(1.3.3)Is new method for resource shortages the underlying cause of German AI problems?

[Dalwin]

I believe it is. The new system is much more painful for major powers if a shortage is severe. If Germany is not played with a decent dose of foresight the shortages of oil, rubber, and tungsten can be severe.

This leads to them not having enough forces. This in turn means that after the fall of France they do not have enough to accomplish their three main tasks at the time, which are guard Atlantic Wall, suppress France and Benelux, Barbarossa.

My proposed solution is to rollback to something that is similar to the original method of calculating shortages, but which accomplishes one of the stated goals of the new system, that being eliminating the 90% cap on penalty.

The old system took the shortage and prorated it against demand to calculate the penalty and did so with a relatively mild cap for each involved resource. If something only used one resource it was prorated against a 90% cap, for two each was able to give a maximum of 45% penalty, and for three resources the max penalty for each was 30%.

I propose changing that to a target cap of 120%, meaning that the maximum for 1/2/3 resources being used is 120/60/40 percent of penalty for each resource per line. The reason to target 120% instead of 100% is so that in situations where you still have some of what is needed it is not so generous in the face of the completely missing ingredient(s). Before, if I had enough of one ingredient but none of the other two on something like medium tanks I would still get 40% production from the line. Under my proposal that would be 20% instead.

 

[Pro_Consul]

From looking at the near-ARR posted in another thread, and based on the fact that AI Germany's problem date back long before 1.3.3, all the way to the original release of HOI4, I would say that your conclusion that the change in shortfall calculation is the source of AI Germany's problems is not supported. You have been an active poster in that thread, so you might want to look at posts prior to your first reply there.

Example: Germany is seen to perform textbook invasions of Poland and France, knocking both out efficiently and more or less on schedule. It then properly masses it forces on the Eastern Front and launches Barbarossa. Then, for reasons which mock the concept of sanity, it strips the Eastern Front of most of its divisions, sending a few to Yugoslavia and the rest to Africa. That is NOT a problem with the production system or a lack of divisions. It is a psychotic employment of the divisions it had.

That said, the production issue still may be a problem in itself that bears investigation, testing, and then perhaps some tweaking to resolve it.

 

[ringworm]

Pro, I am not so sure, I played only 2 games with the new patch but both have seen early loss for axis.

Germany could not even take the Netherlands and Belgium. So this is before things go crazy with redeployment.

 

[Cooling Machine]

1.3.3 messed up AI front management, that is Germany's biggest problem

 

[grandad1982]

1.3.3 messed up AI front management, that is Germany's biggest problem

Those screens make me sad. Seen that happen on a smaller scale with my own battle plans when they don't get distributed well.

 

[TheOrangeGuy]

1.3.3 messed up AI front management, that is Germany's biggest problem

God, I can just imagine that half of them will do a baltic cruise soon as well

 

[Gwydion5]

1.3.3 messed up AI front management, that is Germany's biggest problem

Are you using mods? One thing to consider is that the AI will at times do something, and then the next day or two later correct it. I know ideally it shouldn't do what you posted in your screenshots anyway. But if you select the divisions and then observe over a few days it might give more insight.

The new system is much more painful for major powers if a shortage is severe. If Germany is not played with a decent dose of foresight the shortages of oil, rubber, and tungsten can be severe.

This leads to them not having enough forces. This in turn means that after the fall of France they do not have enough to accomplish their three main tasks at the time, which are guard Atlantic Wall, suppress France and Benelux, Barbarossa.

A good way to test this would be to give AI Germany in 1936 a few refineries and see if it changes things. That was the purpose of my 1939 and 1940 human played games and then handing control over to the AI to see how it would do with a better economy and long term production goals / surplus to equipment. Also if the AI is consuming say more tungsten then it is using and it does not make adjustments to its production to put balance that, wouldn't that be the thing to fix instead of the deficit penalty?

Then, for reasons which mock the concept of sanity, it strips the Eastern Front of most of its divisions, sending a few to Yugoslavia and the rest to Africa. That is NOT a problem with the production system or a lack of divisions. It is a psychotic employment of the divisions it had.

I'm curious if anyone has downloaded the pictures zip that I posted in the bug report forum. Looking at the post it shows zero downloads for any of the files. But there is a bunch of screenshots capturing theater/army/division realignment. AI_TEST-2 has screenshots from March 20th, to April 20th where the AI is basically doing nothing but shuffling and reorganizing army's on the eastern front while at war with Russia. That's pretty much an entire month where AI Germany is doing nothing but harm while at war with Russia.

At first with the garrison changes, I figured maybe it was that. But even if we assume AI Germany is doing a garrison realignment, it would not explain or justify pulling 200 divisions off the Eastern Front to do so, only to send them back. I actually show this in 4 screenshots on the bug report post, though I captured a day or two after it started. Basically February 13th is my first screenshot showing all (There was about 180-200 divisions originally) but 8 divisions being pulled off, and then on February 17th it's sending 150 divisions back. Like I said there is at least a day or 2 that I missed when it started.

 

[P3D]

Solutions:
a/ Change the second German industry NF from CIC to Refineries.
b/ start Romania on Free Trade, make sure they do not switch back, and that Germany would prefer to trade with them.

 

[Meglok]

1) the new penalties for lack of resources are more harsh than expected (100% shut down of lines) and the AI has issues dealing with them.
2) the overseas trade penalty is too harsh, Japan especially is crippled by it
3) the front management issue still has not been even remotely fixed.
4) the African situation is again a joke, far too many divisions are being sucked down there.
5) the AI does not seem to reliably receive offers of expeditionary forces from minor allies.
6) Axis minor participation is iffy at best
7) production strategy does not match the new industry code or resource availability.
8) template coding does not match industry capability or resource availability, nations are trying to build divisions they have no hope of equipping.

Those are what I have seen in my own tests and from posts from others. It is just not always one thing, but a mixed stew of problems. Some can be overcome by sliders, some like the front management shuffle and Africa are frustrating enough to make your hair hurt.

 

[Cooling Machine]

Are you using mods? One thing to consider is that the AI will at times do something, and then the next day or two later correct it. I know ideally it shouldn't do what you posted in your screenshots anyway. But if you select the divisions and then observe over a few days it might give more insight.

I do use Expert AI 2.0 and the Africa addon mods, without those i refuse to play this game.
Mods can't touch the AI front management, which is up to paradox to fix so it's unlikely that this situation is caused by my mods.
I ran 3 tests with Expert AI 2.0+Africa Addon when 1.3.3 came out and the AI's horrible frontal management was present in all of them.

 

[Gwydion5]

I do use Expert AI 2.0 and the Africa addon mods, without those i refuse to play this game.
Mods can't touch the AI front management, which is up to paradox to fix so it's unlikely that this situation is caused by my mods.
I ran 3 tests with Expert AI 2.0+Africa Addon when 1.3.3 came out and the AI's horrible frontal management was present in all of them.

I understand, but it's always worth mentioning up front, just so people are aware. Also being in IT for 20 years, I know that things that should not be, sometimes can be for things or ways we wouldn't expect. With that said, I have watched about 12 or so AI Germany games and have not seen that kind of pile up. So I will make a point to look closely at the lines and see if that is happening for me as well.

 

[Nevrion]

1.3.3 messed up AI front management, that is Germany's biggest problem

WTF?! If this is a common issue, we urgently need a patch to 1.3.4. before 1.4.

 

[Gwydion5]

1) the new penalties for lack of resources are more harsh than expected (100% shut down of lines) and the AI has issues dealing with them.
2) the overseas trade penalty is too harsh, Japan especially is crippled by it
3) the front management issue still has not been even remotely fixed.
4) the African situation is again a joke, far too many divisions are being sucked down there.
5) the AI does not seem to reliably receive offers of expeditionary forces from minor allies.
6) Axis minor participation is iffy at best
7) production strategy does not match the new industry code or resource availability.
8) template coding does not match industry capability or resource availability, nations are trying to build divisions they have no hope of equipping.

That's a very nice summary. But you forgot one.

9) AI does not utilize air power efficiently (most wings being assigned) or optimally (mission areas that make sense).

Default 1939 start as UK. Added 1m manpower, 7k level strat bombers, level 6 radar in southern England, Gibraltar, Malta and Cyprus, as well as level 10 air fields. This is around 3 weeks in to the war. Plenty of time for the AI to assign these wings. Also while it might say 750 - Lorraine - East Germany. The reality is at least 1 wing is operating like that from the base in that region, the rest might just be on standby or operating in a different air zone. So you need to look at the map to see actual strategic bomber allocation.

10 Days later, still generally the same.

10 days after that, still relatively the same. If you add up what we see for Strat Bombing in the air zones, it's around 1500-2000 Strat Bombers operating in Germany / Italy. Granted, UK would likely never have 7k strat bombers at this point. But still relevant in that we are seeing issues with air power utilization and this evidence proves that the AI does not scale well with large air forces.. It's strange because I remember when the AI was pretty good about using it's air force. Like the Allies with the Southern England, English Channel, and Northern France, where you had thousands of fighters in various air zones.

 

[Meglok]

@Gwydion5
Actually i intentionally avoided issues of air power as those are slated for 1.4. There are several concerns there from region choice to wing size to production. There is the still unresolved issue about the AI refusing to produce anything but tier 1 CV fighters no matter the year or known tech for example.

I also didn't include anything about naval since that is apparently now a 1.5 project. @Axe99 and I were just discussing the issue with the AI setting fleets to only retire at heavy damage and no split off of damaged ships causing AI fleets to be destroyed too easily.

I suspect @SteelVolt has most of these on his little list.

 

[Cooling Machine]

I understand, but it's always worth mentioning up front, just so people are aware. Also being in IT for 20 years, I know that things that should not be, sometimes can be for things or ways we wouldn't expect. With that said, I have watched about 12 or so AI Germany games and have not seen that kind of pile up. So I will make a point to look closely at the lines and see if that is happening for me as well.

I did a test without the Expert AI or Africa Addon mods and here is are the results

The overstacking into one province is not a problem this time around, i guess the reason for this is because without Expert AI 2.0 the AI have less units to do dumb things with and without the Africa addon(impassable Sahara) they are sending their armies to fight ww2 in the Sahara.Meanwhile Germany is leaving large parts of the eastern front undefended.
1.3.3 made the AI front management nearly as bad as it was in the release version.SAD!

 

[Dalwin]

@Gwydion5

I did read your test in the other thread. You running Germany until mid '39 and then handing it off to the AI is an interesting test and it reveals some important information. I do not, however, think it solidly supports your conclusion in that thread that the problem is mainly one of front management. Front management is indeed a problem, there is no doubt about it. As pointed out above by @Pro_Consul that problem predates 1.3.3, likely goes back to 1.0 and has had only minor improvement at a couple of points in the game's post-release progress.

@Meglok 's statement that the current problem is due to a stew of problems all contributing is obviously correct.

Now let me try to support why I still feel that the contributing factor in the OP is still one of the major ones and one that we would benefit from changing. In the test of handing off after human control up to mid '39, what was missing was a detailed comparison of what Germany would have had if human controlled to that point before 1.3.3. In my current game as Germany, I am obviously not collapsing and dying before the end of '42, but I am struggling noticeably compared to earlier versions. Even with great effort to obtain sufficient oil ( a couple of refineries by NF would barely scratch the surface) I could only raise a decent size land force by letting myself be greatly inferior in the air.

At times I am quite frustrated with the developer statement to the effect that the AI needs to have lots of units (effectively more than it can supply) to be able to play effectively. At the moment I think we are seeing evidence of that. There is a critical mass in the area of army size that Germany must achieve between the fall of France and Spring '41. They are currently failing to do so. Another contributing factor is that in the past the AI would ignore suppression. It still is unable to deal with it but the fact that it is now on its list of tasks only makes its confusion worse when it has more things needing to be done than it has units to do them.

Someone in another thread suggested that Germany would be helped by having specialty templates for smaller and cheaper to build units for the purpose of coastal garrison and suppression. I agree with that. Though it is more than just having the templates to fix that aspect. The AI also has to produce those second rate units in quantity and has to deploy them to the proper areas.

As a final point supporting my premise that Germany fails because of lack of reaching critical mass in its land force, we have Meglok's observation concerning expeditionary forces and overall participation from its minor allies, specifically Hungary and Rumania. In my human run Germany game, I still have the entire Rumanian army sitting inside their borders and they have offered no expedition. The Hungarians, at least offered my roughly 90 divisions. This offer seems to not be happening when Germany is AI run. Someone else in a different thread observed that when he played Italy, the expeditions were offered to him instead of to Germany. I believe that sets the precedent for expedionary offer behavior being different depending on who is human played.

As to my use of those Hungarians, one army of 24 took up a sector on the Russian Front which allowed me to condense the German units there somewhat and consequently push the front forward a few provinces. I still have not achieved historical progress (though I am playing with Russia at +3 on slider). I have played with Russia that high in earlier versions and walked straight over them, however. The bulk of my Hungarians are garrisoning the west. Even with that, some of the many invasions are making it ashore. If I did not maintain one of my three mobile armies spread out to various locations in the west, I would have trouble throwing those incursions back into the sea.

I did not use the sixty-odd Hungarian units sent west to free up Germans there to send east. The defenses had been stretched far too thin so that I could launch Barbarossa. I had also (in spite of producing more cavalry than I usually did in the past) been occupying many nations at less than the harshest policy. The Hungarians let me crack down and increase my occupation level.

This is something that had to be done gradually even with the sufficient force there because having the extra factories without having more oil did not help me to make more of what I really needed.

I am also finding myself in that game struggling with shortages of many key equipment items. This is new for me in 1.3.3 and I know that if I struggle with it then the AI will fare worse. The only thing I have in excess is infantry equipment (since steel never runs short). I give much of this to Italy and Hungary.

 

[Pro_Consul]

1.3.3 messed up AI front management...

Not by a long shot. Germany's front management has always been terrible, and in fact is better now, to a point, than it was before. That "to a point" would be up until Barbarossa, at which point it loses its mind once more, relapsing into the same psychotic behavior pattern it has had since 1.0.

...that is Germany's biggest problem

Now with this I agree completely. Its other problems, and Meglok was right to say there are quite a few of them, each pale before this one game-breaking behavioral disorder. It pretty much makes Germany the only worthwhile country to play, since AI Germany always implodes shortly after finishing off France rendering the rest of the game an exercise in absurdity.

 

[Gwydion5]

I did read your test in the other thread. You running Germany until mid '39 and then handing it off to the AI is an interesting test and it reveals some important information. I do not, however, think it solidly supports your conclusion in that thread that the problem is mainly one of front management. Front management is indeed a problem, there is no doubt about it. As pointed out above by @Pro_Consul that problem predates 1.3.3, likely goes back to 1.0 and has had only minor improvement at a couple of points in the game's post-release progress.

I ran 1 game till 1939, saved, handed control the AI, it did really poorly. I then loaded 1939, kept control until august of 1940, then saved, and handed over control to the AI and it did much better. Which I posit is probably because my production lines weren't destroyed in 39, allowing more of my hard work and planning to come to fruition in terms of equipment. I have also tried to say often that the issues observed are multifaceted, and that I have a hard time narrowing it down to one thing. Rather a lot of my tests are trying to rule out and narrow down the scope. If my posts seem to infer singularity, just ignore it, unless I specifically say it is. I'm just trying to draw attention to various observations and provide awareness I guess to issues that may or may not be considered.

 

[Dalwin]

I ran 1 game till 1939, saved, handed control the AI, it did really poorly. I then loaded 1939, kept control until august of 1940, then saved, and handed over control to the AI and it did much better. Which I posit is probably because my production lines weren't destroyed in 39, allowing more of my hard work and planning to come to fruition in terms of equipment. I have also tried to say often that the issues observed are multifaceted, and that I have a hard time narrowing it down to one thing. Rather a lot of my tests are trying to rule out and narrow down the scope. If my posts seem to infer singularity, just ignore it, unless I specifically say it is. I'm just trying to draw attention to various observations and provide awareness I guess to issues that may or may not be considered.

You seem to be saying here that the heart of it lies with production. To that I would agree completely.

What has recently changed in production? There are two obvious elephants in the room, the infrastructure change to construction and the new method for coping with resource shortages. I am obviously blaming the latter though I would concede that the former causes some delay in buildup and should consequently cause Germany to delay its push east. In my few games of 1.3.3 it seems that AI Germany starts Barbarossa between January and March of '41. They might be better with a more historical May or June kick-off.

 

[Gwydion5]

You seem to be saying here that the heart of it lies with production. To that I would agree completely.

What has recently changed in production? There are two obvious elephants in the room, the infrastructure change to construction and the new method for coping with resource shortages.

There is also the issue of how it manages production lines, and production goals. One thing I see consistently in 1.3.3. is that the AI production targets seem like they are optimized for maximum efficiency/utilization, in that it almost never has surplus, and if current production can't met current demand then it falls behind.

For example, here is the division overview with equipment for a Max Boosted AI Germany at the end of 1941.
View attachment 243511

And here is the logistics screen
View attachment 243506

( Bonus points : Now for comparison, load up my 1939 save and hand over control to the AI, and then load up the 1940 save and do the same, look at the differences of those screens at the end of 1941 assuming/forcing war in the 1st quarter of 41.) I just don't see how AI Germany can win like that, or be competitive with equipment shortages like that. It can't all be resources either, Tungsten shortages make sense, but Infantry and Support Equipment? It doesn't add up as a resource shortage issue. Then you have the AI wiping out and removing MIC's from the most productive lines. I mean straight up undeniable that if there are two lines of say artillery, the 1st line has production efficiency of 15% with 4 factories, the second line has 55% with 15 factories, and the AI wants to realign an MIC, it will often take from the 2nd artillery line, in some cases it even gives it to the 1st artillery line.

Anyway I need to get going and this reply is rushed. So if I get anything wrong or anything comes off rude, please forgive me, it is not my intention.

Also if the screenshots down't show up above, here is the post that has them.
https://forum.paradoxplaza.com/foru...-loses-the-war-in-1941.1001605/#post-22473039