Home Trending Latest threads New posts Developer diaries

Improving AI Germany vs AI USSR in 108

Jump to latest Follow Reply
  • We have updated our Community Code of Conduct. Please read through the new rules for the forum that are an integral part of Paradox Interactive’s User Agreement.
Finally moscow falls on july the 4th in 1942. I donnot think the USA approve of that.
On august the 4th JAP joins axis. MAN, MEN and SIA join axis, too. But the rest of the former JAP alliance does not. SIK and CHC are now an alliance at war with Allies and Axis and CHI.
The fall of vladivostok and the bitter peace coincide on august the 28th.
In late january 1939 germany challenges the Royal navy in the mouth of tthe themse. Both fleets have heavy casualties but none of the 10 german capital ships sink when fighting the Carrierfleet. ENG has about 17 CVs by that time.
On july the 8th in 1944 CHI capitultes to JAP. JAP makes CHI a puppet and therefore CHI joins Axis.
 
Pang Bingxun said:
Finally moscow falls on july the 4th in 1942. I donnot think the USA approve of that.
Click to expand...
Do you mean in a real-life situation the USA would've got involved if Moscow fell or just in the game the US AI didn't approve?
Pang Bingxun said:
In late january 1939 germany challenges the Royal navy in the mouth of tthe themse. Both fleets have heavy casualties but none of the 10 german capital ships sink when fighting the Carrierfleet. ENG has about 17 CVs by that time.
Click to expand...
The Royal Navy may have been the biggest in the world at the time, but it had nowhere near 17 carriers :wacko:
 
Denniss said:
Moscow - SOV has (or had) capital move events.
Click to expand...

I changed the moscow event to only trigger if moscow is not controlled by SOV. Before that Moscow tended to be evacuated too early. Now it may have been too late.

Mr_B0narpte said:
Do you mean in a real-life situation the USA would've got involved if Moscow fell or just in the game the US AI didn't approve?
Click to expand...

It is just a coincindece with the independece day in the USA, it is not of relevance to the game.
 
I played a bit farer. After months of hesitation germany finally pulls off sea lion. Norwich is the first target. This position is still far from save but germany attacks the USSR? :wacko:
Maybe the combat value against SOV should be deleted after BP? :confused:
 
If Germany is advancing too easily against the USSR maybe edit most of the provinces in eastern Poland and western Russia to 20/40 infrastructure to simulate the vast logisitcal problems Germany encountered in its advance on the Soviets IRL, thus providing more game balance and historical accuracy. The USSR (or maybe just an AI USSR) could be given extra manpower to cope with the added attrition due to the lower infra.
 
The too low soviet manpower could be considered a problem. The infra is not too low. The real obstacle in improving the results of Barbarossa is keeping things between a too strong germany and a germany too weak. Due to strong selfamplification it is a ride on the razors edge. :confused:
 
The soviets tend to run low on metal. One solution could be increasing the chances that SIK joins CHI. If SOV snd CHI have direct borders they will trade with each oher, won't they?
 
SPR wins the SCW, join allies later and is therefore annexed in 1940.
Barbarossa starts on june the 22nd in 1941. The startup is quite well, although it is quite some luck.
Germany: 54 Inf, 1 Cav, 44 Mot, 1 l.Arm, 14 Arm, 2 Par, 15 Mtn, 46 Gar, 5 HQ, 182 Landdivs, 8 MRF, 37 Int, 7 CAS, 8 Tac, 4 Nav, 3 Tra, 67 Airdivs; 505/326 IC.
USSR: 194 Inf, 5 Cav, 40 Mot, 4 l.Arm, 35 Arm, 11 Mtn, 3 HQ, 292 Landdivs, 8 MRF, 17 Int, 18 CAS, 8 Tac, 2 Tra, 53 Airdivs; 402/277 IC.

The third gpw event fires on july the 28th. 183 german divs vs. 268 soviet ones.
The fourth gpw event fires on august the 5th. 183 german divs vs. 239 soviet ones. Meanwhile german transports have captured brazzaville near Bata in Africa.
The fifth gpw event fires on september the 17th. 184 german divisions vs. 220 soviet ones.
In early october moscow falls because SOV prefers to walk trough finnland.
the last gpw event fires on november the 6th. 184 germn divisions vs. 217 soviet ones.

In january 1944 it is 160 german divisions vs. 263 soviet ones. moscow is still grey but finnland is threatened seriosly and the krim peninsula is pocketed by soviet forces.
 
Barbarossa start on june the 3rd.
Germany: 55 Inf, 1 Cav, 45 Mot, 17 Arm, 5 Par, 9 Mtn, 39 Gar, 4 HQ, 175 Landdivs, 11 MRF, 32 Int, 11 CAS, 7 Tac, 4 Nav, 3 Tra, 68 Airdivs; 489/315 IC.
USSR: 192 Inf, 5 Cav, 39 Mot, 4 l.Arm, 36 Arm, 1 Par, 11 Mtn, 4 HQ, 292 Landdivs, 8 MRF, 18 Int, 18 CAS, 7 Tac, 2 Tra, 53 Airddivs; 408/282 IC.

The third gpw event fires on july the 2nd. 176 german divisions vs. 264 soviet ones.
The fourth gpw event fires on august the 21st. 178 german divisions vs. 242 soviet ones.
The fifth gpw event fires on september the 12th. 178 german divisions vs. 219 soviet ones.
The last gpw event fires on november the first. 179 german division vs. 204 soviet ones.
Germany: 55 Inf, 1 Cav, 46 Mot, 18 Arm, 6 Par, 9 Mtn, 39 Gar, 5 HQ, 179 Landdivs, 12 MRF, 36 Int, 12 CAS, 7 Tac, 3 Nav, 3 Tra, 73 Airdivs; 482/311 IC.
USSR: 134 Inf, 2 Cav, 28 Mot, 4 l.Arm, 25 Arm, 1 Par, 7 Mtn, 3 HQ, 204 Landdivs, 9 MRF, 18 Int, 16 CAS, 6 Tac, 2 Tra, 52 Airddivs; 313/224 IC.

God luck or bad luck, this scenarios solves off good enough. I donnot want to touch it in 108 again. :)

Edit: Here is the front in 1942:

attachment.php


In 1943 it is 160 vs 280 divisions and the soviets advance accordingly.
 
I don't intend to increase Pang's frustration, but it seems to be a proper thread to ask the question... If it's true, indeed, that a player can create his/her own ai switch files (and not just edit existing ones), there surely must have been some modding experiments (in DAIM?) to simulate operative level maneuvers through the chains of ai switch files and events triggering them--through cycles of defining new target provinces and checking their capture as the campaign progresses? Or would it just end in an ineffective mess eventually?

It at least seems that this approach might offer for an AI an additional layer of rationality. Historical campaigns would be divided into fine phases/stages, each with a clear goal, trying to build a positional advantage step by step. Phases (ai switch files) would be "engineered" based on analyze of typical force setups (strength, position) and campaign theater (province connections, terrain, weather). (In case of AI Barbarossa, one classic option might be: start with a heavy concentration of German forces in Königsberg-Torun-Lomza and Zamosc-Przemysl-Presov-Bacau; set high attacking priorities to Suwalki and Stryj (with a solid air support); after those provinces are taken, event that checks ownership of Suwalki and Stryj is triggered and it switches on a new ai switch file that targets now Zhlobin and Vyshgorod, protecting at the same time both pincer claws (by setting high garrison priorities to thrust corridor provinces (Grodno, Lida, Bobraisk etc); after reaching Zhlobin and Vyshgorod, next event is launched that activates ai switch file that coordinates clearing of Pripyat pocket (with a heavy air support); then Kiev and Minsk are attacked; etc.) Obviously each event would have several (three) options which ai switch file (new set of targets) to activate, defining how the next campaign phase would unfold (choice depending on force balance, positions, previous success and containing a random element as well)--ensuring flexibility and unpredictability of offensive. In an ideal also the time factor should be taken into account somehow (with a current "ai vocabulary" difficult)--each stage should have certain period available for being completed; if not completed in due time, plan B (a new ai switch file) should be implemented--otherwise the offensive is almost sure running into some absurd dead end situation sooner or later (probably before historically plausible). Currently the AI Barbarossa is already divided into stages (in a sense), but it could be much finer. What do you think?
 
Of course, there's a serious threat to get carried away with all this. On the other hand, it would be nice to perceive for a change that AI has some broader battle plan guiding its activity. Most campaigns, if particular game hasn't deviated too far from its historical path, have certain limited set of sensible moves available (defined by fixed borders, province connections, distances, terrain, time of year and probable force concentrations) that are clearly better than any alternative--if attacker doesn't make one of them, his/her/its result would be significantly weaker. (It's unlikely that AI would stumble on proper strategy.) In the mildest form of the idea, target altering ai switch files could create new Schwerpunkt(s) to shake up the front balance time to time (if reactive defense has succeeded to distribute its forces to match attacker's forces and freeze the situation). A criterion of success of the approach would be how well offensive (German) AI could cope against (Soviet) AI's static defense. If scripts are highly context sensitive and rational, they might not be that bad necessarily.

But I do understand that even if it does work in theory, inventing and tweaking the system would require a lot of time and energy.
 
Barbarossa start on june the first.
Germany: 61 Inf, 1 Cav, 47 Mot, 3 l.Arm, 15 Arm, 2 Par, 15 Mtn, 44 Gar, 3 HQ, 191 Landdivs, 7 MRF, 37 Int, 7 CAS, 8 Tac, 5 Nav, 3 Tra, 67 Airdivs; 485/313 IC.
USSR: 197 Inf, 5 Cav, 38 Mot, 4 l.Arm, 38 Arm, 1 Par, 11 Mtn, 3 HQ, 297 Landdivs, 7 MRF, 16 Int, 18 CAS, 6 Tac, 2 Tra, 49 Airddivs; 409/282 IC.

The third gpw event fires on june the 16th. 192 german divisions vs. 264 soviet ones.
The fourth gpw event fires on june the 23rd. 192 german divisions vs. 243 soviet ones.
The fifth gpw event fires on june the 30th. 193 german divisions vs. 212 soviet ones.
The last gpw event fires on july the 14rg. 193 german division vs. 188 soviet ones.
Leningrad falls on august the 19th. 196 german division vs. 166 soviet ones.
On Septemberthe 20th the Kola pennsula is cut of from the soviet mainland, 30 soviet divisions are there. 198 german division vs. 145 soviet ones.
Stalingrad falls on september the 26th. 198 german division vs. 139 soviet ones.
Baku falls october the 3rd. 198 german division vs. 137 soviet ones.
Moscow falls october the 5th. 198 german division vs. 132 soviet ones.
Japan, Manchukuo and Mengjiang join Axis on october the 6th.
Swerdlovsk falls november the 20th. 200 german division vs. 97 soviet ones.
Bitter Peace triggers on november the 22nd. This saves 18 severly hurt divisions from the Kola pocket that was about to be closed.

It is amazing to see that the german AI can still beat the soviet AI if it severly beats the soviets in the first 30 days of the campagn. Please note that the ruleset is practically unchanged to my earlier campagn in #210.

Edit: Here is the savegame. I have used some minor changes i will upload later.
 
Last edited:
You must log in or register to reply here.
Top Bottom
Back