AI gets more braindead with every update

  • 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.

papapyro

Captain
33 Badges
Jul 11, 2017
351
694
  • Hearts of Iron IV: Expansion Pass
  • Hearts of Iron IV: Cadet
  • Europa Universalis IV: Third Rome
  • Hearts of Iron IV: Death or Dishonor
  • Age of Wonders III
  • Hearts of Iron IV: Expansion Pass
  • Europa Universalis IV: Dharma
  • Europa Universalis IV: Golden Century
  • Imperator: Rome
  • Europa Universalis IV: Mare Nostrum
  • Crusader Kings III
  • Battle for Bosporus
  • Europa Universalis 4: Emperor
  • Victoria 3 Sign Up
  • Hearts of Iron IV: By Blood Alone
  • Hearts of Iron IV: No Step Back
  • Hearts of Iron 4: Arms Against Tyranny
  • Europa Universalis IV: Cossacks
  • Europa Universalis IV: Common Sense
  • Europa Universalis IV: El Dorado
  • Europa Universalis IV: Res Publica
  • Europa Universalis IV: Call to arms event
  • Europa Universalis IV: Wealth of Nations
  • Europa Universalis IV: Conquest of Paradise
  • Europa Universalis IV: Art of War
  • Europa Universalis IV
  • Crusader Kings II
  • Heir to the Throne
  • Divine Wind
  • Europa Universalis III: Chronicles
  • Europa Universalis III
  • Crusader Kings II: Sword of Islam
  • Crusader Kings II: The Old Gods
I'm seriously considering giving up on this game until they fix the AI. Every update makes the AI worse in both economy and warfare but I think they've finally found the straw to break the camel's back. I've had all the enjoyment taken out of my Levant Turnabout campaign because the Ottoman AI just keeps suiciding smaller stacks with Insufficient Support penalty into my armies, and left a 40k+ stack sitting in the mountains letting attrition eat away at them while I sieged down the Balkans. Maybe some casual players enjoy this stuff but I don't. Beating an AI with the mental capabilities of a child is not rewarding or enjoyable
 
  • 30
  • 3
  • 2Like
Reactions:
A 5 years old child play better at Risk
 
  • 4
  • 1Like
Reactions:
I was playing Hannover. Owned all of northern Germany, Prussia and Denmark. Allied with Russia. Austria owns half of France , Hungary and all of the rest of Scandinavia as well as a good chunk of old PLC land connecting their lands.
austria pu’s Spain. Can’t have that so I Dow’d them, fighting of their hordes, theirs allies annoying nonsense stacks (this is 1750s) and Spain’s (incl colonies) armies.

Russia has a YUGE number of troops. As a war ally I got fog of war lifted on their land and after many years of not seeing anyone from Russia went over to have a look ......... fml.....
The russia AI had almost a hundred maybe more, 1 stacks, yes one stack of inf/cat/art sitting everywhere from Kiev to Vladivostok. Just f****** sitting there ! Hundreds of the little sh***. JUST SITTING THERE!!!!!!!!!!!!!! They didn’t move. For. The entire. War. WTF.... AI kill yourself
 
  • 7Haha
  • 3Like
  • 1
  • 1
Reactions:
This again. Everyone wants the AI to be at least competent, but it never seems to happen does it.
 
  • 3
Reactions:
AI did the right thing: why risk his armies on your war? :D

I would have given him Finland sweeden and Belarus :(

in the end it p*’d me off so much I gave russia (in their national ideas) 10,000 discipline and siege ability, and they still only sent one stack to war. Leaving hundreds of little regiments scattered everywhere.
 
This again. Everyone wants the AI to be at least competent, but it never seems to happen does it.

Not whilst there’s still some random ass tribe/backend area of the world to be given a dlc and mission packs, they’re own mechanics - even though /bar some post 1950s films/ they have made absolutely no contribution to world History whatsoever andif they did, only by ever so slightly slowing down European colonization. - that the AI cant get to grips with and some new random bloat to further sink the ship of a competent AI.
I dread to think where the next dlc will be: Sahara or Siberia. (Obv hopefully it will be eu5)
 
  • 11
  • 10
  • 1Haha
Reactions:
I don't want to be quite as sceptical, but I hope once PDX Tinto is properly up and running they can hire the game at least one (more?) dedicated AI programmer with external experience.
 
  • 1
Reactions:
I was playing Hannover. Owned all of northern Germany, Prussia and Denmark. Allied with Russia. Austria owns half of France , Hungary and all of the rest of Scandinavia as well as a good chunk of old PLC land connecting their lands.
austria pu’s Spain. Can’t have that so I Dow’d them, fighting of their hordes, theirs allies annoying nonsense stacks (this is 1750s) and Spain’s (incl colonies) armies.

Russia has a YUGE number of troops. As a war ally I got fog of war lifted on their land and after many years of not seeing anyone from Russia went over to have a look ......... fml.....
The russia AI had almost a hundred maybe more, 1 stacks, yes one stack of inf/cat/art sitting everywhere from Kiev to Vladivostok. Just f****** sitting there ! Hundreds of the little sh***. JUST SITTING THERE!!!!!!!!!!!!!! They didn’t move. For. The entire. War. WTF.... AI kill yourself
I can accept some bad AI decision making - that doesn't bother me as much as others here.

What I hate is when the AI goes braindead, as you are describing. If you even needed to count on them in the slightest, it can turn what should have been a very winnable war into a lost war.

Usually a save and reload will do the trick. But not always. In which case, hopefully you have a backup save...
 
  • 2Like
Reactions:
First problem: AI cant be sold as dlc, content does.
Second problem: rotation of AI programers.
Third problem: mechanics reworks.
Other problems: meta shifting (hard to program optimum every patch), dead game balance (useless ideas like religious), game design ( creating idea sets for world conqueror or normal gameplay).

The worst part is about adding new stuff into game. I warned about coastal batteries in dev diaries that AI will overuse them (this building is so situational that i dont even use it) and guess what: no reaction. New Merc system, reducing war exhaustion. Those features weren't tested or programed properly for AI. This is only tip of the iceberg...

Ps. Like you see i can only count to 3 and i still play better than AI :p.
 
  • 9
  • 1Haha
Reactions:
Not exactly true.

AI at launch used to be somewhat solid.

But then they made forts worse without telling the AI how to handle the new forts.

And then they introduced new mechanics that the AI spam for some reason

And then they introduce new buildings that the AI also spams for some reason...
 
  • 3
Reactions:
considering the fact that the AI is dead set on stopping the player from achieving any goals they'd like to accomplish? especially in iroman? I enjoy the braindead AI that stops working after it realizes i declared a war that it can't possibly win.
 
  • 1
  • 1Like
Reactions:
Not exactly true.

AI at launch used to be somewhat solid.

But then they made forts worse without telling the AI how to handle the new forts.

And then they introduced new mechanics that the AI spam for some reason

And then they introduce new buildings that the AI also spams for some reason...

How did forts used to work and why are they worse now?
 
On release, there was no fort, or no fort+ZOC (I don't remember exactly). So you had to carpet-siege during a war to get warscore.
Forts make more sense tactically speaking but the ZOC is very difficult to understand
 
  • 1
Reactions:
How did forts used to work and why are they worse now?
Early on every province had a mini-fort and you could siege them faster. There was no Zone of Control (ZOC) so a lot of wars involved beating the AI Doomstack, then running down the AI armies, and then carpet sieging the place (using a lot of covering regiments to stop the AI from building new troops).

The devs thought it would be better if the AI had time to regroup after a failed hit, so they made forts beefier and gave them ZOC so that in theory the AI could retreat past some fort and you could not follow. Unfortunately, rather than doing something simple (like forts act historically to slow the enemy, like a 4x transit multiplier, and harass them, like a 2x attrition multiplier) they opted for a magical "thou shalt not pass" setup. Then then hiked fort upkeep to make so you had to place forts in strategic locations.

The AI, has ever since had to deal with the problem of fort pathing. If the province they want to siege (for wargoal or whatever) is past a fort do they siege or walk further around? Well humans know it depends. You want to skip around the fort if you can do so by just getting MA through somebody else for an extra province or two. However, if you have to walk around the Black Sea it is likely faster (and cheaper) to just siege the first fort before going after the next.

Likewise, where to build forts is not obvious to the AI. AI forts will often leave gaps through which you can maneuver and they will often leave prime chokepoints (e.g. the passes in the Alps) open. Let alone doing better stuff like having your forts on the flat for early game Hordes or forting up some clutch straight landing. Instead the AI pretty much always has its starting forts and those it gets through conquest. Worse, the AI is terrible at upgrading them as technology rolls on so only the wealthiest places (e.g. Italy) have up-to-date forts.

Then we have all the idiocy with mothballing. Even in AI-AI wars all too often you have AIs drilling their units on the border with the forts unmanned. An AI declares war (maybe because it allied with somebody who then declared and promised land) and the AI has a stack-wipe from the morale penalty on drill and has their fort occupied. Worse still, armies will march to wherever the initial order specified, even if the now manned forts prevent reinforcements from following them in or if they will get locked after seizing their target province with no choice but to lay siege to a massive fort in the Siberian winter.

To whit, at game launch forts were simple (too simple to really slow anyone down), but they were implemented in a complicated fashion that AI sucks at parsing. This has offset a lot of AI improvements elsewhere. And AI algorithms have never kept up with the changes. The AI's pathing, strategic value weighting, and dynamic recalculation have never matched what the fort system actually does. Of course, with all the new AIs and all the new calcs for all the old AIs with every patch, it becomes that much harder to do anything about it.

As it stands, EUIV forts are an ahistorical abomination that are a giant nerf to the AI.
 
  • 8
  • 5
  • 3
Reactions:
Early on every province had a mini-fort and you could siege them faster. There was no Zone of Control (ZOC) so a lot of wars involved beating the AI Doomstack, then running down the AI armies, and then carpet sieging the place (using a lot of covering regiments to stop the AI from building new troops).

The devs thought it would be better if the AI had time to regroup after a failed hit, so they made forts beefier and gave them ZOC so that in theory the AI could retreat past some fort and you could not follow. Unfortunately, rather than doing something simple (like forts act historically to slow the enemy, like a 4x transit multiplier, and harass them, like a 2x attrition multiplier) they opted for a magical "thou shalt not pass" setup. Then then hiked fort upkeep to make so you had to place forts in strategic locations.

The AI, has ever since had to deal with the problem of fort pathing. If the province they want to siege (for wargoal or whatever) is past a fort do they siege or walk further around? Well humans know it depends. You want to skip around the fort if you can do so by just getting MA through somebody else for an extra province or two. However, if you have to walk around the Black Sea it is likely faster (and cheaper) to just siege the first fort before going after the next.

Likewise, where to build forts is not obvious to the AI. AI forts will often leave gaps through which you can maneuver and they will often leave prime chokepoints (e.g. the passes in the Alps) open. Let alone doing better stuff like having your forts on the flat for early game Hordes or forting up some clutch straight landing. Instead the AI pretty much always has its starting forts and those it gets through conquest. Worse, the AI is terrible at upgrading them as technology rolls on so only the wealthiest places (e.g. Italy) have up-to-date forts.

Then we have all the idiocy with mothballing. Even in AI-AI wars all too often you have AIs drilling their units on the border with the forts unmanned. An AI declares war (maybe because it allied with somebody who then declared and promised land) and the AI has a stack-wipe from the morale penalty on drill and has their fort occupied. Worse still, armies will march to wherever the initial order specified, even if the now manned forts prevent reinforcements from following them in or if they will get locked after seizing their target province with no choice but to lay siege to a massive fort in the Siberian winter.

To whit, at game launch forts were simple (too simple to really slow anyone down), but they were implemented in a complicated fashion that AI sucks at parsing. This has offset a lot of AI improvements elsewhere. And AI algorithms have never kept up with the changes. The AI's pathing, strategic value weighting, and dynamic recalculation have never matched what the fort system actually does. Of course, with all the new AIs and all the new calcs for all the old AIs with every patch, it becomes that much harder to do anything about it.

As it stands, EUIV forts are an ahistorical abomination that are a giant nerf to the AI.

AI sucks for sure. But the carpet siege system is far from being a good alternative. The not-so-new ZoC system made wars much more strategical. What needs to be done is improving AI, instead of going back to a more primitive system.
 
  • 6
  • 2Like
Reactions:
If the ai is being braindead I always save and reload. That resets the ai_strategy and usually causes the ai to react. Sometimes the various ai sub-sections get caught in a loop in the engine.

Sometimes that doesn't work, usually due to some other factor or ai priority overriding ally support. A mission goal, low manpower, another threat, a flock of geese crapped on the forum as they flew over and the auspices aren't good, the ruler has gout and can't be bothered, sometimes you just have to suck it up and realize your ally is worthless.
 
  • 1
  • 1Like
  • 1
Reactions:
AI sucks for sure. But the carpet siege system is far from being a good alternative. The not-so-new ZoC system made wars much more strategical. What needs to be done is improving AI, instead of going back to a more primitive system.

There were many alternatives to carpet sieging that would be easier for the AI (or even humans) to parse. Many of these would be far more strategic in nature and historical to boot.

For instance, suppose we made forts allow enemy armies to move through, albeit it very slowly. This would make them actually follow Napoleon's maxims on fortifications. It would allow defeated armies time to withdraw and regroup (and recruit new regiments). We could, in theory, increase the time penalty to cross the fort and surrounding provinces to the point that it is comparable to the time a mid-range siege army would take to get through. In theory, this results in the same sort of primary functions, but would make it easier on the AI to calc. For instance, right now it matters massively when and how you merge armies. Do you keep the retreat path to the sea or back to home territories? Do you keep your armies separate with two options for exit or merge them to share attrition reduction? A lot of this would be gone and the AI could instead focus on how to actually move armies. We could even have forts slow transit through neighboring provinces even if they are owned by the enemy (an original goal of ZOC forts that failed).

Ultimately some tasks are harder for AI. Looking at a uniform field that uses an algorithm they AI already works on maximizing? Easier. Adding in something new that is highly dynamic, anisotropic, and requires either continuous recalculation or a lot of up front computation about potential states? Harder.

ZOC is exactly the sort of breakpoint system that AI will always have a hard time managing.
 
  • 2
  • 2
  • 1Like
Reactions:
Russia hasnt declared war Novgorod even once in my games. Also didnt declare on any horde. Muscovy is still the same size in 1444 as in 1540. The solution to the debt problem was making the AI never fight any war apperently.
 
AI sucks for sure. But the carpet siege system is far from being a good alternative. The not-so-new ZoC system made wars much more strategical. What needs to be done is improving AI, instead of going back to a more primitive system.

Carpet sieging doesn’t really involve that much more finicky micro than the fort “system”, if you’re playing anything close to efficiently, and that’s before you take into account the need to wrestle with the obtuse and illogical ZoC rules.
 
  • 1
Reactions: