The hidden, variable, probabilistic cap to CA growth

[GeorgeFloydOverdosed]

I was going to sit on this and put it in my guide, but my guide is going to take a while, and I suppose it would be a long section buried in the middle and be overlooked by all but a handful anyway.

Recently there was this video. I've translated it using AI and created a subtitle file you can download here. You just need to download the bilibili video to play with it. Or you can just open the srt file in notepad and read along. But I'm going to summarize it now anyway.

In short, there is a limit to the amount of PA a player’s CA can fulfill, but also that this limit is unlockable and in fact is frequently unlocked throughout normal play with the human manager oblivious to the fact.

When the human player starts the game and becomes the manager of a selected team, each player is assigned to a ‘tier’ (‘1’ to ‘5’) in this probabilistic table:

The formula that is fed through to this probabilistic table is as follows:

(((Professionalism+Ambition+Determination)/3))-4)+(training facilities/4)+((morale-10)/2)+(2 if Captain)+(1 if Vice Captain)

There are one or two parts of the formula I didn't include because it was vague in translation, but I think they're things you have little to no control over, like how much your player likes a staff member or something.

So, if we consider a likely scenario for our own promising 18-year-old player once they reach the Premier League:

(((14+12+12)/3))-4)+(14/4)+((12-10)/2)+(0)+(0) = 13.16

Their most likely tier is ‘3’ (55%) followed by ‘4’ (30%).

The tiers correspond to the following CA cap as % of PA:

• 70%
• 75%
• 85%
• 95%
• 100%

Therefore 85% of the time, our typical promising player is going to be capped to 85% or 95% of his PA. We probably brought in a player with a high PA of say 150, so 85% would mean he will stop at 127 CA while 95% means he will stop at 142 CA.

Even in the unluckiest situation though, a 150 PA player will reach 105 CA at least.

Even with max points from the formula (‘20’), the probability of tier ‘5’ and therefore reaching 150 CA with 150 PA, is still only 35%.

However, these tiers are not set in stone, they are re-rolled whenever one of the events from this (non-exhaustive) list occurs:

• New manager
• Training facilities change (improvement or decline)
• New or renewed player contract (re-rolls on date new contract starts)
• Player moves on transfer/loan (but not when player returns from loan)

In theory this can be gamed, but given these kinds of changes are still only something you’d attempt once or twice a year, you’d still want at least a formula score of 17 ideally.

I also have slot in somewhere here that a player's CA can stop in-between levels. So for example:

150 PA x 0.85 = 127.5 CA cap
> Event: New player contract > Dice roll > Rolls Tier 4 (new CA cap of 95% of PA)
150 x 0.95 = 142.5 CA cap
> Event: New player contract > Dice roll > Rolls Tier 3 (new CA cap of 85% of PA)
150 x 0.85 = 127.5 CA cap, but he already grew to 134 CA, so now his CA stays at 134 CA until he declines or rolls a Tier 4/Tier 5 later on.

It is reasonable to find players with 14pro/12amb/14det (13.33 average). Morale can be kept high (~15) with an exploit tactic. With 17 training facilities, the formula result is only 16.08, but this is the realistic reality. So it’s worth keeping in mind to try and get all these factors as high as possible, because you’re going to need every bit you can get. ~130 CA is sufficient for an ‘ideal’ meta player with allowance for some fluff around the edges, so 150 PA serves as roughly the minimum for an ideal player in a realistic scenario.

There is a bit more to this that needs to be explored. In using LightningFlik's harness training simulation feature, I was left scratching my head over why 150 Match XP resulted in a player with 144 PA stopping at ~100 CA, while the same player with 100 Match XP stopped at ~108 CA. Now with this knowledge part of the picture is solved:

144 x 0.75 =108
144 x 0.70 = 100.8

But the question that remains is, why does more match experience seem to result in a lower CA cap?
Coach workload also seems a bit wacky. For one player I've tested, 80% coach workload is identical to 1% workload. But final CA took a big hit at 90% and 100% workload. Does this mean only 'very heavy' workload impairs growth, or is this specific to the player's unique set of circumstances? Do training factors work together as a complex interacting web rather than simple linear additions.. or are there bugs with LightningFlik's program..

So there's a few things that need further testing to find out, but I think one thing that can be said overall is that because the CA-PA gap is in practical reality lower than we think it is, and less straightforward to fill it as well, each training factor has less importance and is not a silver bullet on its own - whether its 20 professionalism or 20 5 star coaches, or a meta training schedule. I think it's going to be possible to derive some rough optimums. For example, we could use a starting point of 14pro/14det/12amb as attainable and work from there to derive that say 127 CA is likely and to get that you need X level facilities and X quality coaches.

 

[squirrel_plays]

Interesting stuff.

One thing I'm noticing as the game engine gets more reverse engineered, is how well conceived all these various mechanics are. Fair play to the developers over the years, they have created a really interesting thing that still remains interesting even as it is gradually revealed.

I was expecting my immersion to be impaired by finding out about stuff like this but that hasn't been the case, and I think it's because the logic that underpins it is well considered.

 

[mitt]

I did some digging and I found out people already know about this back around 2019-2020 and SI called it Progression Factor
there is a screenshot of a tool that show Progression Factor with the exact same data as this table
what was not known is how exactly this progression factor work with vague description like calculated with age and performance

 

[GeorgeFloydOverdosed]

I did some digging and I found out people already know about this back around 2019-2020 and SI called it Progression Factor
there is a screenshot of a tool that show Progression Factor with the exact same data as this table
what was not known is how exactly this progression factor work with vague description like calculated with age and performance

You're right I think, a lot of aspects match up. It would either be the same thing, or a sibling of it.

The tool screenshot I knew I'd seen from somewhere, but couldn't recall where. Eventually it came to me - it's on the page of the guide for FM20 Italy researchers, and is from the internal 'mini editor' they use:

The guide says:

For illustrative purposes, there is the Progression Factor, a value given by the combination of attributes: Ambition, Determination and Professionalism.​

​

This value represents, at the beginning of the game and up to a subsequent recalculation due to an event that occurred during the game (transfer, favorite manager / hated, trainee structures, morals, etc.) the likely speed with which the player increases and improves his attributes, being the only mental values and excluding the other factors.​

​

The speed of progression is not related to the CA but to the individual attributes. In the case, therefore, of a more experienced player who does not have room to improve their CA, the value represents how quickly the attributes can be redistributed with training.​

​

So we can see here a score example of '17', and also some of the events that trigger the reroll of the tier.

The quote in orange is giving me pause for thought about what they mean exactly.

One thing I'm noticing as the game engine gets more reverse engineered, is how well conceived all these various mechanics are. Fair play to the developers over the years, they have created a really interesting thing that still remains interesting even as it is gradually revealed.

I was expecting my immersion to be impaired by finding out about stuff like this but that hasn't been the case, and I think it's because the logic that underpins it is well considered.

I was thinking the same recently after some of LightingFlik's explanations of the mechanics.

I was thinking of doing another post on how CA itself supposedly plays a key part in the match engine. But after thinking it through, the conclusions just don't add up. If it's right, +1 decisions should result in more than +1 for almost every mental & technical attribute during a match. You see why I consider this finding about CA in the match engine to be significant if true..

On the one hand, it would be a satisfying explanation for why decisions is so costly in terms of CA. But then, it's fact that +1 decisions doesn't improve the win rate. It could be because the guy who reverse engineered this info, did so from FM10, so it could be just outdated. Or it could be that I'm not thinking about it correctly. I'll think about it more and see.

 

[LightningFlik]

If this is true then the progression factor must exist somewhere in memory for a given player, right? If it's hard-coded based on their mental attributes put through some formula then it wouldn't need storing, but if it's based on random chance + transient factors like whether you were captain at a specific point in time, the value _has_ to be stored somewhere.

I've never seen a scouting tool or real-time editor highlight this though. (Then again I'd never seen them highlight the late bloomer value I discovered, but it would shock me if two major things had been overlooked).

**EDIT:** also @GeorgeFloydOverdosed, my testing revealed that any amount of coaching workload had a negative effect on coach effectiveness (which makes sense; why even have levels if only the last one counts?) I didn't do a lot of research on it though so I can't give you any special insights.

All this talk is enough to make a person consider buying this game.

 

[mitt]

If this is true then the progression factor must exist somewhere in memory for a given player, right? If it's hard-coded based on their mental attributes put through some formula then it wouldn't need storing, but if it's based on random chance + transient factors like whether you were captain at a specific point in time, the value _has_ to be stored somewhere.

I've never seen a scouting tool or real-time editor highlight this though. (Then again I'd never seen them highlight the late bloomer value I discovered, but it would shock me if two major things had been overlooked).

**EDIT:** also @GeorgeFloydOverdosed, my testing revealed that any amount of coaching workload had a negative effect on coach effectiveness (which makes sense; why even have levels if only the last one counts?) I didn't do a lot of research on it though so I can't give you any special insights.

All this talk is enough to make a person consider buying this game.

You are right that the progression factor is indeed stored in memory for players, the video GeorgeFloydOverdosed linked contains a download link to a tool that the video creator made with AI to demonstrate and check the progression factor level.
The tool had a version for FM24 FM26 and FM27 but the FM27 one is created before the release so the tool no longer works on the current version of FM27 and the creator said he will not be updating the tool since it was just a POC and he doesn't want to update the tool for every small update FM27 has.
However the tool is open-source and written in python so it shouldn't be hard to check for yourself.

 

[meanonsunday]

The one thing I don't find plausible about this is that I've had saves where I've edited newgens at 15/16 yo to represent historic players from the past. I've always, 100% of the time, been able to get them to reach their full PA. No transfers, one or at most 2 new contracts. This would simply be impossible if the maximum probability of that happening is 35%. There definitely are players that reach a point well before mid twenties where CA only increases very slowly or not at all. But there must be other details to the system; it may be a rate limit to CA growth rather than an absolute maximum, and perhaps different effects at different age ranges. What I'm sure of is that if you have a newgen that spawns with good initial CA, and high professionalism, determination and ambition, you will be able to reach their full PA with the right training and match experience.

 

[GeorgeFloydOverdosed]

**EDIT:** also @GeorgeFloydOverdosed, my testing revealed that any amount of coaching workload had a negative effect on coach effectiveness (which makes sense; why even have levels if only the last one counts?) I didn't do a lot of research on it though so I can't give you any special insights.

I fiddled around with the training simulation settings for a fair while, but kept coming across inconsistencies, so I haven't been documenting much or taking too seriously the results.

When I found that coach workload penalty only kicks in at 90%+, other conditions were largely ideal I think, so I don't doubt that with different conditions, you would have got the result that coach workload matters at other % levels if you say you did.

I couldn't replicate your 100 vs 150 matchXP results, and it seemed to me there were other issues. It all made me wonder if it worked for you via the CLI, but the GUI doesn't work as intended perhaps? I posted on the github as an issue, but I don't think you've seen it.

Or perhaps you only tested 1 month? Because I've noticed significant differences in the first month, but then it seems after 36 or 60 months or whatnot it generally catches up to the same final amount in spite of the initial difference. I figure maybe it's a case of there being a problem of simulating the same thing over & over artificially rather than a true natural progression of the player.

I guess I'd like to know if you can confirm the GUI is working as intended and shows the right results over 36 months say.

The one thing I don't find plausible about this is that I've had saves where I've edited newgens at 15/16 yo to represent historic players from the past. I've always, 100% of the time, been able to get them to reach their full PA. No transfers, one or at most 2 new contracts. This would simply be impossible if the maximum probability of that happening is 35%. There definitely are players that reach a point well before mid twenties where CA only increases very slowly or not at all. But there must be other details to the system; it may be a rate limit to CA growth rather than an absolute maximum, and perhaps different effects at different age ranges. What I'm sure of is that if you have a newgen that spawns with good initial CA, and high professionalism, determination and ambition, you will be able to reach their full PA with the right training and match experience.

The results from LightningFlik's training simulation though suggest it really does put a stopper to CA.

There are more triggers beyond the ones I mentioned. I decided to denote it was a non-exhaustive list rather than to include them all.

Remember too that CA doesn't have to decline if the player drops to a lower tier, it can simply stay at the CA it reached, and then continue growing to 100% PA once its tier increases again.

 

[LightningFlik]

I couldn't replicate your 100 vs 150 matchXP results, and it seemed to me there were other issues. It all made me wonder if it worked for you via the CLI, but the GUI doesn't work as intended perhaps? I posted on the github as an issue, but I don't think you've seen it.

Ah, my bad. This is exactly how it works. The CLI repeats the match_experience value but the GUI doesn't. I've just pushed a new version which sees the GUI repeat it too. Fill in the first page as normal and then, on the Simulation tab, add a value to the new Match Experience box; this is months 2 and onwards.

Normally the game zeroes your match experience every time you develop, meaning you need to get more match experience afterwards. The CLI let you specify experience for months 1, 2, 3 etc. but the GUI just let it zero the value and then read _that_ for every month afterwards.

To my shame, I rushed this feature out. I was getting so burnt out towards the end and just wanted to get something out before FM 27 dropped and made it all redundant anyway.

 

[LightningFlik]

I found where this progression factor is stored in memory for FM 24 (Player+0x263, if anybody is interested, four bytes after where their Happiness is stored) add added it to my development simulator app.

I took a youngster, gave him 57 CA, 166 PA, decent mentals, good coaches, lots of match experience and good facilities and then ran five years of development across 100 seeds with values of 1 and 5 for the progression scale value.

Scale	1​	5​
12 months	88.53	97.2
24 months	109.54	118.1
36 months	116.13	135.64
48 months	116.29	147.19
60 months	116.44	161.17

Blimey.

Here's how it works from what I can see (apologies if I'm repeating you @GeorgeFloydOverdosed or our Chinese friends):

What it does
Values 1-4 stop a player's growth at 70, 75, 85 or 95% of his PA; tier 5 has no cap.

How to see it
The game tells you the values for a player in the social feed of all places if they're 5 or 1.

• Value 5: journalists post praise such as "He's putting the work in, and [club] staff have confirmed as much to me".
• Value 1: worried posts such as "[club] are definitely concerned that [player] doesn't have the work ethic required to succeed". These only appear when an flag on the player is set, which I've not identified yet.

How it's determined
When a player's tier is 0, the monthly development computes a score from 1 to 20. It then rolls 0–99 against that score's row in a fixed table of tier chances. A high score shifts the odds toward tiers 4–5 but never guarantees tier 5: even score 20 gives it only a 35% chance.

How the score is built, in order:

1. Personality: (Ambition + Professionalism + Determination) / 3 - 4
2. Training facilities: + facilities / 4 (at most +5)
3. Contract flag: -3 if the player is transfer listed. (Transfer listing also triggers an update of the value.)
4. Morale: + (morale - 10) / 2
5. Club value from 1 to 20: + (value -10) / 3. It sits next to training facilities but isn't a facility. It drifts with the first team's reputation and sets a new signing's starting morale, so it looks like club atmosphere
6. Relationships: capped at +/- 5 in total:
   +/- 3 for liking or disliking the first-team manager;
   +/- 1 or 2 for each staff member on the same senior/youth side (2 for assistant manager and 1 for other staff);
   +/- 1 for each squad teammate.
7. Captaincy: +2 for captain or +1 for vice-captain of his own team, but only with Leadership 9+.
8. Language: if the player doesn't speak the language of the first team's stadium's city (i.e. he has a 0 in that language), -1 if someone at the club speaks one of his languages at level 2 or more and has been there over 100 days (or is a staff member); -3 if nobody shares a language with him.
9. League or manager standing: +1 if the first team's league is in a certain state (possibly an unloaded, view-only league?). Otherwise, at a human-managed club, -2 to +2 from a manager standing value from -100 to 100; in my test game, it's 71, worth +2. AI clubs get nothing from this. (I need to dig in to this more).

There are several places where the value is re-rolled and they work in two modes.

Replace mode

• First roll: at monthly development, when the value is still 0.
• A move completes: transfer, trade or a loan buy-back clause.
• A contract starts
• Club-wide, everyone at the club:
  • any change to training facilities
  • a new manager is hired

Lower-only mode
The new value is the lower of the old one and the new roll, so it can only fall.

• Transfer/loan listed
• He goes AWOL.
• Long term injury or ban (14 days or more or high severity)
• Personality change after an event

Latest version of fm24ctl, which makes the field configurable in tests.

 

[caiograco]

@LightningFlik beat me to it, but I also found the same thing on FM27 save files. We can actually use an updated version of fmsave to read those values. In short, Every player has a stored value from 1 to 5. Comparing two saves of the same career, it behaves almost exactly as described above.

What I did to confirm was to point an agent to FM27 save files over 12 seasons, as well as some other saves, like one I had on the FM26 save file.

First, if you want to look it up on your own saves. You can use fmsave + an AI to modify it (I will push my version later today):
These values are in the save's game_db section, inside each player's record: one byte, 7 bytes after the start of the current ability field. CA is a 16-bit value at +0, PA a 16-bit value at +2, the tier byte at +7.
The values are indeed 1-5 = tiers 1-5 and 0 = not rolled yet.
FM26 has a byte with the same 0-5 values at the same place, but my FM26 save was too early in its career to confirm it acts the same.
In-memory offsets will differ, but the field should sit next to CA/PA there too.

The caps are exactly 70 / 75 / 85 / 95 / 100% of PA. Players who stopped growing (same tier in both saves, PA unchanged, age 19-24, CA unchanged over 10 months) pile up at:

tier players most common CA/PA median CA/PA
1 1,827 0.69-0.70 0.71
2 5,398 0.74-0.75 0.75
3 10,771 0.84-0.85 0.85
4 3,175 0.94-0.95 0.93
5 882 0.99-1.00 0.94

Growth over the 10 months against room left under the cap (same tier both saves, PA unchanged, age 18-25):

CA points under the cap players mean CA change share gaining 3+
more than 10 above 3,307 -0.1 1%
2 to 10 above 5,308 -0.2 2%
within 2 10,990 -0.1 2%
2-5 below 2,631 +0.6 19%
5-10 below 4,051 +1.3 29%
10-20 below 5,657 +1.9 33%
20+ below 5,867 +2.2 35%

So it is a hard stop, not a slowdown. CA above the cap doesn't drop back. A player re-rolled to a lower tier keeps what he has and just stops, as George described.

It applies at every age

Share gaining 3+ CA in 10 months:
age 10+ below cap 1-10 below at/above cap
15-17 79% 51% 5%
18-19 62% 37% 4%
20-21 28% 17% 2%
22-23 24% 16% 1%
24-25 10% 9% 1%
26-27 3% 5% 0%
Growth slows with age anyway, but at the cap it stops at every age.

When it's rolled
- On the first day of a new career, almost every player is 0: 99.7% in my career, 98.8% in a fresh 27.0.1 career.
- By the season-2 save, everyone has a tier. So it's rolled soon after the start, not stored in the database.
- New regens show 0 until their first roll.

What triggers a re-roll?

Share of players whose tier changed between the two saves:
event (between the saves) players tier changed
none, same first-team manager 24,441 0.7%
first-team manager changed (no personal event) 1,041 62.7%
renewal signed but not started yet 650 0.5%
new contract started at the same club 22,299 59.6%
moved club (transfer, free, release) 29,508 45.2%
loan started or ended 39 62%

A re-roll can land on the same tier. If every renewal were a fresh roll, about 68% would change tier. 59.6% did, so roughly 87% of renewals re-roll. For moves the same sum gives about 68%.

The re-roll happens on the day the new contract starts, not when it's signed, as the first post says. Renewals that were signed but hadn't started yet changed no more often than players with no event at all.

Manager changes are clear club-wide events. In 35 of the 37 clubs whose first-team manager changed, half the squad or more changed tier. Under the same manager, almost no club had more than a handful of changes. Three clubs had a mass re-roll with no manager change, which would fit a facilities change, but I can't confirm that yet.

What decides the roll
Tier after a renewal or a move, by (Professionalism + Ambition + Determination) / 3:
average tier 1 tier 2 tier 3 tier 4 tier 5 players
< 7 23% 53% 19% 5% 0% 1,660
7-9 20% 46% 27% 6% 1% 4,319
9-11 14% 36% 40% 8% 2% 11,146
11-13 8% 25% 48% 14% 5% 15,464
13-15 5% 15% 50% 22% 8% 15,161
15-17 3% 10% 45% 31% 12% 3,795
17+ 1% 5% 25% 48% 21% 112

The club matters about as much. Same players, average 12-15, by club reputation:
club reputation tier 1 tier 2 tier 3 tier 4 tier 5
< 2,000 10% 29% 50% 8% 2%
2,000-4,000 7% 24% 53% 11% 5%
4,000-6,000 4% 13% 51% 24% 9%
6,000-8,000 2% 8% 39% 36% 14%
8,000+ 1% 5% 28% 44% 22%

That fits training facilities being in the formula, since bigger clubs have better ones. I haven't located the training facilities rating in the save yet, so I can't test it directly. I couldn't test morale or captaincy either.

For comparison with the formula in the first post: a 13-15 average at a mid-sized club gives tier 3 about 50% and tier 4 about 24%. That's close to the 55% / 30% quoted for a score of 13.16.

Notes on meanonsunday's point
- Reaching full PA needs tier 5, which is rare. But tiers re-roll on every renewal and manager change, and CA kept from a good roll isn't lost on a bad one. So a player who gets 2-3 rolls during his growth years has a decent chance of hitting tier 5 at some point.
- High drive at a big club makes tier 4-5 the most likely roll: about 66% at an 8,000+ reputation club.
- Edited players may also be different. If you edit a player's personality, his stored tier only changes at the next roll.