Summary of recent findings for optimal play in FM24

Step 2: Each template has a "style"

The game gives every template a style by adding up three of its multipliers in each group below. The highest total wins.

StyleThe three attributes added
AttackingFinishing + Off the Ball + Vision
DefendingMarking + Tackling + Positioning
FitnessStrength + Stamina + Pace
TacticalOff the Ball + Decisions + Vision
TechnicalTechnique + First Touch + Dribbling
MentalBravery + Concentration + Work Rate
Set piecesCorners + Free Kicks + Long Throws

A tie goes to whichever style is higher in the table. But if the average of the Attacking, Defending and Fitness totals is more than 80% of the winning total, the template is "all-rounder" instead.

Example. A template with Technical total 4.2, Attacking 3.0, Defending 2.4 and Fitness 3.3. Technical wins with 4.2. The average of Attacking, Defending and Fitness is (3.0 + 2.4 + 3.3) ÷ 3 = 2.9. 80% of 4.2 is 3.36. 2.9 is not above 3.36, so it's a Technical template.

Step 3: Your HoYD's coaching style

Your HoYD gets a style the same way, from his coaching attributes. His highest attribute out of Attacking, Defending, Tactical, Technical, Mental, Fitness and one more hidden value wins. Ties go to the first one in that list. The hidden value counts as "Set pieces". I couldn't match it to anything you can see in-game.

He's an all-rounder instead if the average of all of those except Mental, rounded down, is more than 80% of his highest.

Examples.
  • Attacking 16, Defending 12, Fitness 10, Mental 14, Technical 15, Tactical 13, hidden value 8. The average without Mental is (16 + 12 + 10 + 15 + 13 + 8) ÷ 6 = 12.3, rounded down to 12. 80% of 16 is 12.8. 12 is not above 12.8, so he's an Attacking HoYD.
  • Attacking 14, Defending 14, Fitness 13, Mental 12, Technical 15, Tactical 14, hidden value 12. The average is 82 ÷ 6 = 13.7, rounded down to 13. 80% of 15 is 12. 13 is above 12, so he's an all-rounder.

With no HoYD, there's no style and no bias.

Step 4: Picking the template — region, and the HoYD's nudge

Which region.


  • If the new player's nation has a playable league, he uses his nation's templates.
  • If not, the game uses his continent's templates to set the size of the roll (below). Then it uses the roll itself as a coin flip: if the roll is an even number it walks only his continent's templates, and if it's odd it walks the whole list* (every nation and continent).
That odd-roll half looks like a bug. The roll is sized for his continent's templates alone, but the walk starts at the front of the whole list, and the list starts with the continent sets in this order: South America, Oceania, North America, Europe, Asia, Africa. So on an odd roll, he nearly always gets a template from the first continent sets in that order that have templates for his position, whatever his own continent. I confirmed this live (see "What the templates look like in a real save" below): most odd-roll Europeans got South or North American templates.

Example. A Scottish youngster, with Scotland not playable. Say his continent's templates for his position and height have a total weight of 300, so the roll is 0–299. On an even roll he gets a European template as you'd expect. On an odd roll, the walk starts with South America's templates for his position. It only reaches Europe's if the roll is larger than the weight of every template ahead of them.

If nothing gets picked (for example, his nation has no templates for his position and height), the game tries again over the whole list and sizes the roll for the whole list.

The roll. The game lines up the matching templates and rolls a number from 0 up to (total weight − 1). The total only counts templates in his region, for his positions and his height band. It then walks down the list, counting off each template's weight, and stops at the first template in his height band where the count reaches the roll.

Two things take a share from templates further down the list:
  • Templates of the other height bands use up the count as the game walks past them, even though they can't be picked and weren't in the total.
  • A template with the same style as your HoYD counts double, but the total the roll came from does not include the doubling.
Example. Three templates in this order, each with weight 10 and in his height band: A (Technical), B (Attacking), C (Defending). The total is 30, so the roll is 0–29.
  • No HoYD: A takes rolls 0–10 (11 in 30, 37%). B takes 11–20 (33%) and C takes 21–29 (30%).
  • Technical HoYD: A counts 20, so it takes rolls 0–20 (70%). B takes 21–29 (30%). C can never be picked.
  • Defending HoYD: C is last, so doubling it changes nothing: 37% / 33% / 30%.
  • No HoYD, but a 10-weight template for a different height sits between A and B: the total is still 30. A takes 0–10. The other-height template uses up 11–20 but can't be picked, so those rolls fall through to B, which takes 11–29 (63%). C can never be picked.

So a matching style gets up to double the chance, depending on where its templates sit in the list. Because the most common template for each role comes first (Step 1b, point 6), the HoYD's nudge does the most for the most common templates for each role.

When the template matches his style, the HoYD is also added to the player's list of important people. My guess is that's how he shows up as one of the youngster's favoured personnel, but I haven't confirmed it.

Step 5: Turning the template into attributes

For each attribute:
  1. Base value. The game works out a level for the player's CA: CA ÷ 20 + 6.5, rounded down. It works out a second level the same way from either his PA (1 time in 3) or a random number between his CA and PA. The base is somewhere between the two levels.
  2. Position importance. Attributes that matter for his position get a random bonus, and ones that don't can lose a little. The result is kept within 1–20.
  3. Template value = base × the template's multiplier for that attribute, rounded down and kept within 1–20.
  4. The template value usually wins. About 23 times in 24, the attribute is replaced:
    • If the template value is at least the base, the new value is the template value −2 to +2.
    • Otherwise it's the template value +0 to +2.
    • Then, 1 time in 4, it's averaged with the CA level from point 1, rounded down.

      If the attribute would naturally have come out higher than the template value, there's an extra 1-in-20 chance the higher value is kept. A few attributes have extra checks of their own.
Example. CA 80 and PA 160. The CA level is 80 ÷ 20 + 6.5 = 10.5, rounded down to 10. Say the PA was used for Pace: 160 ÷ 20 + 6.5 = 14.5, so 14. The base is 10 plus a random 0–3, so 10–13. Say it's 12, and position importance leaves it at 12.
  • The winger template has ×1.3 Pace: 12 × 1.3 = 15.6, rounded down to 15. 15 is at least 12, so Pace becomes 13–17. 1 time in 4, say he rolled 16, it's averaged with the CA level instead: (10 + 16) ÷ 2 = 13.
  • The same template has ×0.6 Heading. With a base of 12: 12 × 0.6 = 7.2, so 7. 7 is below 12, so Heading becomes 7–9.

After this the game does one more pass over all the attributes. I haven't traced it fully, but it looks like it tidies them to fit his CA.

Bottom line: the attribute spread comes from the template, picked by position, height and region, and scaled by CA and PA. Your HoYD's strongest coaching attribute makes templates of that style up to twice as likely. His formation and tactics play no part.
 
What the templates look like in a real save

I read the whole template list out of a running save. I also logged every template pick on a March intake day: 62 youngsters across all the clubs with an intake that day. The rules in Steps 1–5 reproduced every single pick exactly. England is the only playable nation in this save, so the numbers below are for this save. Yours will differ, but the patterns should be similar.

The list. There are 335 templates, built from 10,091 real senior players.

RegionTemplates
England (its own set, because it's playable)75
Europe76
South America60
Africa51
North America45
Asia28
Oceania0 (every one was too rare and got deleted, Step 1b point 5)

The styles. "Share" is the style's share of the total weight, which is roughly how often it's in play.

StyleTemplatesShare
All-rounder22062%
Technical8130%
Mental176% (all 16 goalkeepers, plus one European pressing forward)
Fitness91.2%
Attacking40.6%
Tactical30.4%
Defending10.1%
Set pieces00%

So an all-rounder HoYD has something to boost in most positions, a Technical HoYD in many attacking ones, and a Mental HoYD for keepers only. An Attacking, Defending, Tactical, Fitness or Set-piece HoYD has almost nothing to boost.

The odd-roll quirk is strong. 36 youngsters in my intake were from European nations without a playable league. 18 rolled even and got European templates. Of the 18 who rolled odd, only 4 got a European template. 9 got a South American one and 5 a North American one. Two European keepers (177–185 cm) both got the same North American sweeper-keeper template.

Many shapes can never appear. Templates of other heights eat into the roll as the game walks past them (Step 4). In some positions that means the first template of the youngster's height always wins, and the rest are never reached.

Example: a tall English striker (186 cm or more, natural only up front). The striker templates for his height are Advanced Forward 27, Target Man 56, Poacher 11 and Pressing Forward 13. The total is 107, so the roll is 0–106. But the list is in role order, and other heights come first:
  • Deep-Lying Forward (177–185 cm), weight 17;
  • Advanced Forward (177–185 cm), weight 63;
  • Advanced Forward (under 177 cm), weight 33;
  • Advanced Forward (186 cm+), weight 27: the first one of his height.
By the time the walk reaches it, the count is 17 + 63 + 33 = 113, already more than any roll. So he always gets the Advanced Forward template. Target Man is the most common shape among real tall English strikers, and it never appears.

What English youngsters actually get. This is for a youngster natural in one position, with no HoYD:


PositionUnder 177 cm177–185 cm186 cm+
Goalkeeperno template (below)Goalkeeper 100%Goalkeeper 66%, Sweeper Keeper 34%
Full-backWing-Back 100%Full-Back 11%, Wing-Back 43%, Complete Wing-Back 46%Wing-Back 100%
Centre-backno template (below)Central Defender 100%Central Defender 67%, Ball-Playing Defender 33%
Wing-backWing-Back 100%Wing-Back 56%, Complete Wing-Back 44%Wing-Back 100%
Defensive midfielderDeep-Lying Playmaker 100%Deep-Lying Playmaker 36%, Ball-Winning Midfielder 64%Defensive Midfielder 18%, Deep-Lying Playmaker 82%
Central midfielderDeep-Lying Playmaker 48%, Box-to-Box 52%Deep-Lying Playmaker 17%, Box-to-Box 27%, Advanced Playmaker 45%, Ball-Winning Midfielder 11%Box-to-Box 100%
Wide midfielderWinger 100%Winger 83%, Defensive Winger 17%Winger 100%
Attacking midfielderAttacking Midfielder 32%, Advanced Playmaker 68%Attacking Midfielder 49%, Advanced Playmaker 51%Advanced Playmaker 100%
WingerWinger 100%Winger 69%, Advanced Playmaker 31%Winger 100%
StrikerAdvanced Forward 100%Deep-Lying Forward 12%, Advanced Forward 42%, Target Man 46%Advanced Forward 100%

These shapes never appear for these youngsters in this save, even though the templates exist: Inverted Full-Back, Libero, Wide Centre-Back, Mezzala, Wide Playmaker, Shadow Striker, Inside Forward, Wide Target Man, False Nine, Poacher and Pressing Forward. Sweeper Keeper only appears at 186 cm+, Ball-Playing Defender only at 186 cm+, Target Man only at 177–185 cm, and Complete Wing-Back and Ball-Winning Midfielder only at 177–185 cm.

Two things change the table:
  • Youngsters natural in more than one position. 35 of the 63 youngsters in my intake were, e.g. left-back and left wing-back. The templates for all their positions go into one pool, so their odds differ from the table.
  • Short keepers and centre-backs. There are no goalkeeper or centre-back templates for under 177 cm anywhere in the list: too few real players that short, so they were deleted. A keeper or centre-back under 177 cm, natural only in that position, gets no template at all**. Point 4 of Step 5 is skipped, and his attributes come from the base value and position importance alone. One keeper in my intake was like this.
What each HoYD style actually does here (English youngsters, one natural position):

HoYD stylePosition and heightNo HoYDWith this HoYD
All-rounderFull-back, 177–185 cmFull-Back 11%, Wing-Back 43%, Complete Wing-Back 46%Full-Back 22%, Wing-Back 78%, Complete Wing-Back 0%
Al-rounderCentre-back, 186 cm+Central Defender 67%, Ball-Playing Defender 33%Central Defender 100%
Al-rounderCentral midfielder, 177–185 cmDeep-Lying Playmaker 17%, Box-to-Box 27%, Advanced Playmaker 45%, Ball-Winning Midfielder 11%Deep-Lying Playmaker 34%, Box-to-Box 54%, Advanced Playmaker 12%, Ball-Winning Midfielder 0%
Al-rounderCentral midfielder, under 177 cmDeep-Lying Playmaker 48%, Box-to-Box 52%Deep-Lying Playmaker 9 %, Box-to-Box 6%
Al-rounderStriker, 177–185 cmDeep-Lying Forward 12%, Advanced Forward 42%, Target Man 46%Deep-Lying Forward 23%, Advanced Forward 77%, Target Man 0%
TechnicalAttacking midfielder (centre), 177–185 cm Attacking Midfielder 49%, Advanced Playmaker 51%Attacking Midfielder 97%, Advanced Playmaker 3%
TechnicalWinger, 177–185 cm | Winger 69%, Advanced Playmaker 31%Winger 100%
MentalGoalkeeper, 186 cm+Goalkeeper 66%, Sweeper Keeper 34%Goalkeeper 100 %
 
Two lessons:
  • It's a reshuffle, not "more of that style". In the under-177 cm central midfielder row, both templates are all-rounders. The all-rounder HoYD still pushes Box-to-Box down from 52% to 6%. The doubling helps whichever matching template comes first in the list, and takes the share from everything after it, whatever its style.
  • Shapes, not quality. CA and PA are already set before this (Parts 4–6). The template only decides which attributes are high and which are low.
Bottom line: the attribute spread comes from the template, picked by position, height and region, and scaled by CA and PA. The templates are averages of real players in your save: a youngster from a playable nation is shaped like that nation's senior players in his position. Your HoYD's strongest coaching attribute makes templates of that style up to twice as likely. In practice only an all-rounder, Technical or Mental (keepers only) HoYD has much to work with. Even then, he reshuffles which shapes arrive rather than making them better. Because of how the pick works, many shapes can never appear at all. His formation and tactics play no part.

---

God I'm glad that's over with. That was not enjoyable. This really isn't the place for almost 1,000 lines of information. I'd kill to be able to see my character count as I'm typing, too.
 
Premier League 3.0 weights for FM24

Based on 280,000 samples of every attribute in a realistic range for outfield + ~600,000 for GK.

FMST26 weights (click to view)

Accuracy:

85.9% - Harry Kane W 2974 D 2595 L 4431 1.57-1.86
85.2% - Robert Lewandowski W 2982 D 2589 L 4429 1.58-1.88

81.6% - Tammy Abraham W 2897 D 2539 L 4564 1.55-1.89
81.5% - Daizen Maeda W 2803 D 2556 L 4641 1.47-1.84

Seems pretty good to me.

GK isn't as accurate, but it's decent. I have a player list I compare to with 10,000 samples for dozens of GKs, I would say its about 85% of the way there.

I've included ST and DC specific weights, as these positions in particular require higher jumping reach. I've partly based these changes on caiograco's findings with FM27, which is a different FM but should still roughly apply to FM24. I myself have noticed the pace/acc and jumping reach differences between different positions when manually testing in the past, so I'm fairly confident these are valid changes - just without precision right now.

Attribute boosts:
  • ST
    • +5 acceleration
    • -5 pace
    • +10 finishing
    • +20 jumping reach
    • +3 strength
    • +5 leadership
  • DC
    • +10 acceleration
    • -10 pace
    • +30 jumping reach
    • +10 leadership
Personality boosts:
  • Outfield
    • +13 Professionalism
    • +8 Ambition
    • +12 Determination
  • GK
    • +1.3 Professionalism
    • +0.8 Ambition
    • +1.2 Determination
You can remove these adjustments if you want, but don't use the unadjusted version, because that lacks injury proneness, natural fitness, etc., which don't have effects in simulation but do in real play.

Most outfield attributes I took 2 x 2000 samples for. Not the best, but I didn't want it to take too long. I can always return to it and update attributes individually with more samples in the future. There's more attribute interactions to be added, I know of a few, but this will do for now.

For pace/acc the range I used is 10 vs. 20. For other attributes, its typically 6 vs. 16. My thinking is that you wouldn't sign or leave a player at 6< in a key attribute, and 17+ in attributes other than pace/acc is going to be rare - plus there may be a flatline going above 18 that distorts the picture.

I learned a few things from doing this btw:

- Left footedness does appear to have a performance bonus over right footedness
- Being of a different nationality to the nation the match/team is in didn't require or benefit from adaptability; there was no penalty for different nationality, at least not in the simulation testing. Adaptability itself matters for all players though, to a small extent.
- Long Throws is the worst negative attribute by far. I did 2 x 3000 samples a second time to confirm this. I don't know why this is.
 
Comparison of 2.1 vs. 3.0 unadjusted weights:

SGFnW6m.png


Note that nothing has been done to fit the players with either the unadjusted or adjusted weights. I simply used ChatGPT to calculate weights for each attribute based on the samples, and added these weights together into one formula, and have left it at that.

Surprisingly my 2.1 weights have held up pretty well. A few days prior to FM Arena going down, I discovered that I had made an error of not removing international match eligibility when testing players. The main issue with this was in regards to the Japanese players, as their team being greyed out means Maeda and Furuhashi benefited from not leaving to play international games. Osimhen getting called up also resulted in 2-3 games without him, unbeknownst to me.

So I had fitted the formula onto slightly erroneous results. It explains why it was so difficult to account for Maeda and Furuhashi.

I thought this must have ruined the whole thing a bit, but we see here that only Maeda & Furuhashi are really affected. Everyone else is within ~1-3% of the accuracy made possible through simulation testing.

This got me curious to compare past GS weights. I entered the GS data into FMST, excluding natural fitness, loyalty & injury proneness, as well as leaving the personality weight as intended rather than actual. Then I used Claude AI to analyze it.

hSqynwx.png


UPBa3Rw.png


So 2.1 did in fact suffer as a result of the error, but it could have been a lot worse.

The chronology of my files is:

1.0 GS
Pure Performance & Blended GS
2.0 (not included here)
2.1

Meta FMA 2.0 was Falbraav's I think, based on FM Arena's attribute testing with some adjustments.
Orion wasn't made by Orion, but someone who took his findings and put it in GS.
ykykyky05251 was I think a chinese user who purportedly was using AI in the cloud with hundreds of thousands of brute force tests or something - well before any of this latest gen AI stuff. Based on the results, you have to wonder what their methodology was..

Bafici's one dates all the way back to January 2025.

I'm pleased with my results overall, and surprised that it was the Blended one that ended up closest.
 
So what I've found when looking at training is modules train the attributes they say they do, but these attribute increases are competing with each other for a share of CA, as CA growth is fixed by other factors.

I hypothesized that if I want to increase dribbling say, then I should choose modules not only in which dribbling exists, but where there are few other pointless attributes, and more importantly fewer high CA pointless attributes such as decisions.

Turns out the first two prime candidates I selected based on this hypothesis are in fact the best (I've highlighted in blue the best that came later):

284.16 - [Physical]x2[Attacking]x2[Defending From the Front][Ball Distribution] | 130% intensity (125% GK)
282.46 - [Match Practice][Physical][Attacking][Defending] | 105% intensity (105% GK)
282.05 - [Physical][Tactical][Possession][Attacking Shadow Play][Defending From the Front][Ball Distribution][Shot Stopping][Play From the Back]
282.00 - [Physical][Attacking][Defending From the Front][Transition-Press][Handling][Shot Stopping][Distribution]
281.97 - [Physical][Attacking][Defending From the Front][Ball Distribution][Handling][Shot Stopping][Distribution]
281.91 - [Physical][Tactical][Possession][Defending From the Front][Ball Distribution][Shot Stopping]x2[Play From the Back]
281.81 - [Physical][Tactical][Attacking Shadow Play][Defending From the Front][Ball Distribution][Handling][Shot Stopping][Distribution]

281.75 - [Handling][Shot Stopping][Attacking][Physical][Chance Conversion][Aerial Defence][Ground Defence][Distribution] | 97.5% (130% GK)
280.95 - [Physical]x2[Attack]x2 | 100% (100% GK)
280.30 - [Physical][Attacking][Defending From the Front][Ball Distribution] | 80% intensity (70% GK)
274.49 - [Quickness][Match Practice][Attack]x2 | 115% intensity (115% GK)
261.14 - [Rest]

I've tried a whole bunch of combinations after the green ones, they're all 278-282.

One reason I wanted to get the new weights done is to test the training properly, as I knew using the 2.1 weights as a basis wasn't that accurate. These new ratings use the 3.0 weights derived purely from 280,000 outfield samples using controlled simulation of a change in each attribute.

Recovery affects the result negatively, don't use it - 3x recovery added reduced the score from 284.16 to 280.93. Adding 2x match review did similar. My theory is that its lower because its taking up a little bit of CA in contrast to rest.

I've only done this on an ST so far. Different positions will have slightly different results due to having different CA weightings, and then it all needs to be combined into one final figure, so there's more testing to be done on it.

More on the theory:

A cap on CA growth is determined by other factors, like professionalism, and later on also 'progress factor' (when the player comes within 30% of PA). But this CA growth isn't a fixed monthly increase based on professionalism, it's filled up to the extent of your training. So full rest is far below the CA growth limit for each month for a high PA player. But most schedules will hit the cap easily, as it only takes a few modules.

The most efficient training will hit the CA cap without going over it, otherwise you're putting in more modules but merely changing the proportions of what attributes are gained.

However what also seems to be the case is that if you can manage to go under the monthly CA growth cap, which is actually quite difficult to do, then you can change the proportions of attributes without compressing pace/acc gains.

The highest monthly CA growth cap is typically ~2 CA (20 CA/year).

If I do 2 x Quickness, I'll only gain say 1.3 CA/month, giving +0.6 pace/acc but leaving 0.7 CA on the table.
If I do 2 x Quickness + 3 x Attacking, I'll be gaining 2.2 CA/month, proportionally reduced to 2 CA/month, so +0.6 pace/acc x 0.9 = +0.54 pace/acc & +0.751 CA in other areas
If I do 2 x Quickness + 2 x Attacking + Ball Distribution (0.1 CA), I'll be gaining 2.0 CA/month, no proportional reduction, so I get +0.6 pace/acc & +0.7 CA in other areas


Because pace/acc are the best meta attributes, the 0.51 going from pace/acc to other attributes will always be inferior. So going above the 2 CA cap (or whatever the player's monthly max possible CA growth is) is always bad.

Now look at attacking vs. ball distribution. Ball distribution ostensibly trains mainly 'dribbling, passing, decisions, teamwork, vision'. Let's say its 20% dribbling, so Ball Distribution = Effectively +0.02 meta CA. I won't give the full list for 'Attacking', but let's just surmise dribbling and composure are 2 of the 12 = 16.6%. So like-for-like, ball distribution is more efficient, but the main reason we're using it is just to fit the 2 CA cap precisely.

Now take a look at a dud module, of which there are many, to varying degrees:

'Attacking Patient' = Finishing, First Touch, Passing, Technique, Composure, Decisions, Flair, Off the Ball, Vision, Teamwork.
There's nothing there except composure, and even then, it's 1 in 10. So it has a ~10% efficiency value.

'Defending Wide' = Heading, Marking, Anticipation, Decisions, Positioning.
Anticipation is good here, and there's not many attributes diluting it either, but unfortunately they're all CA-costly junk that turn it into a shit sandwich. Remember, its CA that's being added up to make up that 2 CA cap, not the attribute values. So at first it seems to be ~20%, but because of the CA costs, it's more like ~10% or less.

The CA cost would be calculated using the position's weights. If you look at HarvestGreen's excel sheets showing the stats for individual modules, you'll notice that endurance is curiously the lowest CA module apart from rest - it's even less CA than recovery. The reason is:

Endurance
+0.3 natural fitness (+0 CA)
+0.3 stamina (+0.18 CA)
+0.3 work rate (+0.06 CA)

Compare to Quickness, which would be:

+0.3 acceleration (+0.30 CA)
+0.3 agility (+0.18 CA)
+0.3 pace (+0.21 CA)

That's using weights for ST position.

If true, then what this also means is that intensity isn't the CA gain, as Endurance is 40% while Quickness is 45%. It is hard to say if it reflects effect on condition, or attribute value gain, or both.
 
Last edited:
YkOskG3.png


Upon further investigation, I was able to comprehensively buckbreak the FM24 training system.

It turns out that the reason why certain training schedules (i.e. [Physical]x2[Attacking]x2) do so well, beyond selecting the modules with the most efficient set of attributes, is that they are exploiting what to my mind must be an unintended bug, where certain technicals (usually set pieces) fall precipitously, freeing up CA to be redistributed into physicals as rest does.

I have witnessed this before, but never really gave any thought to it. But now when looking into it, I noticed again the technical drops in [Physical]x2[Attacking]x2, and thought to try and utilize it.

This sent me down a rabbit hole. Long story short, I've deduced that what causes the unending decline of useless technical attributes is when you don't use training modules that use individual roles. Even if it's the goalkeeper 20% section of the module that uses individual roles, it prevents the decline in technical attributes. For some reason, 'Attacking Wings', which does not have individual roles, also prevents the decline. But everything else is fair game.

Further, I found that the decline of useless attributes can extend to even more attributes, with the right combinations, which flows into the meta attributes even more for 0 CA cost. This leads to crazy good results like anticipation 65 > 93, concentration 65 > 87, pace 50 > 88, etc. while technique, crossing, heading, passing, etc. all 40 > 25.. and CA doesn't hit the cap even after 5 years (these stats are from a real result!).

Now the reason I think why no one else has cottoned on to this, at least as far as I know, is that a good deal of the modules have individual roles, and they're quite mixed together. So for instance, 'outfield' and 'goalkeeping' has individual roles, while 'overall' and 'defending' doesn't. Some of the combos also just don't make intuitive sense if you were going off just the names and looking for the magic combo. Physical + Defensive Shadow Play just doesn't sound like the winning ticket, does it?

Once you know this, it opens up a realm of possibilities for training options. I've tested a whole bunch of potential combinations and have found the best so far is simply [Physical][Defensive Shadow Play][Quickness Focus][Double Intensity]. It's so good that not only does it surpass the pace/acc gain of full rest, but it grows most other key attributes as well, and overall has a superior meta attribute score even to [Physical]x2[Attacking]x2.

agdPRl5.png


For further comparison, [Physical]x2[Attacking]x2, which is already a fairly light schedule, is 115 CA @ month 24, and ends up with 77.8 pace, 76.1 acc, 76.5 jump, 61.9 work, 63.6 bal. So this blows it out of the water. These stats are for ST (attacking unit), but I've chosen this particular schedule because I've also tested it does best of a few top candidates I tried with DC (defensive unit).

I've also manually tested this for 2 years just to make sure its valid, and to see what it looks like in normal play:

ST (Attacking unit) [Physical][Defensive Shadow Play]:

Sg64GHP.png


ST (Attacking Unit) [Physical]x2[Attacking]x2:

70Izlmu.png


DC (Defensive unit) [Physical][Defensive Shadow Play]:

TmW9kym.png


DC (Defensive Unit) [Physical]x2[Attacking]x2:

ctZ7D4f.png


I've found literally dozens of training schedules that are seemingly superior to the previous best, and there's a lot of variety.

This tactic I've chosen because high pace/acc trumps deficiencies in finishing, concentration, and dribbling.

But there are certainly more balanced, or even higher performance, options:

Highest score & equal highest Pace - [Physical]x2[Tactical][Defensive Shadow Play]x2 - 367.51 score, 89.2 pace, 85.4 acc, 82.2 jump. Low finishing (25.9) and dribbling (47.8).

Best all-rounder - [Quickness][Defending][Defending Unengaged] - 358.63 score, 84.7 pace, 83.1 acc, 80.0 jump, 48.2 finishing (decent), 73.7 dribbling (very good)

Near top dribbling + Very good overall - [Quickness][Possession][Play from the Back] - 357.87 score, 85.2 pace, 83.2 acc, 80.8 jump, 23.4 finishing, 77.1 dribbling

Best dribbling + Very solid all-rounder - [Quickness][Defending][Attacking Direct] - 349.44 score, 81.1 pace, 79.7 acc, 76.8 jump, 53.3 finishing, 78.0 dribbling

[Physical]x2[Attacking]x2 for comparison - 346.95 score, 77.9 pace, 76.1 acc, 76.5 jump, 61.8 finishing, 76.7 drib.
Fairly competitive at this pace/acc level, but sacrificing some finishing & drib for pace/acc is just better.
 
Last edited:
Amazing discovery!

How many slots per week do you put physical/defensive shadow play compared to rest?
 
Amazing discovery!

How many slots per week do you put physical/defensive shadow play compared to rest?
So for this training regime: [Physical]x2[Tactical][Defensive Shadow Play]x2

In a week you have 21 total slots:
2 Physical
2 Defensive Shadow Play
1 Tactical
16 Rest
 
Thanks! I realize now how stupid the question was, the answer is there in plain sight.
 
Interesting that the eight-friendlies trick works regardless of squad quality - since anything below the played level gets clamped to that 0.85 floor anyway, it is pure cheap volume rather than quality farming. Did you see any variance by opponent level within friendlies, or is MatchXP flat once you are under the floor?
 
Thanks! I realize now how stupid the question was, the answer is there in plain sight.
No worries. Since these tests started, a lot of the explanation has been increasingly technical, and sometimes even seemingly obvious things can be hard to wrap your head around. That's why I explained it in simple terms. :)
 
I spent hours playing around with it yesterday, so there's a lot of nuances to it. Here are most of them:

First a reminder to not add 'recovery'. If you are low on match sharpness, schedule friendlies instead. The light schedule probably actually works quite nicely with the fact that young players need 80 friendlies to hit max MatchXP for growth.

You can add match tactics and I think a few others like community outreach, without any change to the training results. I don't know if this means they have no effect at all when added in addition, but whatever the case, its another way to help manage fitness without using recovery. If they do work, then I would add one or two of them every week to build tactical familiarity, team cohesion and happiness. I'll experiment with this soon probably.

Quickness focus is still essential, nothing else I tried works.

If you double up modules, i.e. [Physical][Tactical] > [Physical]x2[Tactical]x2, it generally adds a small ~2% bonus. This wouldn't be worth the extra time setting up the schedule, injuries, and fitness costs (even in low sharpness situation, friendlies are superior). I obviously had to include the schedule that scored the highest, but I wouldn't recommend using it (it also doesn't as well for defensive unit btw, which is the more important reason).

Adding extra of the same module does also change proportions slightly to the same extent. So it will have ~2% bonus, but the jumping reach proportion will also increase or decrease ~2% on top of that for example.

The blue general blocks have strong deterministic effects on the outcome of the training schedule, here are the most important to understand:

Attacking - High growth, but usually puts a stopper to the technical decline. You'll notice it doesn't feature in any of my top new schedules, it's for this reason. It doesn't have 'individual roles', so it doesn't stop technical decline entirely. In [Physical]x2[Attacking]x2 it only declines set pieces. Others like crossing are hampered, but still stagnate rather than decline - the difference in outcome ends up being huge, because a slightly different rate is small at the beginning, but becomes 55 vs. 23 by the end.

Tactical - Lowest CA of the blue options, which means max pace/acc growth. Declining attributes: Crossing, Dribbling, Finishing, Heading, Marking, Passing, Tackling, Technique, First Touch, Corners, Long Throws, Free Kicks. The three in red are a bit of a problem. But finishing and first touch you can get away with on the whole, while dribbling is important, but only as critical as say agility.

Defending - 'Defending' modules in general do not do what you'd intuitively think they do. Ironically, all the 'defending' modules are actually the worst for defenders. And it's a hidden boon that I think is deliberately set up to mislead so you don't exploit it. You'll notice most of the best schedules I've listed either have 'defending' or defensive modules in them. Unlike 'Tactical', it has high dribbling and preserves finishing. The only attributes that decline are: Corners, Long Throws, Free Kicks - yet it ends up with fairly low CA. So where did the CA decline come from? It stagnates a whole bunch of (lesser) attributes, so it ends up being one of the most well-rounded and beneficial options. Because it stagnates rather than declines most useless technicals, that's its main downside.

Possession - Inbetween tactical and attacking. Excellent dribbling growth. Declines in: Crossing, Finishing, Corners, Long Throws, Free Kicks. So not much technical decline, and one we don't want. The other quirk of this module there is stagnation in certain semi-important attributes such as stamina, strength, concentration, natural fitness. This isn't quite so bad, because in reality you only need middling amounts of these attributes and it's conserving CA which will be redirected in other more important attributes.

---

We add another module to try and correct the deficiencies, while not activating individual roles or containing attributes we absolutely need to decline (so no 'technique' for instance). Normally you should read the attributes they affect and add based on that, but at this level where we are trying to alter progress rates of attribute types, it ceases to work intuitively and you just need to test different combos. It's clear to me that there are 'types' being activated between the scenes, even sometimes when no attribute is mentioned in the module that would be of that type. But if it is mentioned, then you should probably expect that type to be changed for certain. An example: A module says it has 'dribbling'. It will likely also stop decline of a few other technicals, even if those aren't mentioned on the module. This would be probably because its increasing 'progress_ball_control' rather than dribbling specifically.

I couldn't get anything to reduce decisions, in spite of it being the most CA costly attribute. Looking at HarvestGreen's spreadsheet, I see why - no module can decline it, not even rest, and the lowest is +0.5. If you think about it, since its going to grow anything, you may as well max out decisions, as you can probably avoid other bad ones like 'technique' (which can decline) by doing so.
Before finding what combos work, I first had to deduce what attributes I needed to try and stagnate/decline for each of the attacking unit (AML/AMC/ST) and defensive unit (DC/DL/DR/DM) and GK unit. They are different because positional weighting is taken into account for the attributes. So decisions and technique are bad for everybody, but marking is only bad for defensive unit.

For attacking unit (ST/AMR/AML):

reduce

agility (6) — 6.00
first touch (5-6) — 5.50
finishing (2-8) — 5.00
decisions (5) — 5.00
strength (3-6) — 4.50
technique (4) — 4.00
off the ball (2-6) — 4.00
crossing (2-5) — 3.50
heading (1-6) — 3.50

Very good:

transition (restrict) - none
penalties - none
endurance - none
community outreach - none

Good:

attacking overlap - heading (3.5)
chance conversion - heading (3.5), technique (4.0)
ball distribution - decisions (5.0)
transition (press) - first touch (5.5), decisions (5.0)
ball retention - first touch (5.5), decisions (5.0)
attacking shadow play - decisions (5.0), off the ball (4.0)
set pieces routines - heading (3.5), off the ball (4.00)
resistence - strength (4.5)
quickness - agility (6.0)
tactical - decisions (5.0), off the ball (4.0)

Bad:

attacking wings - crossing (3.5), finishing (5.0), heading (3.5), off the ball (4.0), technique (4.0)
attacking direct - finishing (5.0), first touch (5.5), technique (4), off the ball (4.0)
chance creation - crossing (3.5), technique (4.0), decisions (5.0), off the ball (4.0)
attacking - first touch (5.5), finishing (5.0), technique (4.0), off the ball (4.0), crossing (3.5), heading (3.5)
possession - first touch (5.5), technique (4.0), decisions (5.0), off the ball (4.0)

For defending unit (DC/DR/DL/DM):

reduce

decisions (5-10)
positioning (4-8)
agility (6)
marking (3-8)
tackling (4-7)
strength (4-6)
heading (1-5)
first touch (2-4)

Very good:

play from the back - first touch (2-4)
endurance - none
community outreach - none
defending from the front - none
penalties - none
attacking overlap - none

Good:

resistance - strength (4-6)
quickness - agility (6)
physical - agility (6), strength (4-6)

Okay:

attacking - first touch (2-4), heading (1-5), marking/tackling/heading/positioning/decisions (x0.33 impact)
possession - decisions (5-10), first touch (2-4), marking/tackling/positioning/decisions (x0.33 impact)
tactical - decisions (5-10), positioning (4-8)
defensive shadow play - decisions (5-10), positioning (4-8)
ground defence - marking (3-8), positioning (4-8), tackling (4-7)
aerial defence - marking (3-8), positioning (4-8), heading (1-5)
defending from the front - marking (3-8), positioning (4-8), heading (1-5), tackling (4-7)
overall
outfield - decisions (5-10), first touch (2-4), heading (1-5), marking (3-8), positioning, tackling
attacking direct - marking, tackling, positioning
transition (press) - first touch (2-4), tackling (4-7), decisions (5-10)
transition (restrict) - tackling (4-7), marking, positioning
attacking shadow play - decisions, positioning
set piece routines - heading, marking, positioning

Bad:

defending engaged
defending disengaged
defending wide
defending - heading (1-5), marking (3-8), tackling (4-7), decisions (5-10), positioning (4-8)
match tactics - decisions (5-10)
attacking wings
attacking patient
ball distribution
ball retention

If it wasn't listed, it's bad. I won't do GK here, but let's just surmise Shot Stopping, Handling & Possession are the best non-physical options.

You'll notice there are very few good options for the defensive unit.

------

Combining all this, I came up with:

Transition (restrict) - 0.4/2/7 anticipation, concentration (both)
Endurance - 1/3/3 stamina, natural fitness, work rate (both)
Penalties - 0.2/1/2 composure (both)
Play from the back - 0.6/2/7 DC | 0.2/2/7 ST dribbling, composure
Defending from the front - 0.6/1/5 ST concentration 0.2/2/6 DC dribbling, anticipation
Tactical - 1/3/8 (both) anticipation, composure, concentration
Attacking Overlap - 0.6/2/6 DC dribbling, anticipation 0.2/1/5 ST concentration
Chance Conversion - 0.6/2/6 ST anticipation, composure 0.2 DC roles
Attacking Shadow Play - 0.6/2/5 ST anticipation, composure 0.2/2/5 DC anticipation, concentration
Transition (press) - 0.4/2/8 (both) anticipation, work rate
Ball Retention - 0.6/2/5 ST anticipation, composure 0.2/2/7 anticipation, concentration
Set Piece Routines - 0.2/2/6 (both) anticipation, concentration

How to read this: 0.2 growth x 2 key attributes, diluted by a total of 6. The two key attributes are anticipation, concentration.

I further reduced this to:

play from the back 0.6 DC 0.2 ST
penalties 0.2 DC 0.2 ST
defending from the front 0.2 DC 0.6 ST
attacking shadow play 0.2 DC 0.6 ST
attacking overlap 0.6 DC 0.2 ST
physical
quickness

As mentioned previously, this is for just standard training module calculations. When you are talking about what declines attributes in the right way, that is a separate matter that can used in conjunction with this info, but it works completely differently. You'll notice 'defensive shadow play' is not listed here for instance, because normally it sucks.
 
God I'm glad that's over with. That was not enjoyable. This really isn't the place for almost 1,000 lines of information. I'd kill to be able to see my character count as I'm typing, too.
I did read these extra posts, but didn't respond because while its good info to know, I didn't see anything too consequential in there to comment on. Plus I need more time to digest it. And then I was doing weights and training, of the FM kind.

I'll come back to it again later on. I just wanted to say I appreciate your efforts in finding out the info.

Interesting that the eight-friendlies trick works regardless of squad quality - since anything below the played level gets clamped to that 0.85 floor anyway, it is pure cheap volume rather than quality farming. Did you see any variance by opponent level within friendlies, or is MatchXP flat once you are under the floor?

I don't blame you for not getting it entirely correct, its very complicated, but I think you have misread it:

the x0.85 level multiplier is applied in friendlies to the combo of:

- average CA of all players in the match, weighted by total minutes played by each player
- opposition club rep

So opposition quality does matter, even in friendlies.

And also, in case you don't realize, if a player hasn't played 11+ competitive matches in one season yet, his multiplier for friendlies is 1.5 not 0.85.

You want to keep opposition quality as high as possible, but at the end of the day it's not going to reduce the need for 80 friendlies really. It more helps ensure players will still have max growth in spite of injuries, non-selection for matches, etc. by allowing a buffer (i.e. can miss 1 in 6 matches).
 
Back
Top