Disappointing military AI - too aggressive

[HelmuthM]

If I attack Poland, the Polish army should be there defending its heartlands. Not wandering away to besiege my vassal’s provinces in Burgundy.

Such strategies are gamey and guarantee a defeat.

Would it be possible that the AI considers properly its own military strength when making plans? Let’s face it, a good player is militarily completely superior at least 90 percent of the time.

If the AI is weak, it should DEFEND!

 

[darth254]

yeah that would be nice if they tried to minimize their own attrition and tried to counterattack on their own besieged forts or something when overmatched. rather see that than "the best defense is a good offense.....derp dee derp"

 

[fleetothemoon]

I speculate that the AI tries to take out enemy secondary war participants when it thinks it cannot directly fight against the main participant or the invading army. The idea is that by peacing out enemy allies, you will balance the overall army scale to your favor. This also helps shorten long wars, since AI allied secondary participants are far more eager to stay in the war than the primary participants, so even if you annihilate the primary war enemy, you still cant get the peace deal you want unless if you occupy their allies or peace them out.

As the player, I take this strategy too sometimes, when dealing with enemies like Austria, who have 10-12 allies, or the War of the Protestant League.

Of course, the problem here is that your vassal cannot be peaced out and will fight until they are completely wiped out, making this strategy utterly useless.

Also, I think in your scenario, you would have won either way. If the AI engaged you with their armies, they will just be stackwiped/defeated. By sieging Burgundy on the other side, they can offset some of the inevitable warscore advantage you would have gotten anyway. If you managed to siege most of their provinces without having to engage their army, then at least they get to keep their manpower/army reserve. This can, tie into the other AI strategy that comes in play sometimes - moving their troops out of harms way even if you are besieging all their land because they know that they cannot beat you in an open battle.

 

[cetvrtak]

If I attack Poland, the Polish army should be there defending its heartlands. Not wandering away to besiege my vassal’s provinces in Burgundy.

Such strategies are gamey and guarantee a defeat.

Would it be possible that the AI considers properly its own military strength when making plans? Let’s face it, a good player is militarily completely superior at least 90 percent of the time.

If the AI is weak, it should DEFEND!

Send an army their size and watch how they DEFEND!

 

[sparta105]

I think the problem lies more with the neutral AIs willingness to give military access basically on demand. The AI is being semi-smart here: why should I engage the stronger armies when I can simply go around them, wreck their country and balance out the warscore? This is alse why conditional military access shouldn't be a thing, by the way.

 

[cetvrtak]

conditional military access shouldn't be a thing

Without it armies would be able to hide in neutral territory. This was one of the most frustrating things in eu3.

 

[master_kong]

The key code for that is this:

    REGION_PLANNING_HOMELAND_PRIORIZATION = 5.1,                -- Homeland priorization when assigning armies to regions (only applied when actually threatened)

Actually, they increased this define very recently. I guess it was the latest patch. But it surely doesn't seem enough.

However, I think AI behavior isn't the only problem with defensive warfare. If AI were to prioritize homeland in most situations, they would get crushed most of the time as well. So defensive warfare isn't strong enough is as important as the former. Some potential improvements off the top of my head would be:

• Bringing back dynamic combat width dependent on the terrain. Removing this might still be one of the most single obnoxious changes in this game's history.
• Fixing unlimited reinforcement in foreign lands
• Undoing scorched earth being insignificant and teaching AI to use it etc.

Implementing regional manpower and levies would almost single-handedly improve the situation to a comfortable level although we know from the scope of such a change, it's not possible anymore for EU4. I'm sure it will be in EU5 though after seeing it's coming in Imperator and looking awesome.

Edit: Also, forgot to mention that one line code is only a define. I'm not an expert by a large margin but the real code is behind that define and we can't see it. Maybe the problem lies there instead of the define value. For example, it says only applied when actually threatened and maybe the coding part about the AI diagnosing the threat is flawed instead. I'm just spitballing here, I didn't even play properly after that value changed so maybe I could tell if I did or maybe you guys could tell the difference and diagnose the problem better.

 

[sparta105]

Without it armies would be able to hide in neutral territory. This was one of the most frustrating things in eu3.

And with it in, you either have to build forts on all of your provinces to prevent the AI to walk around your defences or you have to fight alone (and I mean actually alone: no allies, no vassals, nothing).
To me, it just goes to show how multilayered the problem is. Maybe if neutral AI is less willing to allow troop movement in its territory it could be an emergency band-aid to figure out a better solution?
I do not mean to derail the thread, but I think this is one of the main reasons why the AI is so bad at waging defensive warfare.

 

[Roland Traveler]

The key code for that is this:

    REGION_PLANNING_HOMELAND_PRIORIZATION = 5.1,                -- Homeland priorization when assigning armies to regions (only applied when actually threatened)

Actually, they increased this define very recently. I guess it was the latest patch. But it surely doesn't seem enough.

However, I think AI behavior isn't the only problem with defensive warfare. If AI were to prioritize homeland in most situations, they would get crushed most of the time as well. So defensive warfare isn't strong enough is as important as the former. Some potential improvements off the top of my head would be:

• Bringing back dynamic combat width dependent on the terrain. Removing this might still be one of the most single obnoxious changes in this game's history.
• Fixing unlimited reinforcement in foreign lands
• Undoing scorched earth being insignificant and teaching AI to use it etc.

Implementing regional manpower and levies would almost single-handedly improve the situation to a comfortable level although we know from the scope of such a change, it's not possible anymore for EU4. I'm sure it will be in EU5 though after seeing it's coming in Imperator and looking awesome.

Edit: Also, forgot to mention that one line code is only a define. I'm not an expert by a large margin but the real code is behind that define and we can't see it. Maybe the problem lies there instead of the define value. For example, it says only applied when actually threatened and maybe the coding part about the AI diagnosing the threat is flawed instead. I'm just spitballing here, I didn't even play properly after that value changed so maybe I could tell if I did or maybe you guys could tell the difference and diagnose the problem better.

Personally I would have army reinforcements be contingent on distance from the capital/core territory/owned and controlled centers of development, local infrastructure (the military buildings, supply depots, local supply limit, etc), and national modifiers. This would allow attritional warfare to actually matter, since capped max attrition means you have to slowly exhaust manpower instead of exhausting an army. In addition to putting dynamic combat width and buffing scorched earth (it is almost literally worthless for its cost, even as a tool to torch enemy lands outside of spite), this would do a lot to help defensive warfare in general and allow the AI to think about remaining in their homeland without considering it a death sentence.

That’s close to how attrition and reinforcing worked in Victoria II, and it made campaigning abroad actually dangerous instead of just dumping an army somewhere and letting numbers take care of the rest.

 

[TheMeInTeam]

The AI is avoiding your army out of cowardice, not being especially aggressive. It can be scared out of doing anything if you have enough troops to shoo it away, literally moving its soldiers into uncolonized land and hiding there if possible. They will then sit there until you take 100% war score. Even if 100% war score is game over for the AI.

Other than that I agree it should have priority on defending itself.

 

[henriqueb1]

Maybe the AI would be more successful in their offensives if they would actually finish their sieges, I've seen the AI armies lose their breachs so many time over fear of an incoming army still way under fog of war it's awful. Other than that, yeah, they should mostly defend their lands when overmatched, more fear of being cut off, scorching the frontier and try to break through without forsaking a line of defense.

 

[cetvrtak]

incoming army still way under fog of war

Note that the AI can see two provinces deep, unlike player who can see only one away from owned/controlled/allied land.

 

[Meglok]

l fail to see a problem here. If the ai refuses to fight because I outnumber their forces I use my armies to take their land with less losses to manpower. If they run around and go behind me I catch them with a reserve army. Either way I win.

The programming priority seems to be in most cases the ai will not suicide attack vs a stronger opponent. It will either attempt to attack where you are weak, snipe off captured provinces, or try to preserve it's army in case you don't annex the entire country. Given the ai options vs a much stronger foe, that is probably the best it can do.

 

[Ekcrbe]

Without it armies would be able to hide in neutral territory. This was one of the most frustrating things in eu3.

Fine. And remove the Military strength factor that doesn't go down to zero until you repeatedly stackwipe the enemy's armies from the peace acceptance calculation while we're at it. If I fully occupy you and you go hang out in Ruthenia, I win. It shouldn't matter how many troops you still have if you're not using them to, you know, do war things.

 

[cetvrtak]

Fine. And remove the Military strength factor that doesn't go down to zero until you repeatedly stackwipe the enemy's armies from the peace acceptance calculation while we're at it. If I fully occupy you and you go hang out in Ruthenia, I win. It shouldn't matter how many troops you still have if you're not using them to, you know, do war things.

Look, the game wasn't made yesterday. It's 7 years old and if in that time they couldn't figure out a better way for a crucial mechanic like this, it's highly unlikely they will ever. At least not until eu5.

 

[henriqueb1]

Note that the AI can see two provinces deep

I actually didn't knew that, I'll sure take it into consideration in my later games, but still they appear to see more than that, how many times have you guys ordered your armies to go attack a besieger army and they instantly start running away? My strategy has been to order them to a neighboring province for some years now and, would you guess it, the AI doesn't move a muscle until they can see my armies.

 

[Ekcrbe]

Look, the game wasn't made yesterday. It's 7 years old and if in that time they couldn't figure out a better way for a crucial mechanic like this, it's highly unlikely they will ever. At least not until eu5.

Oh I agree and I understand. I think the war meta is horribly off all around, with completely insignificant battles, war enthusiasm far too difficult to bring down, overly resilient secondary participants, etc. (this applies to the player too, so I'm not just saying make everything easier), but it's clear the developers either don't agree with those criticisms or are uninterested in changing them so far into the game's life cycle. It's a good game nonetheless. I do have some hope that an EU5 could be different, given that I think CK3 comes a lot closer to a good balance, but there's no telling whether that ethos carries over between the development teams. That's a different conversation in any regard.

 

[Quickmelty]

AI does not have FOW and can see the troops of nations its at war with. This counts for other AI as well.

 

[AxeBeard]

if the ai is doomed i want it to fight and die, not run away to seige siberia.

 

[Damedius]

The AI does really wonky things in general.

It declares war before their armies are in position and ready. It will send al their armies to chase an army one or two stacks can handle.

It really needs an overhaul but it's either too hard or they don't think that they can market and sell a DLC that is just AI improvements.