A pattern to the current German collapse (and probable cure in sight)

[Dalwin]

In the past we have discussed some of the reasons for the current German AI collapse which typically takes place around the end of 1941. Some of that has been centered around changes in 1.3.3 regarding economics or how the Russians deploy their units etc.

This time I want to take a slightly different tack to it and illustrate why I think the items we have already seen in the limited amount that has been revealed about 1.4 may include the fix.

The Pattern I have noticed now across several games is that the northern part of the front will immediately stall in the vicinity of Memel. The Center often reaches the outskirts of Kiev and the south sometimes makes it past Dnepropetrovsk. As the center and south advance, German lines become longer.

At some point the unit controller decides that it needs to reinforce some thin spots in those sectors which inevitable results in pulling too many from the north and leaving fatal gaps. This collapses so quickly that even before units get sent back north, the Russians have taken all of east Prussia and often Danzig and Warsaw as well. Worse, the shape of the new line makes it even longer thereby increasing the German overextension which led to the collapse.

Two things that have been revealed about 1.4 should help with this. First, the revealed changes to the unit controller which have shifted units taking shorter trips and re-evaluating periodically to make sure their destination is a current need, should make the effects of overextension less severe.

Second, it has been mentioned both by @Meglok (and other players) as well as @podcat that in 1.3.3 Germany's allies are not sending expeditionary forces as often as they used to. Increased expeditions should also reduce the chance and impact of gaps caused by overextension.

Unless I am reading the pattern wrong, those two things by themselves should cure what is currently ailing Germany. (At least cure the current biggest problem. There are of course others remaining on the list.)

 

[Axe99]

Both good points . I've only done a handful of hands-off games, but also noticed that in those games Germany didn't concentrate force as effectively in the east as I'd like (something impassable terrain in Africa should help at least), so that may be another factor (or my hands-off games were just unlucky). I definitely expect the changes announced to help offensive nations on long fronts though (Japan in China as well, which also suffers unit-shuffling pretty heavily).

 

[Dalwin]

Well even my SP games include No Mans Land, so I expect no relief from that change in the main game.

 

[Meglok]

Unless I am reading the pattern wrong, those two things by themselves should cure what is currently ailing Germany. (At least cure the current biggest problem. There are of course others remaining on the list.)

You are forgetting this

4) Unit usage such as garrison and template and production management were far from optimal.

These two in my opinion could be a huge improvement, depending upon how they are handled.

Let's deal with the easy one first. The improvements to garrisons in 1.3 were a big step forward. The problem now is that while the ai will assign units to stop mostly stop resistance it will shuffle and steal them away a the drop of a hat. The actual assignment priorities and what the units are assigned to garrison needs improvement. If the ai makes better choices on what it uses for suppression (for god's sake change the block on making cavalry and learning MP tech) and then keeping the units on site this will really help Germany. It does no good for ai Germany to capture the West only to see everything go up in flames due to resistance.

Template and production management fixes could be huge. Every nations starts with generic templates and production scripts. Some majors have modifications but for the most part every nation is coded to eventually make the same division templates given enough time. And for the most part these templates are based on what is needed/was used in the Western European conflict. So you have infantry with a heavy dose of artillery, AT, and support companies, and mobile units that all loaded with Sp Art etc etc.

This of course leads to huge production issues as being able to produce all of the equipment the generic templates is beyond most nations besides the Big 4, Germany, Russia, the USA, and the UK. And the actual template choice made by these less capable nations makes it even worse. These generic templates add production calls that are beyond the MIC and resource capabilities of these smaller nations.
Examples from my current game.
1) Japan has 20 motorized inf with 8 MOT/3 MSpArt and ART, AT, ENG, and Recon support and more on the queue. While that might be a great template for the Ukrainian steppe it sucks in Asia and is a huge MIC drain. So Japan has no MSpArt and these divisions are deployed short over 1600 guns now.
2) The current ai meta for Inf templates appears to be 7 INF/2 ART with ENG, recon, ART, and AT. Once again, a decent division for fighting in France or Russia but in the jungles of Asia not so much, especially if you are MIC challenged. And AT is almost useless, unless fighting Russia Japan will not see many tanks.. Japan has 160 of these, plus more in the queue.
3) Marine divisions would be what you would expect to see for Japan, but she only has 9. The generic scripts limit the number of marines and mountain units to a small ratio of overall units iirc.
4) Just to show this is not limited to Japan, Hungary is also producing the same meta inf template, and is down on AT, ART, and support eq. It is incapable of both fulfilling replacement equipment required by this template for attrition and battle losses and meeting new deployment requirements. Hungary has produced far more divisions than it can possibly equip and support. The ai appears to just look at available manpower when making new divisions. It does not appear to consider whether it can actually produce the required equipment, nor does it consider whether it can replace battle losses.
5) If that wasn't bad enough, Hungary has in it's deployment queue a motor template, a mech template, and a light tank template. She only has 24 MIC, this is early 1942.

In 1.3.2 these issues were not quite so bad. It was annoying but the ai could still pump out some production. In 1.3.3 these issues are glaring due to the penalties for not having resources.

If PDS goes to more appropriate templates for each major nation's geography, needs and capabilities then this will be a HUGE improvement. For minors they could at least break the template designs into tiers based upon current amount of MIC.

Whatever they do they have to code production and division deployment so that it takes into account deploying numbers of units the nations can actually support. Hollow shells do nothing but eat supply and become speed bumps. This is Germany's issue. Both Germany and it's allies can not support what the templates are calling for and the numbers of deployed divisions. By 1941 they are a bunch of hollow shells and get run over.

It all depends on how Podcat's statement is addressed in the upcoming DD.

A lot of Japan's issue would be helped if they would Add Aluminum in Asia.

 

[Barbu]

I see none of these issues using expert ai 2.0 so it would be interesting if chondrite (the mod designer) could chime in.

 

[Dalwin]

I am not disputing any of those points. I do, however, fail to see their direct connection to why Germany falls flat in the same manner at the same time pretty much every game. They always collapse in the northern part of the Russian front after advancing in the south and center.

I recently played a game as Russia. In the hope that the Germans would be able to put up a fight I played on veteran and gave them +4. I also deployed more units on the border than I normally would but had a strong second line at the usual spot and some strategic reserve armies in addition.

Even with all that, they failed to push in the north, but I could not really stop them from pushing in the south and center. About the time they were getting close to Kiev, they pulled most of the units away from the north. They only abandoned half of the provinces and only those for a day or two at most. It was still enough to send them into a death spiral from which they could not recover.

I only ever used my border guards. I kept my second and third lines back in their positions, though I did commit one 12 division army of reserves. I was in Berlin by Christmas of '41. This was not because I played well. I went out of my way to handicap myself. It was not for economic reasons since they had much more than they normally would. It is because even with more units than they can currently field in '41 they cannot distribute them properly along the line.

It is not merely the fact that they shuffle. It is the way they go about it. The AI sets up one large army along a wide front. It has no hesitation to send a unit from 4 or 14 provinces away when it should be sending one from the neighboring province instead. A single wide front in Russia is the absolute worst setup for the unit controller. To be able to deal with that wide front it needs more units, a lot more. The Russians get away with it because they have more units and more room to give up when a mishandled section of front allows the enemy to advance.

Having better templates will be important, but it has nothing to do with the current collapse of Germany. Most of your examples are concerning how bad the generic templates are for Japan. You are, of course, correct about that, but I do not see it having any bearing on the topic at hand. I am not speaking of the quality of AI play in general, but specifically of the current German collapse which happens in every game unless the human intervenes directly to stabilize the front.

 

[blusarge27]

Every single game I've played as the Soviets the northern front lines have been woefully under-manned, while the southern sections get the majority of the troops. It's like Germany would rather protect Romania than its own lands. And as soon as you reach Germany proper with your northern armies Germany starts moving troops from the south to try and hold the line in the north, which then leads to the southern section falling, and then it's like watching the heads at a tennis match, left, right, shuffle, shuffle...
I took this pic in my last game; Germany was also a bit behind schedule for some reason, I think I had to DOW him in the end.

I can't remember where the Hungarian and Romanian troops were at this point, but I believe they eventually showed up along with a few more German troops, but the ratio of troops north/south remained the same, or even worse maybe.

 

[Magraev]

Working with the current AI (as a human player) would it be best to split the front into smaller segment, and then give a ½ army (around 12 units) the responsability for that line? You'd have to constantly adjust the length of the different fronts, but at least you won't be sending single units from Leningrad to Stalingrad anymore.

I do have the expert AI mod, but I still see divisions hopping on the train going from the south to the north of my german frontline with Poland. I admit it took me by surprise, that the Polish had made their own baltic faction, and that Königsberg was threatened by their allies for a while, but some of the units in the north could have gone in stead. Maybe a couple of different front-lines (or at least offensive lines) would help?

 

[Dalwin]

I don't usually favor half size armies except for my mobile spearheads. The idea of narrower frontages being assigned to regular size armies is pretty well accepted at the moment. Time will tell if this precaution is less needed after the coming update. For an army of 20 or so, I try to give it a frontage of roughly 6 provinces if I can. I will stretch that to 8 if I must, but thinner than that makes attacking difficult.

 

[lihp]

To me it seems as if we mix several AI issues up, which are probably unrelated.

1. In-theatre-strategy: To me it seems the AI doesnt easily distribute units equally on a front. From some cases I do get the impression that Marines are close to lines for attack over a river or similar, Mountaineers closer to hills/mountains and tanks close to plains. Only excess units are then used wherever. Apart from that it seems the AI also does "Schwerpunktbildung" to create a specific attack focus. Somethign which leads to reshuffling of AI units inside the front to balance a changing front.
2. Cross-theatre-strategy: By times the AI is reavaluating how much armies it needs in which theatre. Once a new "theatre" comes up by eg. an invasion, it redistributes units. This becomes more complex, when allied nations are successfully invaded, eg. for Germany, when Italy is invaded. When does Germany intervene and at which force.

Imho the problem seems to arise with 2nd, which imho is the most complex and difficult decision. Examples:

• England, France, Canada successfully invaded along the french Channel border, by landing 4 divisions each in seperate places. Where do I take the division to fend them off? How many do I take? Which type do I need? Are my garrissions sufficient? Maybe overkill the invasion by 200% knowing a lot of armies are bound to also arrive soon? ...
• Italy was successfully invaded and Sicily was taken? Do I as Germany intervene? Italy has sufficient divisions, they can handle it.
  Now southern Italy was taken too including even Rome. Do I intervene as Germany now? Italy has exp. forces. They dont need me,...
  Italy is eradicated, non-existant . I mut heavily intervene with all my forces,...

Here the AI totally breaks. It just seems to happen at that time in mid-Barbarossa, that the other fronts break somehow or too much successull invasions happened. If I understood @SteelVolt correctly, he aims to have an extremely flexible AI, who should be able to make such decisions and react accordingly.

Thats my impression where HOI4 struggles atm.

Add Aluminum to Asia.

 

[Misaka_Complex]

The Axis AI still sends truckloads of units to the Japanese Soviet front when the Japanese join the war against the Soviets. I played UK on SP, got the convoy raiding doctrine and reduced a +4 Germany's divisions from 500 ish to around 80 just by killing convoys...

 

[Meglok]

I am not disputing any of those points. I do, however, fail to see their direct connection to why Germany falls flat in the same manner at the same time pretty much every game. They always collapse in the northern part of the Russian front after advancing in the south and center.

I recently played a game as Russia. In the hope that the Germans would be able to put up a fight I played on veteran and gave them +4. I also deployed more units on the border than I normally would but had a strong second line at the usual spot and some strategic reserve armies in addition.

Even with all that, they failed to push in the north, but I could not really stop them from pushing in the south and center. About the time they were getting close to Kiev, they pulled most of the units away from the north. They only abandoned half of the provinces and only those for a day or two at most. It was still enough to send them into a death spiral from which they could not recover.

I only ever used my border guards. I kept my second and third lines back in their positions, though I did commit one 12 division army of reserves. I was in Berlin by Christmas of '41. This was not because I played well. I went out of my way to handicap myself. It was not for economic reasons since they had much more than they normally would. It is because even with more units than they can currently field in '41 they cannot distribute them properly along the line.

It is not merely the fact that they shuffle. It is the way they go about it. The AI sets up one large army along a wide front. It has no hesitation to send a unit from 4 or 14 provinces away when it should be sending one from the neighboring province instead. A single wide front in Russia is the absolute worst setup for the unit controller. To be able to deal with that wide front it needs more units, a lot more. The Russians get away with it because they have more units and more room to give up when a mishandled section of front allows the enemy to advance.

Having better templates will be important, but it has nothing to do with the current collapse of Germany. Most of your examples are concerning how bad the generic templates are for Japan. You are, of course, correct about that, but I do not see it having any bearing on the topic at hand. I am not speaking of the quality of AI play in general, but specifically of the current German collapse which happens in every game unless the human intervenes directly to stabilize the front.

The template design and the production shortages do affect Germany, in that she and her allies including exp forces if any suffer from the hollow shell syndrome. Too many units to be supported by too little production. The new production penalties hammer the Axis production and don't really affect Russian production. Russia can usually still trade with the Allies and even Japan. My examples were just convienent ones of the overall problem.

Add to that the blown up factories in France and unit shuffle and you have German collapse in 1941-2. Germany might make up the losses suffered in the Battle of France but she can't possibly keep up with equipment losses in Russia now. And she will try to keep pumping out new unnecessary divisions diverting equipment because she has manpower, irregardless of the ability to equip and support those new units. The deployment and template design needs to be tied to production and resource capabilities for the ai to be competitive.

Add Aluminum in Asia

 

[Chondrite]

I see none of these issues using expert ai 2.0 so it would be interesting if chondrite (the mod designer) could chime in.

It's hard to point to any one area that explains the vanilla AI's performance aside from the unit controller. To give you some idea, the vanilla AI (Germany or otherwise) doesn't use half of its country's potential. I don't know if Paradox has any intention of changing this part of the AI or if they're happy with the way things are but regardless in Expert AI I've edited every part of the AI that is moddable (and twisted the game where it isn't moddable) to create the AI that I want. In a perfect world, the "veteran difficulty" would work similar to this mod (minus the things I've had to do to dodge bugs or limitations) but I doubt Paradox will ever have time for that.

Some of the vanilla AI's weakest areas off the top of my head (that are not related to its army management) are construction (GER building CIC/NIC and no refineries), political power usage coupled with poor law management, production priorities (air production is in very poor shape), poor armor designs, poor AT capability, research priorities (GER should be beelining to Modern Blitzkrieg etc.) and occupation laws/resistance management.

 

[Meglok]

It's hard to point to any one area that explains the vanilla AI's performance aside from the unit controller. To give you some idea, the vanilla AI (Germany or otherwise) doesn't use half of its country's potential. I don't know if Paradox has any intention of changing this part of the AI or if they're happy with the way things are but regardless in Expert AI I've edited every part of the AI that is moddable (and twisted the game where it isn't moddable) to create the AI that I want. In a perfect world, the "veteran difficulty" would work similar to this mod (minus the things I've had to do to dodge bugs or limitations) but I doubt Paradox will ever have time for that.

Some of the vanilla AI's weakest areas off the top of my head (that are not related to its army management) are construction (GER building CIC/NIC and no refineries), political power usage coupled with poor law management, production priorities (air production is in very poor shape), poor armor designs, poor AT capability, research priorities (GER should be beelining to Modern Blitzkrieg etc.) and occupation laws/resistance management.

Yep, yep, and yep.
Really wish PDS would get past their "we don't want to script the ai" design philosophy when they obviously do script the ai in many areas. Your concepts and ideas make the ai far more competitive. I know they have a grand design vision, but a vision goal is useless if it is never reached.

 

[lihp]

Yep, yep, and yep.
Really wish PDS would get past their "we don't want to script the ai" design philosophy when they obviously do script the ai in many areas. Your concepts and ideas make the ai far more competitive. I know they have a grand design vision, but a vision goal is useless if it is never reached.

Just cleared my agree button again. Because you are right as a short-term solution and there I would agree. Yet on the long term I actually like they are going down that road. Except it looks to me as if they are using the wrong tools for that

Add Aluminum to Asia.

 

[Praetori]

Two things that have been revealed about 1.4 should help with this. First, the revealed changes to the unit controller which have shifted units taking shorter trips and re-evaluating periodically to make sure their destination is a current need, should make the effects of overextension less severe.

Second, it has been mentioned both by @Meglok (and other players) as well as @podcat that in 1.3.3 Germany's allies are not sending expeditionary forces as often as they used to. Increased expeditions should also reduce the chance and impact of gaps caused by overextension.

Unless I am reading the pattern wrong, those two things by themselves should cure what is currently ailing Germany. (At least cure the current biggest problem. There are of course others remaining on the list.)

The current AI CAN put up some fierce resistance under the right circumstances. It's not the battlefield AI per-se that makes or breaks it IMO. The most common issue for German AI seems to be a lack of equipment and strict adherence to supply which usually means that there are no spearheads or chance of even a remotely successful Barbarossa (in the majority of cases, it sometimes works).
Post 1.3.X Germany under AI control seems to go deep into deficit on most types of equipment and then her divisions are doomed to wither and die due to crappy combat stats (which means even higher losses and even less equipment).
A lack of critical resources usually results in very limited MIC output and Germany going to the strictest laws doesn't help production in any way (since the resources she's lacking isn't domestically available in sufficient amounts anyway and just ends up with production penalties from the laws).

Better operational AI behavior might improve the situation but then it's not in any way certain as the Soviet AI will benefit from the same changes.

 

[Dalwin]

The template design and the production shortages do affect Germany, in that she and her allies including exp forces if any suffer from the hollow shell syndrome. Too many units to be supported by too little production. The new production penalties hammer the Axis production and don't really affect Russian production. Russia can usually still trade with the Allies and even Japan. My examples were just convienent ones of the overall problem.

Add to that the blown up factories in France and unit shuffle and you have German collapse in 1941-2. Germany might make up the losses suffered in the Battle of France but she can't possibly keep up with equipment losses in Russia now. And she will try to keep pumping out new unnecessary divisions diverting equipment because she has manpower, irregardless of the ability to equip and support those new units. The deployment and template design needs to be tied to production and resource capabilities for the ai to be competitive.

Add Aluminum in Asia

The current AI CAN put up some fierce resistance under the right circumstances. It's not the battlefield AI per-se that makes or breaks it IMO. The most common issue for German AI seems to be a lack of equipment and strict adherence to supply which usually means that there are no spearheads or chance of even a remotely successful Barbarossa (in the majority of cases, it sometimes works).
Post 1.3.X Germany under AI control seems to go deep into deficit on most types of equipment and then her divisions are doomed to wither and die due to crappy combat stats (which means even higher losses and even less equipment).
A lack of critical resources usually results in very limited MIC output and Germany going to the strictest laws doesn't help production in any way (since the resources she's lacking isn't domestically available in sufficient amounts anyway and just ends up with production penalties from the laws).

Better operational AI behavior might improve the situation but then it's not in any way certain as the Soviet AI will benefit from the same changes.

It is true that the AI for every major, even Russia, has problems with equipment shortages. This problem alone would normally kill the AI in every game, but I do not think it is what has been causing the early German collapse since 1.3.3. The AI has always had trouble with shortages and it is only specifically in 1.3.3 that we see this early German collapse. The production shortages may likely have become more acute because of the new system for resource shortages. The unit shortages on the Russian front are also affected by better garrisoning France and by the reduction in Axis expeditionary forces.

The German unit shuffle, along with for some reason valuing the southern sector of the Russian front higher than the northern, is causing Germany to collapse before the equipment shortages can really take their toll. When it is what Meglok calls hollow shell syndrome at work, the AI loses battles and its front gets pushed back. What we are seeing instead is the Soviets advancing rapidly into gaps left in Prussia and northern Poland. It is not because the units there are weak but instead because the units are not even there.

I do agree that all of the majors should get some scripting in the area of production preferences. Make sure that Germany goes with dispersed industry. Make sure that they put enough factories in the key areas and above all that they stop throwing away their efficiency by changing lines too often.

Human players will design their high level strategy around either acquiring the resources they need through conquest or tailoring their forces to what resources they can realistically expect to have. I am not sure that it is either feasible or desirable to have the AI doing the first of these. It should, however, be doing the second. The earlier example about how bad the current templates are for Japan is a prime example of this concept. Not only do those templates call for resources that Japan will never have, they also are best suited for terrain and for enemies which Japan will not face.

 

[Drewoid13]

At a strategic level, it doesn't help Germany will be losing to country A and then goes and declares in Country B.

Why they're losing? Idk but OP has a good philosophy for the tactical and industrial aspect. The AI doesn't do itself any favors though at the grand level.

 

[daemonofdecay]

I am not disputing any of those points. I do, however, fail to see their direct connection to why Germany falls flat in the same manner at the same time pretty much every game. They always collapse in the northern part of the Russian front after advancing in the south and center.

I recently played a game as Russia. In the hope that the Germans would be able to put up a fight I played on veteran and gave them +4. I also deployed more units on the border than I normally would but had a strong second line at the usual spot and some strategic reserve armies in addition.

Even with all that, they failed to push in the north, but I could not really stop them from pushing in the south and center. About the time they were getting close to Kiev, they pulled most of the units away from the north. They only abandoned half of the provinces and only those for a day or two at most. It was still enough to send them into a death spiral from which they could not recover.

I only ever used my border guards. I kept my second and third lines back in their positions, though I did commit one 12 division army of reserves. I was in Berlin by Christmas of '41. This was not because I played well. I went out of my way to handicap myself. It was not for economic reasons since they had much more than they normally would. It is because even with more units than they can currently field in '41 they cannot distribute them properly along the line.

It is not merely the fact that they shuffle. It is the way they go about it. The AI sets up one large army along a wide front. It has no hesitation to send a unit from 4 or 14 provinces away when it should be sending one from the neighboring province instead. A single wide front in Russia is the absolute worst setup for the unit controller. To be able to deal with that wide front it needs more units, a lot more. The Russians get away with it because they have more units and more room to give up when a mishandled section of front allows the enemy to advance.

Having better templates will be important, but it has nothing to do with the current collapse of Germany. Most of your examples are concerning how bad the generic templates are for Japan. You are, of course, correct about that, but I do not see it having any bearing on the topic at hand. I am not speaking of the quality of AI play in general, but specifically of the current German collapse which happens in every game unless the human intervenes directly to stabilize the front.

I'd say the large front lines are a major cause of disaster for the AI. That's been apparent to me for a while, when using the battle planner: the AI couldn't handle broad fronts with broad goals, such as using one large army against the USSR. It would stall, shuffle units around, and just struggle to find traction. But if I first split up my divisions into 4 or 5 different armies with their own sectors of the front, I could use the battle planner more effectively. There were still issues, but it seemed far more capable in action.

 

[lihp]

I'd say the large front lines are a major cause of disaster for the AI. That's been apparent to me for a while, when using the battle planner: the AI couldn't handle broad fronts with broad goals, such as using one large army against the USSR. It would stall, shuffle units around, and just struggle to find traction. But if I first split up my divisions into 4 or 5 different armies with their own sectors of the front, I could use the battle planner more effectively. There were still issues, but it seemed far more capable in action.

I am more the kind of player with massive amounts of art-heavy infantry and some minor tank armies.

I do not see such behaviour. Even long fronts from black sea to siberia, even when really wonky, work pretty well. Sure always with some interference but minimal.