that a player's CA is a multiplier for all other attributes when it comes to determining their overall strength. So given two players whose only difference is CA + whatever inconsequential attribute bumps it takes to justify the increased CA, the player with the higher CA will perform better.
An entire team of such players should outperform the same team without boosted CA. That is, a team of 99 CA players should be consistently outperformed by a team of 100 CA players even if their attributes are almost 100% identical.
If I'm wrong then bumping a useless attribute to the point where it increases CA by 1 should make no difference. Expand Nah, I don't think this is the case because decisions has mid to high CA weight for all positions, but HarvestGreen found it has no positive effect.
Can say the same for almost all of the technical attributes.
We also see a player like Maeda, who has relatively low CA, finishes near the very top in a real performance test.
It stood out to me that in the Championship Manager era, CA did come into account, where certain attribute combos were not assessed until CA was 120+. It's possible something similar might have been inherited from CM into FM.
LightningFlik said: I was digging around in the simatch.fmf file and found the attributes which are used to determine how often a player succeeds at a given task. I don't know if this is useful to anybody but it does present the opportunity for a fun experiment where you mod the file to replace everything with something like "Flair" and then give yourself a bunch of Flair: 20 players to see if that makes you incredibly OP. (If not, then it probably means this data can't be trusted anyway).
Come to think of it, it's probably a good time to have a think about this again.
What the data seems to show is that even if we assume these combos are valid (which I think they probably are), they don't necessarily result in wins/losses from having them.
I suspect certain combos are a lot more consequential than others according to what tactic you use. Since only certain tactics actually win, that narrows down the attribute combos that matter, and consequently the attributes that matter.
For me, the most solid starting point are the attributes that matter. From that we can rule out some of those combos leaving us with:
| Fouls made | Dirtiness, Aggression | | Dribbles | Dribbling, Flair, Pace, Acceleration | | Offsides | Movement, Work Rate, Acceleration, Concentration | | Distance run | Work Rate, Stamina, Pace | | Sprints | Work Rate, Stamina | | Pressures attempted | Work Rate, Anticipation | | Pressure completion | Work Rate, Anticipation, Acceleration |
I've left dribbles & offsides in despite it having flair/movement listed, as perhaps the other attributes are doing the heavy lifting.
I have theories on why Knap's 424 tactic is the best, but what we can say for sure in general is that high intensity & pressing is essential to winning.
I guess though it would be interesting to see if positioning helps a team that is so inadequate that not even the Knap tactic can save it. For strong teams, they often get 0 goals against them anyway, but for a newly promoted team, perhaps the interceptions/blocks/clearances/headers from the high positioning would achieve more draws than the Knap tactic does wins. I think it would have to be a team worse than Luton.
When I was in the early stages of constructing my tactic, I found that things can get pretty wacky:
This isn't of much practical use to know, because the limits of tactics have been largely explored, but what it means is that you aren't really constrained by the quality of your players.. the tactic can make Luton score 4-6 goals week in, week out - even against Man City, and when losing overall.
And of course we see with the Knap tactic that it can make a bunch of duds win the Premier League, so it seems to me more like the tactic is pulling something out of a hat rather than squeezing the extra 10% out of your players.
In the opposite direction, a certain tactic results in Luton losing ~24-0 every match.
Edit: I guess one potential use is FM clownmaxxing?
Would be fun to play with 9-8 results and still end up finishing in the top half.
I would like to say I got 1st with Luton using my own tactic, but actually I used the Knap EF 424 IF HP V2 P101 AC tactic (with 'Blue' set piece routines). However I did use it combination with team selection based on my FMSS weights and other little things that help such as having a lot of pre-season easy friendlies.
There were no transfers or anything like that, and everything was hands-off after day 1, so I consider this a default run, but obviously its different to a standard plug-and-play test.
The best result for a standard plug-and-play test, with Blue set pieces, has been 2nd with again Knap EF 424 IF HP V2 P101 AC.
I haven't been able to replicate 1st since, so it's a bit of a fluke. I've come up with 2 strong variations of the Knap tactic of my own, I'll call them A1 and B1 here, and B1 was awfully close to achieving 1st:
I haven't been keeping the neatest data, but here's some of what I've got:
Luton A1 - 5th, 7th, 10th, 3rd, 7th, 6th, 7th, 10th = 6.875 position Sterling meta team fixed A1 - 6th, 7th, 7th, 7th, 13th, 12th = 8.666 position Man City A1 - 1st 107pts, 1st 99pts, 1st 86pts = 1st position
Luton Knap - 2nd, 7th, 7th, 4th, 8th, 5th, 4th = 5.286 position Sterling meta team fixed Knap = 6th, 7th, 7th, 7th, 6th, 6th, 6th, 6th = 6.375 position Man City Knap - Going by the tactic name, 101 pts? Luton Knap - (optimized team) - 1st, 4th, 3rd, 5th, 3rd, 4th, 3rd, 4th, 3rd, 3rd = 3.3 position (note: samples beyond first three had a role adjustment or two)
Luton B1 - 11th, 5th, 8th, 6th = 7.5 position Sterling meta team fixed B1 - 6th = 6th position Man City B1 - 1st 98pts, 1st 95pts = 1st position Luton B1 (optimized team) - 2nd, 4th, 4th, 6th, 8th, 2nd = 4.333 position
So as you see, I haven't yet been able to top Knap's tactic. Observe also how the result becomes a lot more consistent when the team is pre-selected. The 'randomness' factor seems to only account for ~1-2 positions difference in the vast majority of cases; most of what is considered 'random' seems to actually just be different player selection. And you can't bring up injuries or whatnot, because that 1-2 position difference is inclusive of injuries, morale, and so on.
Here are some results of other tactics (Luton default):
PERFECT STRANGERS 42112 WIWBA P107 EF - 12th, 5th, 4th, 11th = 8th position HIGHWAY STAR 4231 P107 FC - 4th, 10th, 8th = 7.333 position FM24.4HGF4231 V3X RM P107 ALL CUPS - 3rd, 6th = 4.5 position (1st 94pts for Man City) FM24.4KASHMIRKnap 4213 RPM RM P114 ECCC - 5th FM24.4EFKnap 4231 A MT MC P112 FA - 8th FM24.4OXFORDKnap 4141 RM P110ALLCUPS Luton - 9th FM24.4MADASAHATTERKnap 4132 PF2 RM P105FACC - 7th EF 424 MC P101 EC FA CC - 6th 424 Mountain King - 4th SPEARHEAD 4132 - 6th
I wouldn't ordinarily say you can take anything from a single sample, but I think what we can say at least is that it doesn't seem like there is an overlooked tactic out there that actually makes Luton miraculously finish 1st consistently.
HGF 4231 V3X RM P107 ALL CUPS is a promising candidate though that could do with more testing.
Some observations from my testing:
- A single adjustment to a top tactic, whether it's attacking > positive, or setting 'work ball into box', will most likely ruin it. Usually the effects aren't even intuitive. Don't do it lads. - Adjustments to roles can be more forgiving or even work, but most of the time you're going to botch it - High intensity/pressing seems strictly essential. There are some cases however where 'positive' instead of 'attacking' say, and not going hard in tackles for certain positions, is viable or is even best for that particular formation. For Knap's best tactic though, 'attacking' and hard tackling on all players is simply what works best. - Alternative formations, even ridiculous ones, can work surprisingly well. The main aspect to consider seems to be position proficiency - you should choose the formation which is inclusive of your best player's positions. Certain formations such as 424 do seem to have an inherent edge though aside from position proficiency. I have speculative theories on as to why, but I can surmise it as 'ideal angles of attack'. - It does seem to me that 'roles' are simply half-blinded judgements for your staff to use when assessing players. To me this is reinforced by the 'pick roles & duties' dropdown option under 'quick pick', which actually changes the roles & duties in your tactic to what your assistant thinks is best. Attributes have nothing to do with roles; every outfield player performs according to the exact same attribute formula regardless of position let alone role. Yet changing the role does have an impact on results. How to square this fact? The assistant is essentially choosing between jumbled up lists, and what is probably happening is that some lists (i.e. AF ST) are less jumbled up than others (i.e. False Nine ST). If you preselect the players, it reduces the result variation a lot. And also if you then change the role, it does not seem to change the result much. Yet this is not always the case. Possible explanation: Some roles disallow or enforce certain player instructions, and even those that don't, may in theory change the positioning of the player.
LightningFlik said: @GeorgeFloydOverdosed As part of your season-long experiments, would it be valuable to see at a glance all teams in a given league with the average rating of their players?
I'm thinking something like this:
| Man City | 84.75% | | Liverpool | 83.60% | | Arsenal | 82.13% | | Chelsea | 80.44% |
And so on.
I'm not sure what your tests look like but I assume you'd expect the final order of computer-controlled teams in your leagues to roughly correlate to the ratings of the players inside? or else it would mean a tweak is needed to the ratings? Right now it seems like you're testing tactics but this way you could test your weights with all the teams in the league, not just one. Expand Yes, I've been missing that feature a bit since every league became 'jupiler league' in Genie Scout.
I did a test before of the Premier League with 4 teams replaced with 80%, 78%, 76%, 74% players. I only did a few samples, but they all finished in the correct order at season's end.
Wail said: Can you do it automatically through the season ? or we have to do it week by week, its kinda boring copy pasting it. Expand You add it for every week straightaway, but I don't think there's a way around having to copy paste each week
bf3metro said: One practical thing I’m still unsure about: where would you place the 2x Physical + 2x Attacking sessions during the week?
The issue I have is that in weeks with away games, Travel can automatically replace training slots, especially around the day before matches, so sometimes the intended 214 sessions disappear. Expand I think there's an option in the settings to set matches to only be on 2 certain days? I don't use this though, so not sure.
I would do 2 days of [Physical][Attack]
I wish I could remember the details properly, but I can vaguely recall that where you place training doesn't matter like you would think it does. That you can put training straight after a match day, and it won't affect condition recovery rate or injury rate. And that it's actually calculated as a whole 2 week (not even 1 week) block of training.
Not 100% certain on the injury thing, because I would think that unrecovered condition would increase injury rate of training modules, but then again match sharpness (which decreases with time post-match) has a bigger influence on injury rate, so perhaps they cancel each other out.
I won't get into the details because it'll just be confusing, but the Sterling meta team vs. real Luton discrepancy should be fixed now, and you can discard the theory I came up with to try and explain the discrepancy (that roles do take attributes into account in some way).
But this leads me to new ideas.
We see that when position proficiency is equalized, a non-sensical assymetric tactic that I came up with performed equal to Knap's best tactic (5th vs 6th).
We also see that the random variation is subdued a lot with the meta team due it being composed of identical players. Consistent ~6th-7th with Knap tactic, compared to ~2nd-10th for real Luton. And it's unlikely to be mostly on account of injuries, morale, etc. because none of those are controlled with the meta team tests.
A theory I have is that when the assistant manager picks players to play at real Luton, he is choosing based on role rating. Roles are an average(?) of select highlighted attributes. Some of these roles are going to highlight more meta attributes than others.
Now let's take a look at some ST options:
trequartista ST - drib*, fin, ant, com*, vis*, acc*, agil, bal false nine ST - drib*, fin, ant, com*, vis*, acc*, agil*, bal advanced forward ST - drib*, fin*, ant, com*, work, acc*, agil, bal, pace, sta pressing forward ST - agg*, ant*, com, con, work*, acc*, agil, bal, pace*, sta*, str
* denotes its shown as a key attribute for the role.
I've highlighted attributes that contribute 25 or more weight, inferred from the real player testing I've done.
Guess which two of the four Knap's tactic employs.
Now if the assistant manager is selecting based on the role selected in your tactic, then we can suppose if I use AF STs, I will get ~2nd, but if I use False Nine STs maybe I will get ~16th. Not variations between 2nd and 10th. Consider however that because the assistant manager is effectively using a half-blinded assessment, a layer of inconsistency is being added. So for instance, your tactic has 2x AF ST, and your assistant picks 2xST that has very high acc & drib because they have the highest AF ST role score. However, you also have a False Nine ST that has very high det, con, vis, pace, work, and jump that your assistant overlooked, but may nonetheless get rotated in half the time as they are the close 2nd option.
I'll be doing some tests to see if this theory works at all or not.
First result for my tactic with default Man City w/ Blue routines, no player selection or anything. I do set the training intensity to 0-0-0-2x-2x though.
I'm thinking I'll run about 10 tests of Luton & Man City each to see how it goes against the Knap tactic.
I've done 2 tests so far of HIGHWAY STAR 4231 P107 FC, which I've seen Cardoni recommend:
Luton default = 4th, 10th
Another thing I'm thinking of doing is having a 3rd crack at player traits. We know that the tactic is very important, and that player instructions are part of this.. so maybe if we have a team given player traits that align perfectly with the Knap tactic, it will do better - even if its because its giving initial familiarity when normally there is zero starting familiarity. So we'll see if anything results from this approach.
LightningFlik said: With that in mind, how flexible a weighting system do you think is needed to really assess players? Is a simple Pace: 20, Acceleration: 19.2 system sufficient or do you need interdependence between attributes? Expand There isn't the data yet to confirm one way or the other, but my hunch is that there aren't many attribute interactions going on, simply because the evidence shows almost all the technicals don't matter no matter what.
However I believe there is an interaction between aggression and dirtiness for instance that is crucial. Not just because it's a hypothesis that sounds sensible, but because as I was playing around with the weightings, it just seems to make things work out better.
So it would be good if you could include the ability to create complex formulas as FMSS does, but I don't know how difficult that would be to implement.
fb = tackling (cross success %)
fb = crossing, off the ball (high intensity sprints)
dm = low flair (posession won)
dm = off the ball, low marking (pressures attempted)
winger = low decisions, low off the ball (dribbles)
winger = crossing, teamwork, low heading (pressures completed)
st = off the ball, low technique (xG overperformance)
'Low' attributes were reduced to 1, while the others were increased to 20. CA was generally higher, with the highest gain being +20 CA on fullbacks.
Results: 7th, 7th
So no performance difference
I then tried just crossing 16 on fullbacks & wingers.
Results: 9th (sacked), 1st, 3rd, 7th, 6th (sacked), 8th (sacked) = 5.666 average
Normal result is ~5th-7th, so perhaps a hint of a benefit to crossing in there due to the 1st which is not unprecedented but quite rare. Not quite enough to conclude crossing has any benefit though. I think a few more samples with high consistency & technique would settle this one way or the other.
Next, bravery 20 for DCs.
Results: 6th (sacked), 3rd, 5th, 9th (sacked), 5th = 5.6 position
No revelations here. Again, perhaps a very marginal benefit outweighed by CA cost, similar to strength or vision.
I forgot to change Man City's media expectation to 20th, hence the high rate of sackings. Doesn't invalidate the results, just means some samples are semi-wasted (9th sacked means they wont end up first, but they could have ended up 4th say).
I am also doing this in FM24, not FM26, but I doubt that that is why he finds crossing matters and whatnot.
Now I'm unsure, but I've had great success promoting Warrington up the leagues using your old, old weights.
Before I played around with tactics, I would have made this critique of his findings, citing my attribute brute force testing, but now I am more open to the idea again of there being positional differences.
My impression now is that what is going on is that the Knap tactic favors a certain set of attributes, and this would extend to all the very similar top tactics in general (different roles or even formations, but all high pressing/intensity).
So if you play a defensive tactic, maybe then positioning would matter. Thing is, all defensive tactics lose. So effectively positioning never matters.
So it can be true at the same time that (1) only certain attributes matter for winning games, and (2) different tactics need different attributes.
Now I have tried positional variation in my attribute tests with the Knap tactic, and found there isn't any I can see (this is after being convinced for the longest time that concentration mattered for defenders but not forwards), but it's possible I am mistaken on one or two, or perhaps I didn't try the right combo (i.e. maybe I tested flair alone on ST, but not flair alone on MC).
I'll be doing some tests today based on Mustermann's findings, so we'll see if there are in fact position differences that actually increase/decrease win rate.
LightningFlik said: I just can't find where that code above is executed. I've decided to search for "weig" in hopes that I can find somewhere else in the code where that logic is performed. In doing so, I've found the nuttiest string references in the binary. What the actual hell is in this game?
That's funny
I wonder what they've included that brought that stuff into it. I noticed there was some chinese characters, that links to a chinese website in the code for free template resources and whatnot - must be that they used an asset from it.
How did you get it to display the phrases as single lines like that? Maybe its cause I didn't let it fully finish processing I guess.
I would surmise overall that most staff do matter moderately, both for growth and performance. It can also matter for players being willing to sign for your club.
People are going to have different impressions, but for me I'd say staff are about 70% as impactful as I'd intuitively think they should be.
I came here today with the intention of making an assessment of this video lol
Very interesting findings
Flair is strongly negatively correlated with possession won for midfielders. Is this because of the attribute itself altering the role, or the high flair causing a low-possession-won role to be selected? And then, does more possession won lead to more or less wins?
I think overall this can actually be triangulated with the key attribute template and EF 424 IF HP V2 P101 AC tactic to determine what tactical measures are beneficial and which are detrimental, and then refine the key attributes - to per position specificity - based on that. Perhaps it might even clue us into how the tactic can be edged out too.
We see that flair and off-the-ball have good & bad results for different positions/measures. And he finds bravery & one-on-ones for GK were significant in relation to goals conceded. So I will do some testing to see.
There are some caveats to his conclusions I think most will miss. For instance, the GK goals conceded thing, isn't actually measuring team goals conceded, and finishing produces more goals on wingers, but that doesn't mean the team scores more goals as a whole (my own testing on finishing shows that it simply take off goals from ST and give it to the winger).
But there's a lot of valid and potentially fresh information in it.
Also I may have found the location in code where role/position suitability is calculated, but I can't get the code to fire in the course of the game, so testing it is proving difficult. It looks like that's what it's doing though.
So just to clarify, it is the case that 0x239 = acceleration, right?
What led you to 0x0144b40180 anyway?
I've been looking through instances of 0x239 for example for something interesting, but maybe there's a better way of going about it
Would be interested in seeing the suspected role location
Hm.. isn't it the case though that 0x239 = acceleration?
Based on something my LLM said, it did occur to me that perhaps 0x239 just represents a generic offset (i.e. read ahead 0x239 amount starting at param1), but I saw interactions between attributes in lines of the code that made sense it was referring to specific attributes (i.e. I saw 0x244 (Aggression) with 0x240 (Dirtiness)).
I know your #defines list isn't the code itself, but you've nonetheless identified 0x239 effectively means acceleration, no?
I've been disassembling the "fm.exe" application for FM 24, which means I've turned the machine code back in to something resembling C so that I can look at it. I'm doing this because I want to find where shortlists/search results are stored in RAM so I can read them in my scouting tool but I might have found something else.
There's a function beginning at memory address 0x0144b40180 (at least in my Windows, Steam copy of FM 24.4.2+2081827), which appears to be grouping together player attributes and scoring them. I'll share the approximated C code here (feel free to ask Chat GPT what it means) as well as what I think it's doing. If you do feed this in to an LLM to decipher, give it this file too. That will map the offset codes (e.g. 0x23d, Pace) to the player's actual attributes.
From what I can see, this code is reading player attributes in groups, multiplying them together and creating a score for each group. It then returns the highest scoring group and the group itself.
For example, starting on line 65 of the code snippet:
Spoiler sVar1 = (short)((uint)(*(char *)(param_1 + 0x239) * 0x6667 + 0xccce) >> 0x10);
uVar8 = (sVar1 >> 1) - (sVar1 >> 0xf);
if ((short)uVar8 < 2) {
uVar8 = 1;
}
uVar6 = (uint)uVar8;
if (0x13 < uVar6) {
uVar6 = 0x14;
}
sVar1 = (short)((uint)(*(char *)(param_1 + 0x23d) * 0x6667 + 0xccce) >> 0x10);
uVar8 = (sVar1 >> 1) - (sVar1 >> 0xf);
if ((short)uVar8 < 2) {
uVar8 = 1;
}
uVar11 = 0x14;
if (uVar8 < 0x14) {
uVar11 = (uint)uVar8;
}
if ((int)(short)uVar2 < (int)((uVar11 + uVar6) * 5)) {
uVar2 = (short)(uVar11 + uVar6) * 5;
*param_2 = uVar2;
uVar3 = CONCAT71((int7)((ulonglong)uVar3 >> 8),9);
}
- The lines beginning sVar1 = (short)((uint)(*(char *) read a different attribute from the Player entity (Acceleration and Pace in this example (0x239 and 0x23d)).
- The lines uVar8 = (sVar1 >> 1) - (sVar1 >> 0xf); divide the attribute by 2, rounding down.
- The next two if conditions ensure the number is between 1-20 inclusive.
- The final if condition sums uVar11 (which is Pace / 2) and uVar6 (which is Acceleration / 2) and then multiplies the value by 5. If that result is > uVar2 (which is the highest rating it has seen so far), then it sets that to be the highest rating (*param_2 = uVar2) and sets the return value of the function to be the category ID (9 in this case).
Now here's the exciting part: the attributes it groups together and how it scores them.
Group 1: Set pieces
((Corners + Free Kicks + Penalties + Long Throws) * 10) / 4
Group 2: Attacking technique
(Crossing + Finishing + Long Shots + Passing + Technique) * 2
Group 3: Attacking intelligence
((Anticipation + Vision + Decisions + Positioning) * 10) / 4
(Yes, positioning.)
Group 4: Leadership
((Bravery + Work Rate + Teamwork) * 10) / 3
Group 5: Defensive intelligence
((Tackling + Marking + Concentration + Positioning) * 10) / 4
Group 6: Technical control
(Flair + Dribbling + First Touch + Composure + Balance) * 2
Group 7: Aerial Ability
(Heading + Jumping Reach) * 5
Group 8: Physique
((Strength + Natural Fitness + Stamina + Agility) * 10) / 4
Group 9: Speed
(Acceleration + Pace) * 5
Now I'm not saying that the game uses the same aggregation/scaling in the match engine--the only place that I've found it make a difference is the coach report (see below; this is me hacking the return value from the function so it thinks groups 9, 8 and 3 are this player's best)--but it's worth bearing in mind. I'm going to keep this line of enquiry up and I'll report back anything I learn.
Although what you discovered isn't exactly the golden ticket we are looking for, it's nonetheless a significant insight for me.
Perhaps it's something obvious to a programmer, but as someone who doesn't know programming and couldn't make heads or tails of the code before, finding out what the attributes are called in the code gives me a solid foothold to work with.
I spent a while trying to find attribute combos in the code using this info. I came across a few intriguing ones, such as Reflexes interacting with CommandOfArea for some reason, but no clear formula I'm looking for. Strangely 'versatility' seems to interact with certain other attributes, but maybe I'm just misinterpreting what the code is about. Even without the formula, these attribute associations are perhaps useful clues for forming hypotheses to test about attributes.
Another thing I noticed is that there are references to four-letter names such as 'TrDp' (made-up example), after translating it into plain english with AI. If anyone else has a go at it, I know from previously trying to find the newgen mechanics in the code that these four-letter names are the shorthand they use for the actual mechanic. So a real example of this for instance is you can change the fitness heart icon back to fitness % by changing the mechanic it reads from 'PcOI' to 'PCRF' in the skin. If you go through the other game files, you can find a bunch of these four letter names, even ones that are meant to be hidden mechanics.
So I think I already found an example of say TrDp = anticipation + work rate, where if TrDp was what you were interested in finding out the formula for, you'd have your answer.
I reckon that the attribute combinations, and other things like newgen mechanics, are lurking in there waiting to be found.
Is there also such research about staff and staffroles,
Like what are the most important roles and which attribute are the best.
EBFM has important findings about various staff. This video shows the impact of coach attributes, which I would surmise as somewhat impactful but probably a fair bit less important than you would think.
From memory he also found that you only need 1 physio to get the full benefit, and it doesn't matter if this 1 physio has '1' physiotherapy. Hopefully I'm recalling that correctly. I think the idea is that a physio's attributes only estimate length of injury better, they can't actually reduce injury rate/length aside from their presence as a staff member.
For newgen impact, staff are mostly irrelevant. The only effect is that HoYD will influence 1-2 players coming in each intake with their personality. This possibly extends to other 'youth coaches' who supposedly have half this effect.
IMO an aspect that is emerging as important is stuff to do with tactical & team selection, because these appear to significantly influence winning - even if it's just selecting between similar players you have in your squad. So tactical knowledge, judging player ability, motivating for whoever you set the relevant responsibility for. One thing this would mean in practice is getting an assistant manager who is a great manager, rather than great coach.
I have a theory that a player's CA is a multiplier for all other attributes when it comes to determining their overall strength. So given two players whose only difference is CA + whatever inconsequential attribute bumps it takes to justify the increased CA, the player with the higher CA will perform better.
An entire team of such players should outperform the same team without boosted CA. That is, a team of 99 CA players should be consistently outperformed by a team of 100 CA players even if their attributes are almost 100% identical.
If I'm wrong then bumping a useless attribute to the point where it increases CA by 1 should make no difference.
Nah, I don't think this is the case because decisions has mid to high CA weight for all positions, but HarvestGreen found it has no positive effect.
Can say the same for almost all of the technical attributes.
We also see a player like Maeda, who has relatively low CA, finishes near the very top in a real performance test.
It stood out to me that in the Championship Manager era, CA did come into account, where certain attribute combos were not assessed until CA was 120+. It's possible something similar might have been inherited from CM into FM.
LightningFlik said: I was digging around in the simatch.fmf file and found the attributes which are used to determine how often a player succeeds at a given task. I don't know if this is useful to anybody but it does present the opportunity for a fun experiment where you mod the file to replace everything with something like "Flair" and then give yourself a bunch of Flair: 20 players to see if that makes you incredibly OP. (If not, then it probably means this data can't be trusted anyway).
| Statistic | Encoded attributes |
| --- | --- |
| Passes attempted | Passing, Work Rate |
| Pass completion | Passing, Decisions |
| Key passes | Passing, Vision |
| Crosses attempted | Crossing, Work Rate |
| Cross completion | Crossing, Vision, Technique |
| Tackles attempted | Tackling, Positioning, Work Rate |
| Tackle completion | Tackling, Positioning |
| Interceptions | Anticipation, Positioning |
| Blocks | Anticipation, Positioning, Bravery |
| Clearances | Positioning |
| Headers attempted | Positioning, Jumping |
| Header completion | Heading, Strength, Jumping |
| Fouls made | Dirtiness, Aggression |
| Dribbles | Dribbling, Flair, Pace, Acceleration |
| Offsides | Movement, Work Rate, Acceleration, Concentration |
| Distance run | Work Rate, Stamina, Pace |
| Sprints | Work Rate, Stamina |
| Pressures attempted | Work Rate, Anticipation |
| Pressure completion | Work Rate, Anticipation, Acceleration |
| Progressive passes | Passing, Vision |
HarvestGreen did a followup thread to the one Yarema linked.
Come to think of it, it's probably a good time to have a think about this again.
What the data seems to show is that even if we assume these combos are valid (which I think they probably are), they don't necessarily result in wins/losses from having them.
I suspect certain combos are a lot more consequential than others according to what tactic you use. Since only certain tactics actually win, that narrows down the attribute combos that matter, and consequently the attributes that matter.
For me, the most solid starting point are the attributes that matter. From that we can rule out some of those combos leaving us with:
| Fouls made | Dirtiness, Aggression |
| Dribbles | Dribbling, Flair, Pace, Acceleration |
| Offsides | Movement, Work Rate, Acceleration, Concentration |
| Distance run | Work Rate, Stamina, Pace |
| Sprints | Work Rate, Stamina |
| Pressures attempted | Work Rate, Anticipation |
| Pressure completion | Work Rate, Anticipation, Acceleration |
I've left dribbles & offsides in despite it having flair/movement listed, as perhaps the other attributes are doing the heavy lifting.
I have theories on why Knap's 424 tactic is the best, but what we can say for sure in general is that high intensity & pressing is essential to winning.
I guess though it would be interesting to see if positioning helps a team that is so inadequate that not even the Knap tactic can save it. For strong teams, they often get 0 goals against them anyway, but for a newly promoted team, perhaps the interceptions/blocks/clearances/headers from the high positioning would achieve more draws than the Knap tactic does wins. I think it would have to be a team worse than Luton.
This isn't of much practical use to know, because the limits of tactics have been largely explored, but what it means is that you aren't really constrained by the quality of your players.. the tactic can make Luton score 4-6 goals week in, week out - even against Man City, and when losing overall.
And of course we see with the Knap tactic that it can make a bunch of duds win the Premier League, so it seems to me more like the tactic is pulling something out of a hat rather than squeezing the extra 10% out of your players.
In the opposite direction, a certain tactic results in Luton losing ~24-0 every match.
Edit: I guess one potential use is FM clownmaxxing?
Would be fun to play with 9-8 results and still end up finishing in the top half.
There were no transfers or anything like that, and everything was hands-off after day 1, so I consider this a default run, but obviously its different to a standard plug-and-play test.
The best result for a standard plug-and-play test, with Blue set pieces, has been 2nd with again Knap EF 424 IF HP V2 P101 AC.
I haven't been able to replicate 1st since, so it's a bit of a fluke. I've come up with 2 strong variations of the Knap tactic of my own, I'll call them A1 and B1 here, and B1 was awfully close to achieving 1st:
I haven't been keeping the neatest data, but here's some of what I've got:
Luton A1 - 5th, 7th, 10th, 3rd, 7th, 6th, 7th, 10th = 6.875 position
Sterling meta team fixed A1 - 6th, 7th, 7th, 7th, 13th, 12th = 8.666 position
Man City A1 - 1st 107pts, 1st 99pts, 1st 86pts = 1st position
Luton Knap - 2nd, 7th, 7th, 4th, 8th, 5th, 4th = 5.286 position
Sterling meta team fixed Knap = 6th, 7th, 7th, 7th, 6th, 6th, 6th, 6th = 6.375 position
Man City Knap - Going by the tactic name, 101 pts?
Luton Knap - (optimized team) - 1st, 4th, 3rd, 5th, 3rd, 4th, 3rd, 4th, 3rd, 3rd = 3.3 position (note: samples beyond first three had a role adjustment or two)
Luton B1 - 11th, 5th, 8th, 6th = 7.5 position
Sterling meta team fixed B1 - 6th = 6th position
Man City B1 - 1st 98pts, 1st 95pts = 1st position
Luton B1 (optimized team) - 2nd, 4th, 4th, 6th, 8th, 2nd = 4.333 position
So as you see, I haven't yet been able to top Knap's tactic. Observe also how the result becomes a lot more consistent when the team is pre-selected. The 'randomness' factor seems to only account for ~1-2 positions difference in the vast majority of cases; most of what is considered 'random' seems to actually just be different player selection. And you can't bring up injuries or whatnot, because that 1-2 position difference is inclusive of injuries, morale, and so on.
Here are some results of other tactics (Luton default):
PERFECT STRANGERS 42112 WIWBA P107 EF - 12th, 5th, 4th, 11th = 8th position
HIGHWAY STAR 4231 P107 FC - 4th, 10th, 8th = 7.333 position
FM24.4HGF4231 V3X RM P107 ALL CUPS - 3rd, 6th = 4.5 position (1st 94pts for Man City)
FM24.4KASHMIRKnap 4213 RPM RM P114 ECCC - 5th
FM24.4EFKnap 4231 A MT MC P112 FA - 8th
FM24.4OXFORDKnap 4141 RM P110ALLCUPS Luton - 9th
FM24.4MADASAHATTERKnap 4132 PF2 RM P105FACC - 7th
EF 424 MC P101 EC FA CC - 6th
424 Mountain King - 4th
SPEARHEAD 4132 - 6th
I wouldn't ordinarily say you can take anything from a single sample, but I think what we can say at least is that it doesn't seem like there is an overlooked tactic out there that actually makes Luton miraculously finish 1st consistently.
HGF 4231 V3X RM P107 ALL CUPS is a promising candidate though that could do with more testing.
Some observations from my testing:
- A single adjustment to a top tactic, whether it's attacking > positive, or setting 'work ball into box', will most likely ruin it. Usually the effects aren't even intuitive. Don't do it lads.
- Adjustments to roles can be more forgiving or even work, but most of the time you're going to botch it
- High intensity/pressing seems strictly essential. There are some cases however where 'positive' instead of 'attacking' say, and not going hard in tackles for certain positions, is viable or is even best for that particular formation. For Knap's best tactic though, 'attacking' and hard tackling on all players is simply what works best.
- Alternative formations, even ridiculous ones, can work surprisingly well. The main aspect to consider seems to be position proficiency - you should choose the formation which is inclusive of your best player's positions. Certain formations such as 424 do seem to have an inherent edge though aside from position proficiency. I have speculative theories on as to why, but I can surmise it as 'ideal angles of attack'.
- It does seem to me that 'roles' are simply half-blinded judgements for your staff to use when assessing players. To me this is reinforced by the 'pick roles & duties' dropdown option under 'quick pick', which actually changes the roles & duties in your tactic to what your assistant thinks is best. Attributes have nothing to do with roles; every outfield player performs according to the exact same attribute formula regardless of position let alone role. Yet changing the role does have an impact on results. How to square this fact? The assistant is essentially choosing between jumbled up lists, and what is probably happening is that some lists (i.e. AF ST) are less jumbled up than others (i.e. False Nine ST). If you preselect the players, it reduces the result variation a lot. And also if you then change the role, it does not seem to change the result much. Yet this is not always the case. Possible explanation: Some roles disallow or enforce certain player instructions, and even those that don't, may in theory change the positioning of the player.
I'm thinking something like this:
| Man City | 84.75% |
| Liverpool | 83.60% |
| Arsenal | 82.13% |
| Chelsea | 80.44% |
And so on.
I'm not sure what your tests look like but I assume you'd expect the final order of computer-controlled teams in your leagues to roughly correlate to the ratings of the players inside? or else it would mean a tweak is needed to the ratings? Right now it seems like you're testing tactics but this way you could test your weights with all the teams in the league, not just one.
Yes, I've been missing that feature a bit since every league became 'jupiler league' in Genie Scout.
I did a test before of the Premier League with 4 teams replaced with 80%, 78%, 76%, 74% players. I only did a few samples, but they all finished in the correct order at season's end.
You add it for every week straightaway, but I don't think there's a way around having to copy paste each week
bf3metro said: One practical thing I’m still unsure about: where would you place the 2x Physical + 2x Attacking sessions during the week?
The issue I have is that in weeks with away games, Travel can automatically replace training slots, especially around the day before matches, so sometimes the intended 214 sessions disappear.
I think there's an option in the settings to set matches to only be on 2 certain days? I don't use this though, so not sure.
I would do 2 days of [Physical][Attack]
I wish I could remember the details properly, but I can vaguely recall that where you place training doesn't matter like you would think it does. That you can put training straight after a match day, and it won't affect condition recovery rate or injury rate. And that it's actually calculated as a whole 2 week (not even 1 week) block of training.
Not 100% certain on the injury thing, because I would think that unrecovered condition would increase injury rate of training modules, but then again match sharpness (which decreases with time post-match) has a bigger influence on injury rate, so perhaps they cancel each other out.
But this leads me to new ideas.
We see that when position proficiency is equalized, a non-sensical assymetric tactic that I came up with performed equal to Knap's best tactic (5th vs 6th).
We also see that the random variation is subdued a lot with the meta team due it being composed of identical players. Consistent ~6th-7th with Knap tactic, compared to ~2nd-10th for real Luton. And it's unlikely to be mostly on account of injuries, morale, etc. because none of those are controlled with the meta team tests.
A theory I have is that when the assistant manager picks players to play at real Luton, he is choosing based on role rating. Roles are an average(?) of select highlighted attributes. Some of these roles are going to highlight more meta attributes than others.
Now let's take a look at some ST options:
trequartista ST - drib*, fin, ant, com*, vis*, acc*, agil, bal
false nine ST - drib*, fin, ant, com*, vis*, acc*, agil*, bal
advanced forward ST - drib*, fin*, ant, com*, work, acc*, agil, bal, pace, sta
pressing forward ST - agg*, ant*, com, con, work*, acc*, agil, bal, pace*, sta*, str
* denotes its shown as a key attribute for the role.
I've highlighted attributes that contribute 25 or more weight, inferred from the real player testing I've done.
Guess which two of the four Knap's tactic employs.
Now if the assistant manager is selecting based on the role selected in your tactic, then we can suppose if I use AF STs, I will get ~2nd, but if I use False Nine STs maybe I will get ~16th. Not variations between 2nd and 10th. Consider however that because the assistant manager is effectively using a half-blinded assessment, a layer of inconsistency is being added. So for instance, your tactic has 2x AF ST, and your assistant picks 2xST that has very high acc & drib because they have the highest AF ST role score. However, you also have a False Nine ST that has very high det, con, vis, pace, work, and jump that your assistant overlooked, but may nonetheless get rotated in half the time as they are the close 2nd option.
I'll be doing some tests to see if this theory works at all or not.
any news on trainings? which one came out on top?
My conclusion is here
My preferred training schedule remains 214 [Physical]x2[Attacking]x2[Quickness Focus]
331 [Physical][Match Practice][Attacking][Defending][Quickness Focus] has the best absolute attribute gain
Results: 8th, 4th, 5th, 5th, 7th = 5.8 position
Recall that default was ~6.3 I think it was, so I don't think the player traits are making any difference here.
iezzex said: Where i can find those trainings?
You just manually enter it in your training schedule
First, this is a great tactic. Still proving very difficult to beat.
I have a few questions:
1) Have you ever had a result with Luton (or the worst team) where they finished 1st?
2) What set piece routines do you recommend to use today? I see you recommended Azure before, but that was in 2024.
3) What do you think is the best overall tactic for default FM24 today? Particularly for an underdog team, if that's a necessary distinction
First result for my tactic with default Man City w/ Blue routines, no player selection or anything. I do set the training intensity to 0-0-0-2x-2x though.
I'm thinking I'll run about 10 tests of Luton & Man City each to see how it goes against the Knap tactic.
I've done 2 tests so far of HIGHWAY STAR 4231 P107 FC, which I've seen Cardoni recommend:
Luton default = 4th, 10th
Another thing I'm thinking of doing is having a 3rd crack at player traits. We know that the tactic is very important, and that player instructions are part of this.. so maybe if we have a team given player traits that align perfectly with the Knap tactic, it will do better - even if its because its giving initial familiarity when normally there is zero starting familiarity. So we'll see if anything results from this approach.
There isn't the data yet to confirm one way or the other, but my hunch is that there aren't many attribute interactions going on, simply because the evidence shows almost all the technicals don't matter no matter what.
However I believe there is an interaction between aggression and dirtiness for instance that is crucial. Not just because it's a hypothesis that sounds sensible, but because as I was playing around with the weightings, it just seems to make things work out better.
So it would be good if you could include the ability to create complex formulas as FMSS does, but I don't know how difficult that would be to implement.