How did Germany get almost 500 divisions?

[tom_jones]

This does work. Try changing default templates to add stuff like self-propelled guns and self-propelled anti-tank brigades and then report back, but before you do make sure the game proceeds at least until 1941-1942. Problems start to show or accumulate as time progresses.

If you still manage to have AI divisions looking like this and with more advanced stuff, tell us how you did it.

Keep on mind the AI's performance with the templates improved in 1.3, so the problems present in 1.2 may not necessarily still be there. At least as long as the configuration is limited to generic templates, there seems to be bugs when things are more nuanced, but that's another story.

 

[Ancalimé]

Besides the condecending and semi-agressive tone, you're pretty much right.

Although it might be worth playing around with these numbers:

MIN_FIELD_STRENGTH_TO_BUILD_UNITS = 0.7,            -- Cancel unit production if below this to get resources out to units in the field
MIN_MANPOWER_TO_BUILD_UNITS = 0.7,                    -- Cancel unit production if below this to get resources out to units in the field
MANPOWER_FREE_USAGE_THRESHOLD = 500000,            -- If AI has this much manpower he doesn't care about the percentage
MANPOWER_RESERVED_THRESHOLD = 0.25,                    -- The AI will not deploy more units if he goes below this percentage

Notice the part of the free manpower usage, if I'm correct.. as long as the AI has no less than 500k manpower. it will keep producing units and ignore the 25% of maximum manpower threshold.

Let's say I have 10mil recruitable manpower as the soviet union with a 25% reserve treshold ( that's 2,5 mil), it doesn't matter because the defines lua says to keep building units as long as you're above 500k manpower. Assuming I'm understanding this correctly.

That was literally the first thing I did.

Keep on mind the AI's performance with the templates improved in 1.3, so the problems present in 1.2 may not necessarily still be there. At least as long as the configuration is limited to generic templates, there seems to be bugs when things are more nuanced, but that's another story.

Yeap. I'm noticing a lot of improvements in AI behaviour with 1.3.
I'm doing the tweaks from scratch, see where it takes me.

 

[Nobody987]

Here is update from my side also. In mentioned mod on page before, I created some different templates for AI. Here are the examples

I should have said "as intended".

This does work. Try changing default templates to add stuff like self-propelled guns and self-propelled anti-tank brigades and then report back, but before you do make sure the game proceeds at least until 1941-1942. Problems start to show or accumulate as time progresses.

If you still manage to have AI divisions looking like this and with more advanced stuff, tell us how you did it.



The hard part is balancing.
The first thing I did was to make it hard for the AI to do division deployment. That worked beautifully. The AI was able to keep its armies equipped and the game became more challenging and a lot more fun. Obviously now there were a lot less divisions around, which may or may not be what you are looking for.

However, this approach became a problem when I realised China -consistently- could not stand against Japan. When I'm modding the game, I'm trying to do two things:
1) Make it perform better and more realistically.
2) Make it perform historically accurate.
Hence, having China getting knocked out by Japan by 1940-41 each time (something I'm noticing happens a lot with the new patch) was not acceptable for me. So I started tweaking to allow the AI to build enough divisions for Germany to quickly despatch Poland, France and the Low Countries (one thing I did notice making this hard was that the German AI was declaring war against the Low Countries by 22 September and the war was turning into a two front war, making it a lot harder for Germany. So I changed national focus priorities to give it enough time to take care of Poland first) and for China to hold against Japan. This is the hardest part of modding the AI and division deployment defines, at least for me. Because I want the game to progress historically. If I tweaked something a bit too much for example, China would then overwhelm Japanese forces. If I did it a bit too little then Japan would overtake China by ~1941.

Unfortunately before long, I realised the game is just not designed to run historically, nor, it seems, logically. It was just too unpredictable. Also, one thing you change usually has a domino effect that needs further refinements on other fronts. That's why thorough testing is important.
When testing your mods, make sure the game plays until at least 1942, at least. Stuff that seems to be working early game can **** **** up once the map starts shifting.

With that said; you guys might want to tweak AI naval invasion numbers as well. In cases where Japan is driven out of mainland China, it will keep mounting small scale naval invasions on China and Korea with 1-2 divisions. Sometimes it will succeed at gaining a beachhead and start unloading more and more divisions there. In the meantime, the Chinese army will reorganise and create a front around where Japan landed and will attack, eventually regaining all the naval bases and cutting the Japanese army off. This in turn will result in the loss of all divisions there. This will keep happening and soon enough, Japan will have no divisions left. I mean nothing, zero, nil, naught. At this point, as soon as it deploys a new division, it will mount an invasion and the same thing will keep happening over and over again, like a very shitty groundhog day.
This happens with Britain too. They will keep invading Italy and sometimes they will succeed, but then they will be overwhelmed and will lose all divisions there.
I tried making it harder for the AI to invade overseas, but that doesn't work well in the Pacific, where you want there to be invasions taking place. So I found a compromise by making the AI mount larger invasions to increase their rate of success. This works to an extent. It makes it quite harder for Japan to lose all its divisions, but it also makes Italy easier to be invaded by Britain.

So to conclude; if you're not as compulsive as I am when it comes to historical gameplay, then you may be satisfied with the results. I, unfortunately, cannot be.

Here is update from my side also. In mentioned mod on page before, I created some different templates for AI. Here are the examples

This is done for Spain for example, to show that it can be applied not only for majors. Also using cheats, obviously, but still if the game progresses far enough results would be similar.
Also, AI template upgrade behavior is strange, it creates worse template than the one that already has, and it doesn't use it at all.

And this is the part that I really like about this update, check this:

I have cheated add_latest_equipment to AI and it has everything. So I overflowed the stockpile value, that it went negative, so this is day after:

It puts all missing equipment to the queue! Great stuff. Also, it started to rearrange queue, but everything is missing so it went just back and forth.

My conclusion is that there is place for AI template improvement. We just need to nudge developers to take this as one of the priorities

 

[seattle]

MIN_FIELD_STRENGTH_TO_BUILD_UNITS = 0.7,            -- Cancel unit production if below this to get resources out to units in the field
MIN_MANPOWER_TO_BUILD_UNITS = 0.7,                    -- Cancel unit production if below this to get resources out to units in the field
MANPOWER_FREE_USAGE_THRESHOLD = 500000,            -- If AI has this much manpower he doesn't care about the percentage
MANPOWER_RESERVED_THRESHOLD = 0.25,                    -- The AI will not deploy more units if he goes below this percentage

Notice the part of the free manpower usage, if I'm correct.. as long as the AI has no less than 500k manpower. it will keep producing units and ignore the 25% of maximum manpower threshold.

Let's say I have 10mil recruitable manpower as the soviet union with a 25% reserve treshold ( that's 2,5 mil), it doesn't matter because the defines lua says to keep building units as long as you're above 500k manpower. Assuming I'm understanding this correctly.

Have you already found out how to read those numbers?
Ideally I would like to stop a.i. unit production at reasonable levels. ==> raising reserves

Let's say I'm using these numbers:
MIN_FIELD_STRENGTH_TO_BUILD_UNITS = 0.75, -- Cancel unit production if below this to get resources out to units in the field
MIN_MANPOWER_TO_BUILD_UNITS = 0.75, -- Cancel unit production if below this to get resources out to units in the field
MANPOWER_FREE_USAGE_THRESHOLD = 1500000, -- If AI has this much manpower he doesn't care about the percentage
MANPOWER_RESERVED_THRESHOLD = 0.25, -- The AI will not deploy more units if he goes below this percentage

That would make a big difference for manpower-heavy nations like Germany and Soviet Union, right? Small nations wouldn't get affected by the FREE_USAGE value and instead use the RESERVED value.

--------------------------

I had some good experiences with adjusting the conscription laws. In vanilla, fascism and communism can simply proceed to a high level without being at war. I adjusted the middle levels so that Germany can't go ape on conscription early on. Didn't harm Germany's performance one bit. It does slow down unit production though!

-----------------------------------------------

@Nobody987
One suggestion: Your cavalry template is in my opinion to big. I assume that most players only use something like "2-4 cav + mp support" as a partisan suppression template. Using cav as a combat force would be bad in most cases. Increasing the cav template will lead to wasting experience in order to shrink it again to reasonable levels.

 

[Nobody987]

One suggestion: Your cavalry template is in my opinion to big. I assume that most players only use something like "2-4 cav + mp support" as a partisan suppression template. Using cav as a combat force would be bad in most cases. Increasing the cav template will lead to wasting experience in order to shrink it again to reasonable levels.

I agree, it is bad for combat in any case. Currently in mod most of nations start with 2x5 CAV, so generally AI shouldn't waste xp on changing template. I will take a look and test it a little bit to see how it works.

 

[seattle]

I agree, it is bad for combat in any case. Currently in mod most of nations start with 2x5 CAV, so generally AI shouldn't waste xp on changing template. I will take a look and test it a little bit to see how it works.

I fiddled around a bit with the template thresholds in defines. I tried lowering the threshold for creating templates, but the a.i. is stupid that way. Germany used the experience to create a heavy armour template in early 1936. That wasn't intended on my part.

I'm trying my luck now with a combination of increased conscription requirements and increased thresholds for producing new units, alongside 30-40 width target templates.

 

[seattle]

Keep in mind that a lot of mods are using "workaround" to fix problems.
Paradox can't really do that - we actually have to fix stuff "properly" - and If it would have been that easy it would already have been implemented.

Ahem... from the 1.3.1 thread:
TomJones:
"...it looks like no AI bothers to make more than 3-4 tank divisions for their 200+ division armies, so not all is roses"
Podcat:
"yeah known. they are very sparse on armor. Its something for future tweaking. AI does its best job with strong ifnantry anyway though"

Wouldn't you call rendering WW2 into WW1 and justifying it with the a.i. not being able to utilize tanks properly anyways - after 3 patches and 1 expansion - a 'workaround to fix problems'? Something that 'Paradox can't really do'?

 

[GAGA Extrem]

In this case it's a mix of the AI template problems and balance.

The key issue is that armor - particularly LARM - is pretty bad (in terms of "bang per buck") when not used with proper support models (SP-ART/TD). That's one of the reasons why we increased hardness on all models (and trucks) and moved all MECH techs by 1 year.

Before we tweaked the AI "tank happiness", the GER AI would gladly build up to 50 LARM divisions until Danzig, which - as you can imagine - isn't exactly the greatest idea. Even less so when you just taught the other AIs how to use AT. So the choice was to use THAT specific AI or use one which plays it "safe" in terms of MIC investment in armor.

The thing is it was never intended as a fix for the underlying issue, it's more of a bandaid really. But I suppose that would be arguing semantics.
Point is: Podcat knows that the whole thing needs a lot more wörk - and he is on the job!

 

[Napoleonetniet]

funny i havnt actually changed a thing yet in 1.3.1b, but seems like this 2nd game im already seeing way less divisions.

Its often verry complicated how the game works and what ties into what.. for instance it could well be MY OWN FAULT i saw so many divisions from germany when i played ussr, because i was trying to get a load of units on the border to avoid germany cancelling non-agression pact earlier than i liked.
Now i think that experiment might as well have BACKFIRED at me.. the AI scraping all units it could get its hands on together just to be able to break the treaty?
Actually sounds verry viable to me and more and more likely too

now im playing Germany to test-try some other stuff, actually havnt even put ONE unit at the USSR border, the closest few units are 2 provinces away, has been like that since i gave them half of poland (mrp). And to my surprise Aug.1940 , USSR has 201 (191 + 10 vol. in china) divisions. And ìs about to go for Finland - I have about 190 div's including my portguards and suppression units.

still have to see where this goes in 1941 and 42 but so far it already looks alot better than my other game. Anyway I wont change a thing until i've done 10-20 games and a couple of handsoff runs to get a better overall picture of the current state

its way too easy to change too much , and even 1 change can cascade into unbalancing multiple things, or give unexpected results.. taking you further from where you want to go than you started out with

 

[seattle]

In this case it's a mix of the AI template problems and balance.
The key issue is that armor - particularly LARM - is pretty bad (in terms of "bang per buck") when not used with proper support models (SP-ART/TD). That's one of the reasons why we increased hardness on all models (and trucks) and moved all MECH techs by 1 year.
Before we tweaked the AI "tank happiness", the GER AI would gladly build up to 50 LARM divisions until Danzig, which - as you can imagine - isn't exactly the greatest idea. Even less so when you just taught the other AIs how to use AT. So the choice was to use THAT specific AI or use one which plays it "safe" in terms of MIC investment in armor.
The thing is it was never intended as a fix for the underlying issue, it's more of a bandaid really. But I suppose that would be arguing semantics.
Point is: Podcat knows that the whole thing needs a lot more wörk - and he is on the job!

Fair enough! Nobody should expect such a complex game to work perfectly.
There is still no reason for the a.i. unit spam. I guess most of us can live with a temporary lack of tanks, but not with a horrible framerate.

Here's what I've found out after numerous trials since yesterday:
- I increased target templates to 30-40 width
- changed Germany's 1936 start template for infantry to 12 infantry and 2 arty (which is equal to the target template)
- increased requirements for economic and conscription laws ==> Germany doesn't go total eco mobilization in 1939, fascists and commies don't go to unnecessarily high conscription laws early
- Increased thresholds for unit production so that the a.i. stops production of units earlier

Effects:
- Spanish Civil War, Japanese-Chinese war as usual
- German a.i. vichied France in December 1939 despite having way less units
- Germany and Soviets have 400 units in January 1941 (compared with 600-700 German units in 1.3, don't know about 1.3.1)
- the game still runs lightning fast in 1942 whereas vanilla 1.3.1 started to bog down to a halt in late 1940

Now I ask you: What's the downside of that quick modding job? I didn't use any workaround. I wouldn't even know how to script events.

 

[Kadanz]

Yeap. I'm noticing a lot of improvements in AI behaviour with 1.3.
I'm doing the tweaks from scratch, see where it takes me.

Let me know what your findings are. I'm doing the exact same thing.

 

[yonderTheGreat]

If I may ask (I'll check when I get home), how come none of the threads complaining about the "unrealistic" number of divisions give info on what the divisions consist of? Are they tiny, 6-brigade divisions or massive 15-brigaders? When quoting how many divisions the Germans had at such-n-such time in the war, it MATTERS how big said divisions were. In reality, Germany had a greater NUMBER of divisions on the Eastern Front in January of 1945 than they had in July of 1941... but they sure as heck didn't have more MEN.

Please... please... please... give info on the total number of brigades and men. I don't understand what people don't understand about the concept of "there's more to divisions than simply the number of them." 100% of the time when I've played and the total number of divisions has gotten peculiar in any way, I check the make-up of the divisions and instantly I understand "oooooh, they're crappy, skeletal divisions... that makes perfect sense."

And, again, I"'l check when I get home. I'm just seeing if anyone has done the work for me ahead of time

 

[Kadanz]

Have you already found out how to read those numbers?
Ideally I would like to stop a.i. unit production at reasonable levels. ==> raising reserves

Let's say I'm using these numbers:
MIN_FIELD_STRENGTH_TO_BUILD_UNITS = 0.75, -- Cancel unit production if below this to get resources out to units in the field
MIN_MANPOWER_TO_BUILD_UNITS = 0.75, -- Cancel unit production if below this to get resources out to units in the field
MANPOWER_FREE_USAGE_THRESHOLD = 1500000, -- If AI has this much manpower he doesn't care about the percentage
MANPOWER_RESERVED_THRESHOLD = 0.25, -- The AI will not deploy more units if he goes below this percentage

That would make a big difference for manpower-heavy nations like Germany and Soviet Union, right? Small nations wouldn't get affected by the FREE_USAGE value and instead use the RESERVED value.

MIN_FIELD_STRENGTH_TO_BUILD_UNITS = 0.75 <-- I'm still not sure what this one does.

But the other 3 seem to be doing exactly what I thought they would. I combined it with doing heavy cuts in the manpower laws, modified the AI template builds to 30 instead of 20. I changed the manpower free usage to 1,5 mil

Vanilla 1944
soviet union

654 divisions
Germany
531 divisions
USA
614 divisions
UK
346 divisions
JAP
307 divisions
ITA
284 divisions

Modified 1944

SOV
324 divisions
GER
226 divisions
USA
305 divisions
UK
187 divisions
JAP
166 divisions
ITA
214 divisions *I kept the italian divisions smaller (20 width) for historical reasons

 

[amalric de g.]

Ahem... from the 1.3.1 thread:
TomJones:
"...it looks like no AI bothers to make more than 3-4 tank divisions for their 200+ division armies, so not all is roses"
Podcat:
"yeah known. they are very sparse on armor. Its something for future tweaking. AI does its best job with strong ifnantry anyway though"

Wouldn't you call rendering WW2 into WW1 and justifying it with the a.i. not being able to utilize tanks properly anyways - after 3 patches and 1 expansion - a 'workaround to fix problems'? Something that 'Paradox can't really do'?

In my current Germany campaign, i tagged the SOV to see what they build and how big the Equipment hole is. They had 30 medium tank Divs in the qeue and they build upgraded T34 with two full lines.

 

[seattle]

MIN_FIELD_STRENGTH_TO_BUILD_UNITS = 0.75 <-- I'm still not sure what this one does.

But the other 3 seem to be doing exactly what I thought they would. I combined it with doing heavy cuts in the manpower laws, modified the AI template builds to 30 instead of 20. I changed the manpower free usage to 1,5 mil
Modified 1944
SOV
324 divisions
GER
226 divisions

That are fantastic numbers. That's pretty much how it should be. Both in terms of gameplay and performance. Did the a.i. perform any worse than in vanilla? If not, then the "more is better" philosophy gets even more absurd.

Make sure to upload that mod!

 

[seattle]

In my current Germany campaign, i tagged the SOV to see what they build and how big the Equipment hole is. They had 30 medium tank Divs in the qeue and they build upgraded T34 with two full lines.

It seems that PDS changed it so that nations only start building tanks once their army is huge. And then they go ape on tanks.

 

[Ancalimé]

MIN_FIELD_STRENGTH_TO_BUILD_UNITS = 0.75

If the overall division strength as a whole in the field is under that value, the AI will stop building new divisions and try to equip those that are already deployed.

Vanilla 1944
soviet union

654 divisions
Germany
531 divisions
USA
614 divisions
UK
346 divisions
JAP
307 divisions
ITA
284 divisions

Modified 1944

SOV
324 divisions
GER
226 divisions
USA
305 divisions
UK
187 divisions
JAP
166 divisions
ITA
214 divisions *I kept the italian divisions smaller (20 width) for historical reasons

Are these numbers with the expeditionary disabled?

 

[seattle]

If the overall division strength as a whole in the field is under that value, the AI will stop building new divisions and try to equip those that are already deployed.

I use the 0.75 threshold as well and it's probably still way too low. Watched a.i. Germany against a.i. Soviet. Germany was able to keep up with the losses and still produce a ton of units. The Soviets were down to 0.83 ratio and also produced like crazy (while lacking some 80,000 rifles...).
I don't see why the a.i. should produce 100 new units when it lacks so much equipment.

Vanilla even uses a 0.7 threshold which is totally ridiculous if I'm not nuts. I wouldn't be surprised if values like 0.85 are much better than everything we and PDS have tried.

 

[Ancalimé]

I use the 0.75 threshold as well and it's probably still way too low. Watched a.i. Germany against a.i. Soviet. Germany was able to keep up with the losses and still produce a ton of units. The Soviets were down to 0.83 ratio and also produced like crazy (while lacking some 80,000 rifles...).
I don't see why the a.i. should produce 100 new units when it lacks so much equipment.

Vanilla even uses a 0.7 threshold which is totally ridiculous if I'm not nuts. I wouldn't be surprised if values like 0.85 are much better than everything we and PDS have tried.

Yeah, well, moving the limit higher doesn't work quite well for China. They already only have a handful of troops at the start of the game and they have an equipment deficit. If you move it higher, you're effectively making it impossible for China to deploy any new divisions.

I guess you could make it so that they have an equipment surplus at the start of the game and then move it higher.

 

[seattle]

Yeah, well, moving the limit higher doesn't work quite well for China. They already only have a handful of troops at the start of the game and they have an equipment deficit. If you move it higher, you're effectively making it impossible for China to deploy any new divisions.

I guess you could make it so that they have an equipment surplus at the start of the game and then move it higher.

Hmm, but isn't a.i. China supposed to be on the losing side anyways? Plus, it's only one country and you could find a workaround (like you mentioned) should it indeed only affect China. IMHO the major nations are way more important.

What's your assumption on what a higher value like 0.8 or even 0.85 would do for Germany and Soviet?