AI in war in this patch is even dumber than before

[tip001]

I play a lot of spectate games, especially with Jan Mayen. I noticed some changes to AI behavior which are very bad regarding sieging and confronting enemy troops.
Example: 150K troops of whom 3 sieging a fort. We have 2K enemy troops wandering in sight. All 150K troops try to engage the 2K "stack" and leave their sieges. This was not an issue before, one sizable stack would engage and the rest will remain sieging.
This means wars take at least twice as long as before!
This is in latest patch.
Let me know if you have same experience. This will make AI against player much softer as the headless chickens are even more "headless" than before.

 

[Kingless]

I've had this against me the player, and was not in any way prepared for it. Try to lure one stack away to fight, instead lured 36k towards myself and my allies. It was not fun.

 

[Res_Ipsa]

Similar experience.

Also, AI Ottomans has been sitting at 99% WS with Mamluks for about a decade. Has war goal and capital. Refuses to peace out or siege new provinces. Rebels are popping--from WE, I would bet, even though it's on VH, and they still refuse to peace out or even advance the war. WTF.

 

[durbal]

I've had this against me the player, and was not in any way prepared for it. Try to lure one stack away to fight, instead lured 36k towards myself and my allies. It was not fun.

If only the AI could be taught to leave a 1K stack on forts and bring the rest to the fight then we'd be golden.

To me it sounds like the battle AI is less susceptible to tricks.

 

[TheMeInTeam]

AI once again will try to attack a siege while reinforcement units are adjacent, then cancel.

You can repeat this process to take forts without fighting the AI in some cases. Very manpower efficient, but not ideal behavior from an opponent.

 

[Chaingun]

Upload save games please. I have some clue why behavior OP describes happen having briefly glanced over code changes since my departure, but it would be needed to be able to verify that I fix it...

 

[dunyayi kurtaran adam]

I play a lot of spectate games, especially with Jan Mayen. I noticed some changes to AI behavior which are very bad regarding sieging and confronting enemy troops.
Example: 150K troops of whom 3 sieging a fort. We have 2K enemy troops wandering in sight. All 150K troops try to engage the 2K "stack" and leave their sieges. This was not an issue before, one sizable stack would engage and the rest will remain sieging.
This means wars take at least twice as long as before!
This is in latest patch.
Let me know if you have same experience. This will make AI against player much softer as the headless chickens are even more "headless" than before.

This is why AI OTTOMANS end up the game with 2x more dead from attriiton than the second one. AI is sieging one province with 2x-3x more soldiers than the supply limit. Hanging around in Anatolia with 50-60K stacks is like a suicide. Going from Izmir to Tabriz with 50k stacks make you lost at least 15-20k manpower i guess..

 

[tip001]

This is why AI OTTOMANS end up the game with 2x more dead from attriiton than the second one. AI is sieging one province with 2x-3x more soldiers than the supply limit. Hanging around in Anatolia with 50-60K stacks is like a suicide. Going from Izmir to Tabriz with 50k stacks make you lost at least 15-20k manpower i guess..

View attachment 315417

Yeah, the attrition is huge too. In part because even big stacks are going wack-emo from siege to chasing a single inf. Also wars take much longer (or are even lost 4:1) causing higher attrition.
All because the AI dev at Paradox apparently doesn't see "spectate" a part of the test cycle but starts a game on observer and leaves the office. Seriously, just watching AI a couple of hours and its clear the patch sucks regarding AI war behavior

 

[tip001]

Upload save games please. I have some clue why behavior OP describes happen having briefly glanced over code changes since my departure, but it would be needed to be able to verify that I fix it...

Ah, thx for reply. Try to ignore my remark just a minute ago regarding testing I will check on a save file. Thx again, its important i think

Edit: It seems I overwrote all saves with an insignificant custom run :-( Sorry. But its easy to see if you start a Jan Mayen game or any big countries and speed 5 for 10 years. If you like I can make a new run but it takes a couple of hours

Edit2: Dear Chaingun ,
I included a save file of Ming with first war against Mong Yang and Ava (I hope its the correct file as I am a novice in these things).
-near 10:1
-war took 8 years :O It took Ming also 8 years to siege the capital of Mong Yang down.
-Ming lost 250K of which 195 attrition, Mong Yang and Ava lost about 50K
-Ming ended up with 9 war exhaustion! The country was exploding with rebels.

Hope this helps sir.

 

[tip001]

Upload save games please. I have some clue why behavior OP describes happen having briefly glanced over code changes since my departure, but it would be needed to be able to verify that I fix it...

Just in case you are not flagged for an edited reply here a new one. I put an example file as you requested in the edited reply above. Regards.

 

[Chaingun]

Just in case you are not flagged for an edited reply here a new one. I put an example file as you requested in the edited reply above. Regards.

Thank you! In the future, please post in the bug report subforum as otherwise QA will refuse to look at it, but in this particular case it's not necessary as I'll be continue working with AI army behavior next week.

 

[dunyayi kurtaran adam]

Thank you! In the future, please post in the bug report subforum as otherwise QA will refuse to look at it, but in this particular case it's not necessary as I'll be continue working with AI army behavior next week.

how about changing ottoman national idea (ship costs or stability reduce cost) with -10% land attrition?
say yes make my day