Aggressive AI is pacifist

[Nominus]

Even with AI set as "aggressive", I have yet to see Great Britain attempting to wage war against the Great Qing Dynasty for a treaty port or to open their markets, just for example.
Especially late game the AI ironically is afraid of "great wars".
On the other hand, the AI loves to join minor revolutions/civil wars.

What are your experiences with AI behavior?

 

[MfgLuckbot]

To me many things feel a bit random. Guess would be nice to see at least a broad breakdown of how AI makes war decisions (I understand they don't want to reveal the point values to avoid gaming the system, but they could at least list all the pros and cons that the AI considers, like it was in CK2)

I find it especially hard to foresee the effect of infamy. I had GP allies switching to the opposing side in high infamy plays, and sometimes they stay on my side in the same play.

 

[Vernichtere]

Use mods. Then the AI becomes aggressive. The problem then is that they don't know how to win wars: 300,000 without supplies on any islands instead of on the main front.

 

[MathyM]

they don't know how to win wars

It’s comically easy to defeat Egypt as the Ottomans because of this.

Attack them from all fronts, and they will send every man they have to defend themselves, leaving their capital wide open to a naval invasion. You can take the entire Lower Egypt, plus two or three states, in the time it takes them to send a general over there. By then, they will be near capitulation.

This works with pretty much any country whose capital is near the sea. France has an advantage here.

 

[Vernichtere]

It’s comically easy to defeat Egypt as the Ottomans because of this.

Attack them from all fronts, and they will send every man they have to defend themselves, leaving their capital wide open to a naval invasion. You can take the entire Lower Egypt, plus two or three states, in the time it takes them to send a general over there. By then, they will be near capitulation.

This works with pretty much any country whose capital is near the sea. France has an advantage here.

It doesn't even have to be the capital. The current system is such that it is always beneficial to perform a sea landing. Troops don't drown if they fail anyway. And the AI gets more overwhelmed with each additional front. For the same reason, minions are sometimes kept on the border of strategic opponents. As Russia, I'd rather have Mongolia and Sinkiang alive for a while, and thus three fronts, than one forever long front. There are three battles going on at the same time and the Ki is constantly moving troops back and forth.

 

[DukeLeto42]

I've never not seen the Opium Wars occur, so I don't know what's going on at your end.

 

[Opanashc]

I've never not seen the Opium Wars occur, so I don't know what's going on at your end.

About half the games, UK starts the opium war for me.

 

[Mithkabob]

Even with AI set as "aggressive", I have yet to see Great Britain attempting to wage war against the Great Qing Dynasty for a treaty port or to open their markets, just for example.
Especially late game the AI ironically is afraid of "great wars".
On the other hand, the AI loves to join minor revolutions/civil wars.

What are your experiences with AI behavior?

Not sure if they are afraid of great wars, it's just that half of the major powers will be in permanent bankruptcy by 1900.

I'm guessing on aggressive it's even more likely they will bankrupt themselves at some point.

UK/France usually end up hitting an issue and deleting all of their ports in the mainland which breaks their market and economy.

Prussia/Austria/Russia/US/etc. will use field hospitals in conscription centers without regard for the lack of opium in their markets, so they will bankrupt themselves and end up with 50% shortage penalty in their armies the moment they fight a war when they raise conscripts. (same with recon methods, bikes will use up all rubber, motorized/aerial will use up more autos/planes/oil than they have).

Once they hit bankruptcy and spawn some revolts, it's unlikely the countries will recover.

The AI is way too willing to raise their entire army including conscripts and send them to some front in Africa and have millions of people die from attrition, bankrupt themselves, and achieve nothing from the war.

 

[steefanSS]

Or maybe AI is just better understanding at what is going under the hood (and game needs broader rebalancing rather than purely just changing agressive/passive stance - which can be done through simple scripting as someone already pointed out), the one we are struggling to understand, or gather enough facts and there is randomness factor (and RNG is real in this game). Because as you can see here, it happens, and the AI is set to lenient in my playthrough because I am still learning game.

 

[Nominus]

View attachment 931940


Or maybe AI is just better understanding at what is going under the hood (and game needs broader rebalancing rather than purely just changing agressive/passive stance - which can be done through simple scripting as someone already pointed out), the one we are struggling to understand, or gather enough facts and there is randomness factor (and RNG is real in this game). Because as you can see here, it happens, and the AI is set to lenient in my playthrough because I am still learning game.

Glad to see this. But which year is it?
I bet it isn't late game.

Not sure if they are afraid of great wars, it's just that half of the major powers will be in permanent bankruptcy by 1900.

I'm guessing on aggressive it's even more likely they will bankrupt themselves at some point.

UK/France usually end up hitting an issue and deleting all of their ports in the mainland which breaks their market and economy.

Prussia/Austria/Russia/US/etc. will use field hospitals in conscription centers without regard for the lack of opium in their markets, so they will bankrupt themselves and end up with 50% shortage penalty in their armies the moment they fight a war when they raise conscripts. (same with recon methods, bikes will use up all rubber, motorized/aerial will use up more autos/planes/oil than they have).

Once they hit bankruptcy and spawn some revolts, it's unlikely the countries will recover.

The AI is way too willing to raise their entire army including conscripts and send them to some front in Africa and have millions of people die from attrition, bankrupt themselves, and achieve nothing from the war.

I agree, that's probably the explanation. I wouldn't go to war while having turmoil and going bankrupt either.

 

[steefanSS]

It's not late in the game, so it definitely lines up with your speculation that the decision is economy driven.

Although, I do have AI economy focused mod, so I am curious to see how that plays out (and it's also 'role play' enabled, which should in theory, emulate behavior that is historically more accurate

 

[Question]

Ive been trying to mod the AI and its worse than herding cats. The AI is significantly worse than EU4. In EU4, i was able to get the AI to behave very aggressively because the AI has good logic built in...they are smart enough to wait till someone is fighting a war to start a second front to take back their cores for example.

In vic 3, despite cranking aggression up, i cannot get the ethiopian minors to fight each other or japan to attack the poor defenceless ainu state. They just sit there with belligerent attitude towards each other, even if one side is vastly superior. I cannot figure out why.

 

[steefanSS]

Significantly worse... Or just significantly different.

You don't see half of the things going on, just what is exposed to an extent to mod for modders.

How exactly can you know this things when you don't have same diagnostic and debugging capabilities?

I am not trying to be here smart ass, just as a software engineer. You don't have a full agency over what you are scripting, rather just putting a different paint over what you are allowed to manipulate and see if the things are going to play out the way you want it.

So, it's not about which AI is worse or better you took EU iv as an example, I won't even bother commenting anything on that note, but it's just very different model from one another or rather EU IV is more deterministic while VC3 leans more towards nondeterministic model (from everything so far I saw, trying to be as close to cognitive design model as it can get and fails in many areas, which is to be expected against the scale of this product) whatever atever this was good decision or not, time will tell but in the meantime it's really up to devs to open AI more to scripting and shift behavior one way or another...

 

[Question]

My point is that the built in EU4 AI is much better than the vic 3 AI. This is very obvious if you compare both. Of course you can script the vic 3 AI more, but that means that you have to sit down and do an insane amount of work on your own and you dont have access to tools that will tell you "okay the AI isnt doing X because of Y, so i need to tweak X". All you can see is that the AI isnt doing X but not why.

EU4 comes with better AI built in so it takes much less effort to get it working properly. If you want the devs to do it, expect to wait several years minimum.

 

[steefanSS]

My point is that the built in EU4 AI is much better than the vic 3 AI. This is very obvious if you compare both. Of course you can script the vic 3 AI more, but that means that you have to sit down and do an insane amount of work on your own and you dont have access to tools that will tell you "okay the AI isnt doing X because of Y, so i need to tweak X". All you can see is that the AI isnt doing X but not why.

EU4 comes with better AI built in so it takes much less effort to get it working properly. If you want the devs to do it, expect to wait several years minimum.

I am confused now.

I asked you how can you tell which one is worse and not significantly different because you lack debugging capabilities and full agency over what you can do in either of the games and you basically repeated what I said and then some I didn't even mention.

Okay, I might have been overly aggressive and for that I am sorry, but really bro I didn't say wait for devs to fix stuff... AI is very complex thing and they had their reasons for doing things the way they have and different from EU IV. Although, they possibly have hanged themselves with that decision but the time will tell, EU IV Ai wouldn't work in this game and vice versa.

AI in this game is influenced by VC3 mechanics that most really hate, primarily warfare which is intertwined in one big loop and depended on one another much more than in most PDX titles. So game in a grand scale need brush polish, stroking AI in this or that corner won't do a thing (and that's why devs have to tank this one for the time being)

Relax a bit. My point was, that we need to wait on the game to get seriously patched whatever you like that or not.

 

[Question]

I am confused now.

I asked you how can you tell which one is worse and not significantly different because you lack debugging capabilities and full agency over what you can do in either of the games and you basically repeated what I said and then some I didn't even mention.

Okay, I might have been overly aggressive and for that I am sorry, but really bro I didn't say wait for devs to fix stuff... AI is very complex thing and they had their reasons for doing things the way they have and different from EU IV. Although, they possibly have hanged themselves with that decision but the time will tell, EU IV Ai wouldn't work in this game and vice versa.

AI in this game is influenced by VC3 mechanics that most really hate, primarily warfare which is intertwined in one big loop and depended on one another much more than in most PDX titles. So game in a grand scale need brush polish, stroking AI in this or that corner won't do a thing (and that's why devs have to tank this one for the time being)

Relax a bit. My point was, that we need to wait on the game to get seriously patched whatever you like that or not.

I can tell because ive modded the AI in both games. Cranking AI aggression up in EU4 along with some sanity checks for diplomacy gets you this :

Cranking the AI aggression up in Vic 3 just gets you GPs declaring war on unrecognized minors over and over. I even tried scripting Japan to attack Ainu specifically and they stilll refuse to touch them even in 1890. Ethiopian minors still ignoring each other.

 

[steefanSS]

I can tell because ive modded the AI in both games. Cranking AI aggression up in EU4 along with some sanity checks for diplomacy gets you this :

View attachment 932234

Cranking the AI aggression up in Vic 3 just gets you GPs declaring war on unrecognized minors over and over. I even tried scripting Japan to attack Ainu specifically and they stilll refuse to touch them even in 1890. Ethiopian minors still ignoring each other.

I can tell because ive modded the AI in both games. Cranking AI aggression up in EU4 along with some sanity checks for diplomacy gets you this :

View attachment 932234

Cranking the AI aggression up in Vic 3 just gets you GPs declaring war on unrecognized minors over and over. I even tried scripting Japan to attack Ainu specifically and they stilll refuse to touch them even in 1890. Ethiopian minors still ignoring each other.

This didn't really gave answer to my specific question, however, I can go by this and respect the effort and your verdict - even though we may not align on everything here.

For the record, I understand your frustration and all those hitting "disrespectful disagree" at least give some context behind it. It's a bit passive aggressive not contributing to anything.

No one needs to agree with me and we can have healthy conflicting discussions that can lead to a useful feedback, just saying.

 

[Question]

Not sure what the confusion is, if you have played both games, its pretty obvious that EU4 has better AI built in and its much easier to mod into a working state.

Another example is building weights. In EU4, you set buildings to a high weight, the AI builds them, this is very straight forward and works perfectly (just dont use too high of a number or it will overflow). In Vic 3, you set a high weight, and the AI will ignore it, no idea why.

Also, i noticed something interesting with the AI starting diplo plays. They seem to do so in a streak. For example, I had modded Indian states to be very aggressive towards each other. They became independent, and nothing happened for more than 20 years....then I saw East India Company starting diplo plays against a few Indian minors in rapid succession. I thought it was finally working...then I restarted the game and now they are back to sitting there twiddling their thumbs.

STILL cant get Ethiopian minors to fight each other as well.

 

[steefanSS]

What I noticed about VC3 so far, in my multiple playthrough is that AI tends to have a fixed dosage of randomness that is weighting down other options. Now what a heck that random element is and what form it has that only developers know for sure, and your example above of trying to get things working is what is for me the most frustrating part.

I think that what is more frustrating than anything, we don't have a good documentation to go by and lots of it is a guess work. One of the reasons, I haven't put effort into trying myself is being unable to actually train the models myself (if they are reinforcement learning it's all guess work, as I said), or really can directly work on heuristic algorithms AI is using (which is probably more of the case) . Instead, anyone who wants to give it a go have to relay on restricting, loosely document, exposed bits of API through tools that are the least bit useful for my taste.

And devs are quiet on that front, so I bet it's going to be a lot of what you are doing so far. Semi-success, due to elements you can't figure out because they are obscured by who knows what, without means to dig them out.