1.3.3 (2dd2) Issues / Bugs.

[Gwydion5]

Hello everyone, with the new patch, I have a new report. For those that have read my reports before, you know that I can be very meticulous and post a barrage of screenshots to support my point. I will refrain from doing that, and post screenshots based only on request. The first part of this post will go over the new issues. I will respond to this and post a follow up of resolved issues from 1.3.2 and issues still observed from 1.3.2.

Let's begin:

Game Version : v1.3.3 (2dd2) - New Game (as in not loading from a saved game) - Regular difficulty - Historical Focus
DLC : Together for Victory
Mods : NONE (I have never used mods ever in HOI4, nor edited any files)
Observing : Germany AI

1. Light Tank Production : Early in 1936 AI Germany ramps up Light Tank Production to the point it needs to import 4 CIC's worth of oil. It is doing 1 line of 15 MIC's and 2nd line with 4 MIC's by March 25th of 1936. This is too much too soon for a competent Germany AI. The second issue with Light Tank Production is that the AI queues up a bunch (11 by my screenshots) of light tank divisions in 1936 and by August of 1936, only divisions 9, 10 and 11 remain. The problem with this is that when I go to Division Overview screen, I only have 5 light tank divisions. Which means of the 8 divisions that were queued prior, only 2 of them were deployed.
   
   
2. Concentrated Industry Research : AI Germany starts off researching Concentrated Industry only to switch to Dispersed Industry later on. In one game it went as high as Concentrated Industry 3 and then switching to Dispersed Industry 1. The point at which it changes seems inconsistent, sometimes it only makes it as far as 1 level of concentrated Industry other times it's 2 or 3 before switching to dispersed.
   
   
3. Production Schizophrenia : I believe this is probably related to the Light Tank Production issue. But AI Germany makes wild and drastic swings on MIC allocation based on need. Case in point in late August of 1936 the AI has dropped Light Tank production from 19 MIC's down to 2 MIC's, it even destroyed the line that had the most production efficiency and put 22 MIC's in to Towed Artillery. This has now put a heavy need on Tungsten. So now 7 CIC's are being used in trade. To put this in context, in March of 36, AI Germany was importing 32 oil, 8 rubber, 8 chromium; in August of 36 AI Germany is importing 24 oil, 24 tungsten, and 8 chromium. Rubber has been sort of ('ll explain in point 4) dropped, even though 13 is needed.
   
   
4. Trade Issue : August 1936 to April 1937 the AI is trying to import rubber from France. In August of 36 when the trade is not active, it shows France having 19/51 Rubber to import, with British Malaya having 460, and Dutch East Indies having 358 available for trade. AI Germany initiates a trade with AI France for CIC's worth of rubber, but something happens where France's stock drops from 19 down to zero, and AI Germany only imports 8 of the 16 it supposed to, out of the 19 that France supposedly is supposed to have. The following day AI Germany cancels the trade, and then the day after that it re-establishes it only to have the same thing happen again. So it's definitely caught in a loop. By the end of May 1937, France's export quantity for Rubber has increased from 19/51 to 24/56 and still when AI Germany tries to import 2 CIC's worth (16 Rubber) it shows France dropping to 0 quantity available for export, and Germany has only imported 9 of the 16 Rubber.
   
   
5. Garrison Issues : AI Germany now tries to garrison the conquered land. The problem with this is that it is basically game breaking in its implementation and Russia can steam roll Germany in a few months, and is extremely likely to keep Germany completely bogged down in Eastern Poland lands. The first issue is that it doesn't build the most optimal divisions (cavalry) for resistance suppression. The second issue is that it tries to guard the coast lines which requires a lot of divisions. The third issue is that it's constantly moving units around to the point that most resistance suppression is lost. The fourth issue is that it tends to use units like MARM's and Motorized Infantry, which could be better used else where. The current implementation just doesn't really work. Either it puts too much on Garrison Duty that it adequately suppresses resistance, but loses the war in 41 because it can't mount a defense against Russia. Or it allocates most of the divisions to the Russian front and then fails at suppressing resistance to the point it's repair queue grows faster than it can keep up. Even then Germany has no hope against Russia.
   
   
6. Naval Invasions : AI Germany sent 2 divisions of Infantry out of Denmark to Norway that were constantly defeated, and the day they arrived back at Denmark from the province they launched from, they went right back to doing the naval invasion again. This happened for at least 3 / 4 months, with an invasion cycle (launch, defeat, return) being 7 days or so.
   
   
7. Transporting Logic : AI Germany and AI Italy both have horrible logic still. AI Italy is still sending 10-20 divisions to Ethopia via the English Channel or the Atlantic from France. AI Germany in 1940/1941 is sending 20-30 divisions to Japan in their war with China. I seen at least 2-3 MARM divisions off the coast of Africa, but it looked like a steady stream of divisions were being sent for some time. I have also noticed that on front changes that the AI will use a port to move 2-3 provinces down from it's current position. I see this from time to time around Denmark, Netherlands, Belgium France and Italy. I would say for Germany it is worse now than 1.3.2. Italy seems about the same, maybe slightly improved.
   
   
8. Maginot Line : In the two games I've observed, Germany easily punches through the Maginot Line, well before Netherlands or Belgium fall. Also Italy is dominating Southern France, grabbing more land than Germany.
   
   
9. Man Power Losses : In one game the man power losses of Germany and Italy were double what the Allies were by the time France fell. Germany was already at 1 million, and Italy was at 600k, while the Allies were half of that. Second game that I observed the losses were still about double, but much less this time around. But the losses against Russia are close to 1-1. I think it's 1 German casualty for 1.1 or 1.2 at most Russian casualty.

 

[Skalman500]

Trade Issue : August 1936 to April 1937 the AI is trying to import rubber from France. In August of 36 when the trade is not active, it shows France having 19/51 Rubber to import, with British Malaya having 460, and Dutch East Indies having 358 available for trade. AI Germany initiates a trade with AI France for CIC's worth of rubber, but something happens where France's stock drops from 19 down to zero, and AI Germany only imports 8 of the 16 it supposed to, out of the 19 that France supposedly is supposed to have. The following day AI Germany cancels the trade, and then the day after that it re-establishes it only to have the same thing happen again. So it's definitely caught in a loop. By the end of May 1937, France's export quantity for Rubber has increased from 19/51 to 24/56 and still when AI Germany tries to import 2 CIC's worth (16 Rubber) it shows France dropping to 0 quantity available for export, and Germany has only imported 9 of the 16 Rubber.

I have observed the exact same issue with several nations. I posted a thread involving some minors on the forum just before. I think it affected balance of the nations as well, it's not as obvious in the first few years but as time goes on i diverges quicker, did you observe this too?

 

[Gwydion5]

Previous Thread on 1.3.2 issues that focuses more on the AI.

• https://forum.paradoxplaza.com/forum/index.php?threads/ai-issues-in-1-3-2-e13a.994155/

Issues Resolved/Improved :

• AI now builds other things besides CIC's and MIC's in 1936-1939. I've seen Germany build NIC's and Refineries in 1936.
• AI now tries to garrison conquered lands.

Issues Still Present :

• AI still does the commander swap every other day, you can really see this with Spanish volunteers, but I've been noticing it with fleets as well.
• AI still spends Political Power on ministers and techs sub-optimally.
• AI still does commander and division reassignments that make no sense / are not optimal or goal/strategy orientated.
• AI still does not try to train/exercise troops optimally.
• AI still promotes Manstein to Field Marshal.
• AI still has too much spaghetti with front/battleplanner management.
• AI still uses gigantic single army fronts commanded by a Field Marshal. (Not the end of the world, but still something that should be improved in terms of commander/army efficiency.)

And that is all I have for now. I do see Steam has me queued up for an HOI4 patch in a few minutes, perhaps some of this will resolve the issues reported in the post above this one.

 

[Gwydion5]

I have observed the exact same issue with several nations. I posted a thread involving some minors on the forum just before. I think it affected balance of the nations as well, it's not as obvious in the first few years but as time goes on i diverges quicker, did you observe this too?

From the Trade Screen I believe if you mouse over the exported value it will tell you who is trading for the goods and their trade relation value. Unfortunately I did not do that with the France issue or I might have been able to see who was competing for Germany's rubber. Specifically to your question, I can't say for sure how it's changed the game, and where it's caused imbalance. But I think you are right to be concerned given the values you are seeing pre/post patch. I think the trade system is definitely a little buggy/imbalanced right now, but I think that is driven by the goldfish production AI that is only responding to what is, and not having goals or long term strategic assignments. Which is probably being applied to every nation and thus making them all trade freaks due to the production AI.

 

[Skalman500]

Yeah we'll need some more test runs to say where the imbalances might lie, just from what I saw Sweden got 21 factories from trade by September 1939, most of which should have gone to France (mainly Steel Export).

 

[Gwydion5]

I've been informed that using Observe mode can be responsible for problems observed. So instead of using observe, I am now going to use "human_ai" or play a minor like Guatemala and use the save system to reload if I switch in to observe mode.

Game Version : v1.3.3 (2dd2) - New Game - Regular difficulty - Historical Focus
DLC : Together for Victory
Mods : NONE (I have never used mods ever in HOI4, nor edited any files)
Observing : Germany AI using "human_ai" mode from the console.

Production Issues : There are a couple things that have changed that I think are having a rather large impact on the game design / balance. The first is that the AI is much more reactive and aggressive to the needs of the country. With Germany we see this with major swings of MIC allocation on things like Light Tanks and Artillery. This increased production forces increased need for resources which reduces the amount of CIC's being used to build more CIC's. So let's step through this.

For the most part I took screenshots on the 10th of every month of the production, trade, recruit/deployment and division overview screens. Here is what January 10th looks like:

So in all my games, I almost always put infantry equipment as an absolute must for full production of at least a single line, later on eventually adding a second line and maxing that one out. So seeing Infantry Equipment at the top is a good thing. What has changed for AI Germany is that it now is building tanks, and building tanks requires more oil to be imported. But the real issue behind production is how the AI responds to it's needs, and I think a failure to do long term production planning.

Here is the trade screen for January 10th :

Now one thing I did with the Trade Screen was do a screenshot on each mouse over for the need/surplus section of the resource so we could see where this demand is coming from in addition to understanding the amount of CIC's we are spending. So here we can see everything that oil and rubber are being used for as those are the only 2 resources we are importing. So 14 on ground, 10 on sea and 5 on air. This is a pretty significant change to the AI I know in 1.3.2 as it normally did not focus this much on navy or tanks so early and as I will demonstrate this demand will continue to grow.

Ok so let's get to the heart of the issue, which I think is the production logic.

1. PRODUCTION


February shows that MIC's have been reduced from Infantry Equipment to fill out line 6 of light tanks for 15 MIC's, and a new line 1 of light tanks for 4 MIC's. March shows more or less the stability of the production queues at this point.


May 10th shows that the AI is pulling MIC's off light tank production for artillery production. The problem is that it's pulling from line 6 that has the most production efficiency instead of line 1 which has the worst for light tanks.


2 more screenshots from May and one from June 11th. We continue to see the AI remove MIC's from Light Tanks line 6, the most efficient and productive line to fill in Artillery in line 5, as well as add MIC's back in to Line 2 for Infantry equipment. These adjustments are inducing inefficiency in to the product lines when it lowers and raises by 1 MIC every other month. Clearly others like Light Tanks and Artillery are seeing the biggest swings.


June 19th and the AI has actually removed/deleted line 6 for light tanks, the most efficient line that was producing light tanks. Now if I was to guess, I would say this is because the code/logic is actually picking the last line to modify production values when 2 or more lines exist of the same type. So in this case, because line 1 is the first line, and line 6 would be the second line for light tanks, line 6 was being altered per the needs of the AI without evaluating the production efficiency of the line. I say this because if it was the number of MIC's that line 6 had over line 1, then it should reduced line 1 by 1 MIC when Line 6 was at 3 MIC's. I also observed later in the game (with screenshots if you want them) that the AI opens a new artillery line but leaves at the bottom of the production list and all modifications to artillery production happen to that line instead of line 5 in the screenshots above.

I actually have a bunch more screenshots showing this effect with Artillery lines. in the 2nd half of 36 and in 37. But the AI is really hurting itself by not evaluating the production efficiency of the lines it want's to alter, in addition to not making valuations based on type. Infantry Equipment for example should be the last thing to touch in terms of reducing or doing so should be the exception, because you need a significant surplus of equipment going in to war.

I also feel the AI is way too sensitive in responding to immediate needs, and not making long term decisions on how to address those needs. AI Germany does not need to go so gung ho on light tanks only to do a 15 MIC swing from light tanks over to artillery.

2. TRADE
So with the production swings happening, it creates the need to devote CIC's to trade to get the resources necessary. Here is what that looked like over the same period of time.


One thing to note is that the AI imports 8 chromium, and only spends 1 of the 8 imported for years. But by February 10th it's spending 7 CIC's on trade which it never really did before, and I think that might be helping slow Germany down economically.

3. RECRUIT / DEPLOYMENT


One thing that seems to be happening, and I'm not sure if it's a big deal or not, is that the AI is canceling divisions that are being recruited and trained. In the screenshots above you will notice that the line numbers change for Light Tanks without any new light tank divisions being deployed. Could be a small issue, could be a huge issue where the AI thinks it deployed a tank division when in fact it did not. Maybe it was trying deploy early and the code instead deletes it? I don't know. I do know that some of the tank divisions do end up being deployed, but as evidenced in the screenshots, not all of them.

 

[Gwydion5]

Production AI Redux

So above I mentioned that I seen more of the same regarding production AI with Artillery. This post will back up that claim and hopefully make it indisputable.

September 1936 to December 1936


From September to December we will see a reduction in artillery. The line that is being reduced is at the very bottom of the production queue and thus not visible. But with the trade screens we can see the resource allocation of Tungsten to this line, and thus can confirm that Line 5 artillery are not being reduced. This supports my guess that it may be pulling MIC's from the last line of equipment, when 2 or more of the same equipment lines are present in the queue.

January 1937 to February 1937


In January after the AI just disbanded a line of 3 MIC's for artillery, the AI creates a new line and moves it to the top. Again it starts pulling MIC's from the more productive line, while also increasing the MIC's in the new line. Also I included the trade screens so you could see that the order displayed for Tungsten usage matches the production queue. So from September to December, it's clear that there is an artillery line below the artillery line that has 15 MIC's.

February 1937 to June 1937


Nothing but production screens here capturing the shift of MIC's from artillery Line 6 to other things, including the artillery on line 1. The last screenshot really says it all when you look at what line 1 is producing per day and line 6 is producing per day while considering how much line 6 has lost in terms of MIC's.

 

[Gwydion5]

Research Issues
This post is going to cover some research issues with AI Germany. Specifically AI Germany's inability to decide which Industry Tech line it wants, which will confirm my original post observation.

Game Version : v1.3.3 (2dd2) - New Game - Regular difficulty - Historical Focus
DLC : Together for Victory
Mods : NONE (I have never used mods ever in HOI4, nor edited any files)
Observing : Germany AI using "human_ai" mode from the console.

First Game

So in these 4 screenshots, you see the start of research by the AI, then you see it research concentrated industry. After it completes concentrated industry, it switches to dispersed industry. My intention was to try and capture this in issue in 3 different games, so I exit to the main menu and start up a new game.

Second Game

So in these screenshots I watch the AI start with dispersed industry and so I keep the gaming going to see if it is going to switch. I notice around April that the AI has started researching Advanced Machine Tools, which I found odd, given that Dispersed Industry 3 would be the better choice. So I go to the Research Screen for Construction and see that at some point it switched to Concentrated Industry. Not sure if it was a display issue I check the production screen and confirm it did switch to concentrated industry. Now when playing in "human_ai" mode a lot of the notification pop ups are suppressed, where as in observe mode you would still get the research notifications. So that is why I missed when it actually switched to concentrated industry, but have supported my original assertion/bug report when I was using observe mode.

Being curious I let the game continue checking to see if it changes back to dispersed industry, and if it has this issue with any other split technology trees.

These screens just show the progression of the AI's research starting in January 1939. It does not start researching concentrated industry 2 until May of 1939 and finishing at the end of November of 1939. So essentially Germany has spent mid 1936 to the end of 1939 with Industry Tech 1 only. Not including the time wasted researching 2 levels of dispersed industry only to switch to concentrated industry 1.

Another thing worth noting in these screenshots is that the AI researched encryption and improved computing, passing over decryption. It could be a change for historical focus maybe? Or it might be another issue with research priority. The devs assert that intelligence matters to the decision making of the AI, so decryption and radar would be important to that aspect and thus the AI might be losing all sorts of functionality.

**EDIT** Also notice that AI Germany beats France going through the Maginot Line before Netherlands and Belgium fall. Though they do fall shortly after France. Again this is something I am seeing consistently in my games. I'm not sure how big of a deal it is, but for a historical focus game, you would think France might be a bit more focused on defending the Maginot Line forcing Germany to go around?**/END EDIT**

I'm going to do another post below this one where I let the Second Game continue until 1941 in the hopes of finding more research issues.

 

[Gwydion5]

Army / Command Issues

In the previous post (Research Issues) I did 2 games, since the Second Game went so far, I continued to observe the game and take screenshots and I noticed some game breaking command actions by the AI. All these screenshots are from the second game and I will include the save game at the bottom because the amount of divisions moving all over Europe and Africa would take multiple posts to capture. One other thing to note in the screenshots below, are the units being used for garrison duty, like motorized infantry, medium tanks, and the quantities in various places.

March 31st, 1940

110 German divisions are assigned to Africa.

April 19, 1940

109 divisions removed from Africa Theater, assigned back to Europe.

May 1st, 1940

81 divisions assigned back to Africa.

July 6th, 1940

This screenshot exists because I was watching the Land Doctrine research, but it does provide a purpose in showing the current army/theater makeup as well as current research and resistance suppression or lack there of.

August 19th, 1940

The AI has disbanded the African theater, divisions are assigned to Europe.

September 4th, 1940

African Theater has been recreated 39 divisions assigned. So to recap : Africa has gone from 110, to 1, to 81, to 82, to non-existent, to 39. That's some indecisive strategy/response.

September 8th 1940

Still no research for decryption after completing improved computing.

September 20th 1940

African theater increased to 51 divisions.

October 29th 1940

African Theater reduced to 26 divisions. 159 Divisions are doing garrison duty, 4 are likely on the eastern front with the Soviets. One would assume the AI is starting to prepare for Barbarossa?

November 16th 1940

European Theater reduced to 25 divisions, African theater increased 138 divisions.

November 22nd 1940

European Theater increased by 130 divisions, African theater reduced to 35 divisions. The 120 Divisions in Europe are for the Eastern Front with the Soviets.

December 12th 1940

Soviets declare war on Finland, and Finland joins the Axis. Second screenshot to show the eastern front.

December 16th - 17th 1940

Finland accepts Germany's request to join the war. Garrison army in Europe has been reduced from 34 divisions to 2 divisions, with African Theater increasing by 42 divisions. I would assume 32 of those divisions are from the Garrison army in Europe. In the second screenshot for the 17th of December, we see the Garrison army given 25 divisions from the Eastern Front army. I count 23 divisions being pulled from the front lines near Hungary and Romania.

In the third screenshot I count 66 units from the Eastern Front army along the front. We can also see a bunch of armies in the african theater moving off the eastern front line as well. Because the AI moved 25 divisions to the Garrison Army in Europe, as well as the divisions/armies being assigned to the African theater, this is a division realignment along the front with the existing 66 divisions stationed there.

December 26th, 1940

European theater greatly reduced, Eastern Front army reduced from 83 to 34 divisions, theater itself reduced from 110 to 60 divions. A second African theater created adding 104 division to it. Total forces for African theaters is 130 divisions. Some divisions have movement orders through Russia to Asia. Second screenshot shows current air power assignment. Strategic Bombers in Siam trying to hit Southern England.

December 29th 1940

African Theaters have 128 divisions, Strategic bombers still trying to bomb Southern England from Siam.

December 31st 1940

African Theaters have 128 divisions, Strategic bombers still trying to bomb Southern England from Siam.

January 1st, 1941

2nd African Theater disbanded, African theater now has 5 total divisions. European Theater has increased to 185 divisions.

So let's recap a bit: In the month of December, Germany accepts Finland in to the Axis on the 12th because Russia declared war on Finland. On the 16th Finland accepts Germany's request to join the war. Germany reduces the European theater from 150 divisions to 109 divisions. It assigns all but 2 of the Garrison divisions in to the African theater, and then pulls divisions from the Eastern Front to fill the reduction it just created in the Garrison army, as well as various other divisions being assigned to Africa from the Eastern Front. By the 26th while at war with Russia, the AI decides to further reduce the European theater and create a second African theater with 104 divisions assigned to it, only to completely reduce the african theater down to 5 divisions on January 1st.

I just don't see how the AI can fight effectively, if it's constantly making changes to theaters/army command without any constraints from long term goals and objectives. It seems pretty clear based on my reports in 1.3.2 and now 1.3.3. that the AI evaluates conditions on present conditions and with no regards to what it may do or long term objectives/strategy. Reducing the garrison army down only to fill it back up with units from the eastern front demonstrates that effect. But as I said my previous observations in 1.3.2 demonstrate that as well. It's a compounding problem, because the AI makes a change, then makes changes because of that change, until it establishes a stable condition/criteria. Which probably makes a lot of the AI processes inefficient in terms of cpu cycles, but also desired effect.

January 9th, 1941

At this point I stop playing the game when I look around the map to see all the units moving. You can load the save and take a look for yourself, but it's a complete mess and Russia is going to again roll over Germany in 41 or at the latest 42.