AI Abandoning the front

[coregonas]

Why AI abandons the front?

Spain: Late 1936. Both sides have a more or less regular front. My 2 german divisions try advance here or there with some success. Then... my allied spanish trrops start running as mad abandoning the front and concentrating in one point. The enemy, just start walking happily and capture half the map.

In Russia... this breaks the game.


Why Ai suicides himself against the Forts?

Germany: Some British/French troops start assaulting my German frontier. I just put enough troops to sleep there and masscare hundreds of thousands of troops.


Can it be posible tha the AI be smart enough to just hold trenched at least a single unit in each position in the front before try to do attacks.

 

[kettyo]

Why AI abandons the front?

Spain: Late 1936. Both sides have a more or less regular front. My 2 german divisions try advance here or there with some success. Then... my allied spanish trrops start running as mad abandoning the front and concentrating in one point. The enemy, just start walking happily and capture half the map.

In Russia... this breaks the game.


Why Ai suicides himself against the Forts?

Germany: Some British/French troops start assaulting my German frontier. I just put enough troops to sleep there and masscare hundreds of thousands of troops.


Can it be posible tha the AI be smart enough to just hold trenched at least a single unit in each position in the front before try to do attacks.

I think the suicide fort assault results from some crazy relocation order which makes the divisions pass through that fort so actually they just run headfront into those forts. It is not an actual attack plan. Same as with troop transports passing through sea areas under total enemy rule. There is no threat check for army relocation. It's up to luck.

 

[coregonas]

I think the suicide fort assault results from some crazy relocation order which makes the divisions pass through that fort so actually they just run headfront into those forts. It is not an actual attack plan. Same as with troop transports passing through sea areas under total enemy rule. There is no threat check for army relocation. It's up to luck.

Sad....

 

[yamato2cz]

I think the suicide fort assault results from some crazy relocation order which makes the divisions pass through that fort so actually they just run headfront into those forts. It is not an actual attack plan. Same as with troop transports passing through sea areas under total enemy rule. There is no threat check for army relocation. It's up to luck.

nah, its just stupid AI. build 10lvl fort in luxemburg. see millions of axis soldiers perish under utterly uber luxemburgians. or build 10lvl forts on any fronts. AI will attack it relentlessly until it runs out of manpower.

 

[kettyo]

nah, its just stupid AI. build 10lvl fort in luxemburg. see millions of axis soldiers perish under utterly uber luxemburgians. or build 10lvl forts on any fronts. AI will attack it relentlessly until it runs out of manpower.

I don't see it happening like they carried out an offensive plan so it is not happening all across the fortified front simultaneously, at least not by me. It is more like sporatic 'attacks' from time to time which i think is a relocation issue. I can still be wrong though...

The AI has a filler role in this game anyway so don't expect too much

 

[CharlieFox]

I always suspected that suicidal attacks against forts or straits* are in part a result of the game not estimating correctly the chances to win a battle . What I noticed is that when attacking across straits or to fortified provinces the "chances of winning" (the number that appears on a circle when there is a battle) start quite high but quickly drop as the battle progresses. If the AI is using the initial estimate to decide whether attack or not it would explain a lot.

*If you conquer Sicily as the allies you will watch Italy sacrifice hundreds of thousands of men launching continuous attacks across the strait.

 

[Anani]

Soviets also do that on the causcaus border when I was fighting against them as Turkey. They fall back to their fallback line, wait a bit while I figure out they're gone and push, then they come back, hold line a bit, and go back again.

They kept doing the same thing. I pushed all the way up to stalingrad and actually took stalingrad. After that the front got a bit too long so I just waited for germans go grab the rest.

 

[coregonas]

I always suspected that suicidal attacks against forts or straits* are in part a result of the game not estimating correctly the chances to win a battle . What I noticed is that when attacking across straits or to fortified provinces the "chances of winning" (the number that appears on a circle when there is a battle) start quite high but quickly drop as the battle progresses. If the AI is using the initial estimate to decide whether attack or not it would explain a lot.

*If you conquer Sicily as the allies you will watch Italy sacrifice hundreds of thousands of men launching continuous attacks across the strait.

Completely Agree

for example, only attacking a lone area, just with a 2 areas front, Battle Plan says there is no chance to win, claiming an inexistant major river exists. The River is one area after the expected limit of advance.

So... I force the attack manually, and batle is done in a few hours. Battle even statrs with 60 or 70% green values.

Perhaps is not taking into account all modifiers, including Leaders, Tech... sho knows.

 

[Dan1109]

its a well known issue, the AI Front management is just "sad". It creates its own weaknesses (like abandoning fronts), and doesnt have the intelligence to EXPLOIT its own fault (I've seen empty capitals with loads of enemy divisions nearby, and they refuse to walk in for free). What's worse is that for naval invasions, it will do this well after securing a more than viable beachhead and taking critical ports. It abandons the ports, lets them be taken, and now the invasion force has no supply and is easily eliminated.

They promise improvements for this in 1.4.

 

[C-Breeze]

I always suspected that suicidal attacks against forts or straits* are in part a result of the game not estimating correctly the chances to win a battle . What I noticed is that when attacking across straits or to fortified provinces the "chances of winning" (the number that appears on a circle when there is a battle) start quite high but quickly drop as the battle progresses. If the AI is using the initial estimate to decide whether attack or not it would explain a lot.

Good point. I don't think it's coded to update once it's made the initial calculation. In much the same way as naval invasions only require local superiority briefly and initially. There is no subsequent check, so even if you lose superiority in the sea zone, invasions are green lighted in perpetuity.

 

[stjern]

Completely Agree

for example, only attacking a lone area, just with a 2 areas front, Battle Plan says there is no chance to win, claiming an inexistant major river exists. The River is one area after the expected limit of advance.

So... I force the attack manually, and batle is done in a few hours. Battle even statrs with 60 or 70% green values.

Perhaps is not taking into account all modifiers, including Leaders, Tech... sho knows.

This is true and it could help the AI alot if it just relied a lot more on basing attack decisions on the chace to win. Currently the AI clearly can't see the same thing the player sees. It will often stay in a frontal assault with obviously 0% chance of even dealing damage. The problem is that the AI only sees the batleplanner score for the overall front and then some other calculation based on steps forward in provinces that the player cant see. It doesnt take win/lose chance in a battle into account which would be much better.

 

[Lord Cao]

I found the same weird behavior around forts. In my case, France abandoned the Maginot line to re-position to southern France. I was pushing with a spearhead attack to the south though Italy and noticed a division abandon their post and move South to reinforce. However, when I noticed this I commanded my single division on the Siegfried line to advance to the level 10 fort in France. I continued to cut off supplies to the other areas of the fortification by capturing provinces behind the forts. France, pretty much left a door open for my forces to push through.

It's almost like the AI needs to determine fort levels on offensive and defensive operations. E.g. Do I have Fort along a border? I should garrison troops there all the time no matter what. Does the enemy have forts along the line? Plan B, lets find another way to pursue enemy or build forts alongside their forts to prevent their advancements.

AI Changes:
Time for the AI to account for Forts. They are built for a reason, to be manned.
Heck, if it is too much work to adjust behavior of units, lets just give forts a local garrison. Some default strength number. That can be reinforced or improved of course.

 

[Meglok]

its a well known issue, the AI Front management is just "sad". It creates its own weaknesses (like abandoning fronts), and doesnt have the intelligence to EXPLOIT its own fault (I've seen empty capitals with loads of enemy divisions nearby, and they refuse to walk in for free). What's worse is that for naval invasions, it will do this well after securing a more than viable beachhead and taking critical ports. It abandons the ports, lets them be taken, and now the invasion force has no supply and is easily eliminated.

They promise improvements for this in 1.4.

As was promised in 1.1, and in 1.2, and in 1.3. And Africa was supposed to be fixed, and so were the garrison issues, etc etc,.

But hey, Podcat is tweeting about a new war interface so all is good.

 

[Gwydion5]

The problem is that the AI only sees the batleplanner score for the overall front and then some other calculation based on steps forward in provinces that the player cant see.

Yeah, it drives me nuts to see a set of divisions move from non-river crossing province to an adjacent river crossing province to start an attack that could have happened from the non-river crossing province. The AI is rather poor at valuations and your battleplanner explanation makes a lot of sense.