Why the A.I. being in such a crappy state isn't seen as a serious issue for this game?

[Shebaloso]

For me, the biggest culprit in this supposed "deficient AI" is an intrisic powergaming tendency most people seem to have. While I do agree that there are generally a couple of areas in which we could use an improvement in the AI, the problem is that the character side of gameplay is an utterly irrelevant issue for the player, but means everything to the AI. As someone said, although Ai's behaviour is RNG, it's modulated by its traits, so it's by no means random. When an ambitious zealous Holy Roman emperor dies and a content and kind duke inherits the empire, there actual modifications on how you deal with the emperor as a sunni, for example. There are palpable changes in how you interact with him. And "history" altogether might change because, for example, the holy wars the the previous empire had been "planning" are probably not going to happen now.

But if the same change happened to the player, there would be NO impact whatsoever on gameplay. A player will most often than not not even bat an eye if they're torturing someone while having the kind trait, or if they are content and sympathetic towards islam while trying to conquer the whole middle east. Traits are simply modifiers that function as barriers to the player. A bad military skill with a good learning skill does not mean that the player will spend more time on learning focus and hermetic events. It simply means that he will have fewer soldiers to work with in the wars he was planning while controlling the late brilliant strategist.

For example, i always see players complaining about gavelkind and how it fucks up inheritance. But actually why are players so afraid that we end up after the death of charlemagne in a situation like the 867 start, all of his children inheriting and being indepedent?

My point is that people should actually treat the traits as something that matter and guide character behaviour. Painting the map is really not this game's greatest strength, and much of the original criticism simply boils down to the fact that the character-based nation management represents a truncated chain of priorities for the AI, while remaining perfectly stable for the player.

You have issues with the AI OP. So do I. Though in my case the real question isn't why the AI is so bad at snowballing. That's the wrong question to ask.

The more important question is why the AI is so bad at causing empires to collapse. Recall this is the medieval ages. Past a certain size no empire survived intact for more than a century, two at most (China being the exception.)

Yet the AI is terrible at making the player's empire collapse, let alone another AI's.

You could never model "what made empires fall" in CK2 because the problem has one simple name: Perfect information. You always know with 100% accuracy what the intentions are of ALL the characters in the game towards you (you know the exact numerical value of one's opinion), you know exactly why a character is mad and what you can do to improve his opinion, you know exactly where the enemy's troops are, how much much money they have, how loyal his vassals are to him etc. And there's even the even larger issue that is being able to raise levies at all times and move them back and forth with ease and with no issues whatsoever. The HRE is able to muster an entire army of 80 000 people to fight over a single county in a de jure war and such armies can move through provinces with not a single consequence, like devastation, peasant unhappiness, violence etc....this is not how it should work at all if you want more realistic mechanics when dealing with empire instability.

 

[Malprau]

Yea but you want the OUTCOMES to be what you want and not what might happen. That is control no matter what you call it. It seems only the bad outcomes is why you think the AI is bad. You ignore the good choices and outcomes.

That is not completely true, Im p. sure hes not arguing that, I think hes actually arguing that the outcomes from the A.I.s decisions should actually be what might happen, but the outcomes of "what might happen" have to be at least minimally reasonable, irrationality is ok too, but as long as it serves something outside of sheer stupidity as it currently does(and I stand by saying this).

Theres very few instances in history where the big cardinal decisions(like declaring a war) were 100% irrational and whose outcome/consequence was a complete total catastrophe, most of the time irrationality was useful in some indirect way for the decision maker, even if it's consequence was majorly bad.

Having a 3 realm size feudal country decide to war u out of nowhere with their 1.5k levy troop + 500+ gold wasted on 3k mercs vs ur 15k army, just makes the weaker countries more weaker with time, and u stronger, (u get easy gold for winning low effort war= gold= power) cause of such retardedly unreasonable decisions and this happens way too often to be tolerated, especially mid to late game and yet people dont really mention this as something bad and unacceptable, cause apparently avg. player is content that hes playing vs absolutely incompetent A.I.cause it's working as designed?

I just dont see why exactly would this be ok for a game like this. And bare in mind im not by any means a "power player" or a "chess master" or something like that, I just want to enjoy the game and its hard when u know u gonna have easy low effort time everytime u start a new scenario. It just doesnt feel right.

 

[jere8184]

Yes many ruler's in this era where idiots.

Ck2 is a GAME and when I play a GAME I want to be challenged.

I think right now the ai in terms of diplomacy is as good as its going to get. The only adjustments I would like to see is the ai trying to find better marriages for their children as sometimes its just ridiculous. (I think landed heirs don't know their value marriage wise and so settle for poor marriages but I could be wrong)

Where the ai is really flawed is in how it conducts warfare. I think there is alot of room for improvement here in terms of army management. But compared to other paradox games the warfare ai is pretty good.

Overall I dont think the ai is too bad, there's alot of room for improvement. But the people acting as if the ai as perfectly fine and does not need work confuse me.

 

[GundamMerc]

Yes many ruler's in this era where idiots.

Ck2 is a GAME and when I play a GAME I want to be challenged.

I think right now the ai in terms of diplomacy is as good as its going to get. The only adjustments I would like to see is the ai trying to find better marriages for their children as sometimes its just ridiculous. (I think landed heirs don't know their value marriage wise and so settle for poor marriages but I could be wrong)

Where the ai is really flawed is in how it conducts warfare. I think there is alot of room for improvement here in terms of army management. But compared to other paradox games the warfare ai is pretty good.

Overall I dont think the ai is too bad, there's alot of room for improvement. But the people acting as if the ai as perfectly fine and does not need work confuse me.

I don't play CK 2 for the challenge, I play it for the RP. I don't get this idea where a game HAS to be challenging to be fun. Having the AI be too good would in my opinion ruin much of the opportunity to play characters suboptimally just to have fun. Seriously, half the fun of CK 2 is just doing crazy things that aren't exactly the "good" thing to do, but are the fun thing to do.

Honestly as soon as you add AI that is always competent, then you inhibit the ability to play your characters according to their traits. It kills what makes CK 2 fun for me.

 

[Karlington]

And this is where I really disagree. History is filled with what are irrational decisions by leaders. But here's why. Something is irrational simply because we know the outcome. The game gives us WAY too much information. Yea it's stupid for 2K troops to attack a land with 10K. But what if that data point was unknown? Seldom did the real world leaders KNOW what the other side had till they met on the battlefield or learned via other sources. There are way way more than a "very few instances" in history situation. Not sure if it wasx this thread or antoehr where I cited a number of rulers that did absolutely crazy decisions during their reign.

This is true, but in the game both we and the AI do know each other's strengths, so to make decisions as if we don't know when we really do know would be highly irrational indeed.

It'd be different if there were a CK2 game rule that allowed us to play without knowing exact numbers, with perhaps a spy mechanic and technology added that would allow to gain more and more information about the other realm's strengths. Counter-espionage abilities and technology might even be added so that one could hide this information from other lords' spies, or even feed them misinformation in either direction.

But I think a major change like that, which is only likely to be used by a small minority of the playerbase, is very unlikely. Would be fun, though.

From a pure game perspective you are right. We have this knowledge of how good someone might be. But again what if those were hidden or not known? In history most marriages were picked between rulers of around the same rank marrying off their kids. They could be complete idiots but a son of a king and a daughter of another king still got married. That's why plots were so important in real life. They eventually found out the person was a complete loser so they had to get rid of him to save the kingdom.

Same comment here as above.

Now your comment that you want to be challenged is very personal. You need to define what that would be for you because it will be different for other players.

Ageed. See for example @GundamMerc's comment above. I also play more for the RP than the challenge at this point. It was challenging when I first started out. Once you learn the system of the game, it becomes much easier, and with more experience it gets to the point where anything but the most suicidal starting point becomes easy to build from.

 

[Naughtius Maximus]

You could never model "what made empires fall" in CK2 because the problem has one simple name: Perfect information. You always know with 100% accuracy what the intentions are of ALL the characters in the game towards you (you know the exact numerical value of one's opinion), you know exactly why a character is mad and what you can do to improve his opinion, you know exactly where the enemy's troops are, how much much money they have, how loyal his vassals are to him etc. And there's even the even larger issue that is being able to raise levies at all times and move them back and forth with ease and with no issues whatsoever. The HRE is able to muster an entire army of 80 000 people to fight over a single county in a de jure war and such armies can move through provinces with not a single consequence, like devastation, peasant unhappiness, violence etc....this is not how it should work at all if you want more realistic mechanics when dealing with empire instability.

While perfect information is true, we can make the vassal much more unruly, enough that at a certain size they will form unmanageable factions that can break up your empire. So that even if you know he will revolt and at what strength, at some size/point you can't stop it. What I am proposing here is to change the limiters set on the AI. AI is prevented from doing a lot of things that the player can exploit with perfect information. My suggestions below are here to change those limits, so that we stop seeing realms larger than HRE stay intact for centuries.

We can handle borderline exploits first. These are nebulous for me, just suggestions to try and mitigate north korea mode.

1. Allow abdication at will when imprisoned, typically taken after imprisonment. This will mitigate north korea attempts to imprison every vassal.

2. Speaking of, an outight negative income, army size, prestige, piety, and increased retinue maintenance once past say, 10 over desmense. Partly for the same reason, the other to mitigate north korea tribal retinues.

2a. Retinue may revolt if unpaid. Similar to a merc company declaring war and invading your lands if you cannot pay.

3. Tyranny also places a +5% revolt chance and +20% revolt strength on all provinces in the realm per -10 tyranny opinion modifier. Stacks.

4. Intrigue focus no longer finds random reasons to imprison a vassal. Only gives a higher chance to discover any plots or secret societies/religions the target may be under.

Now for the main point. For making vassal management a more empire breaking affair regardless of perfect information.......

1. First thing's first. AI does not treat nonaggression pacts as permanent features. If they hate you enough, at say, -20, they're liable to break the pact. The current setting of AI treating pacts as permanent is often exploited by players below.

2. Better yet, non aggression pacts not stopping factioning should be the default option. It's much too easy to lock a huge vassal or two comprising 50% of the realm from being unruly by sacrificing one child for marriage purposes. If pacts did not stop factioning, not even perfect information would save you should this vassal hate you. At the moment once you have this pact you can do as you please no matter how much you antagonize your vassals, as you have 50% of the realm locked from factioning.

3. At some point in a huge empire you/AI are going to need some vassal kings. There's just no way around it with the vassal limit. Well there is, but it's very tedious to make supercounts under dukes and still be prevented from forming a kingdom, let alone another duchy. As vassal kings are inevitable, make vassal kings unruly. They are kings and would naturally prefer independence. Kings are eligible for independence factions regardless of de jure along with a steeper opinion modifier to get them to reconsider (the old 80 opinion will do.)

4. War Council nerf. Instead of outright banning factioning should war council be enacted, lower the threshold from the current 50 to say, 30 opinion. That's still a very significant difference. You would need to **** up pretty hard to get a -20 modifier amongst your long reign and prestige bonus.

This one is really an option. One or the other, I do not suggest both.

5a. Remove stacking effects on artifacts. Way of Life was already pretty bad regarding stat inflation. The introduction of artifacts and societies blew that right out of the water, and these days you can grab massive diplomacy, prestige, and opinion bonuses even from your typically bad starting ruler. 50 opinion modifier to keep vassals from factioning is a joke to reach these days. Keeping these from stacking would go a long way to some stat deflation for once.

5b. As mentioned above. 50 opinion modifier is a joke to reach these days amongst all those stats we get. Set it back to 80.

That's for factioning. To encourage internal breakups, we should increase mechanics for when a faction or claimant/adventurer does succeed.

1. Losing/neutral side of the claimant faction gets a prompt upon loss. They break off and become independent. Old liege gets a strong claim on the top titles of the guy breaking away. Kings will almost always choose this, as well as vassals with non de jure lands.

2. Winner of a tyranny revolt gets the same independence prompt. Old liege gets strong claim.

3. If adventurer wins, same as above. Vassals get prompt, liege gets strong claim on anyone who leaves.

Regarding factioning, those 5 suggestions I mentioned are perfect information resistant. These are not so dependent on information as they are limits set on the AI. Limits we as players currently exploit.

You would need to invest a lot of time and attention to maintain a huge realm under these circumstances regardless of what you know about your vassals, which makes sense. It is a huge realm. The goal is at a certain size you are just fighting to keep everything, rather than blobbing out further. You can blob out further, but risk the empire imploding earlier.

 

[Karlington]

The game has difficulty levels. What I'd like to see is the hardest level being with the AI playing optimally (as far as it can be written to do so) without traits and things like AI rationality and zeal affecting it. Medium would be as it is today (can always be tweaked, of course), and easiest as the AI basically being an idiot.

 

[Rupmalya.K]

The game has difficulty levels. What I'd like to see if the hardest level being with the AI playing optimally (as far as it can be written to do so) without traits and things like AI rationality and zeal affecting it. Medium would be as it is today (can always be tweaked, of course), and easiest as the AI basically being an idiot.

Good point.
Making "strategic mode for AI" optional is a good way to do it as we are so divided about this.
Now it is about how hard is to improve the AI and what is the overall demand for it basically how much are they willing to work on it.Some improvements can be applied to both as sometimes they are too irrational even with the right traits.

 

[GundamMerc]

I doubt it works this way at all or that it's even possible. Over in the HOI forums it was stated that the exact same AI is used for all difficulty levels as it would be next to impossible to have different ones implemented.

Agreed, but mostly in that the amount of resources needed for that are beyond what Paradox currently has.

 

[Karlington]

I doubt it works this way at all or that it's even possible. Over in the HOI forums it was stated that the exact same AI is used for all difficulty levels as it would be next to impossible to have different ones implemented.

It doesn't work this way. I'd like to see it, though.

The trait checks are done in the game files so those can easily be changed (though the amount of time and effort it would take to for example introduce a game rule or similar about this would be pretty huge).

As for the rest of it that's probably a driver hack and I have no idea what it would take to accomplish that. Many other games have different AIs for different difficulty levels, so I know that it's at least technically possible.

It's probably like @GundamMerc said:

Agreed, but mostly in that the amount of resources needed for that are beyond what Paradox currently has.

 

[Malprau]

I don't play CK 2 for the challenge, I play it for the RP. I don't get this idea where a game HAS to be challenging to be fun. Having the AI be too good would in my opinion ruin much of the opportunity to play characters suboptimally just to have fun. Seriously, half the fun of CK 2 is just doing crazy things that aren't exactly the "good" thing to do, but are the fun thing to do.

Honestly as soon as you add AI that is always competent, then you inhibit the ability to play your characters according to their traits. It kills what makes CK 2 fun for me.

Although I know that u were replying to other person, many people that care so much about RP and not so much about strategy, naively start thinking in panic mode that most of people who care about strategy more than u personally, care so much that they would like to make game 100% strategy game, no, please leave ur self out of that thinking, its quite irrational if u think generally like that since u almost impulsively in ur replies in this thread act like a "crappy A.I. apologist".

In my OP im arguing not against RP elements and not in bias for some sort of "perfect game balance" but actually for enchanting RP elements and against "perfect game balance" by simply caring for what makes game credible(and with this RP+ strategy mechanics credible even more too) in its own right, the A.I., cause the A.I. controls more than 99% characters and lets face it, this game tries to be realistic to a solid degree, in sense that its characters that matter and not some institutions or systems, just like in real life, cause its them who make decisions and who interplay and make world go in the game, just like in rl too.

Putting that aside, if u have A.I. that is gruesomely incompetent (as it is now), the game loses realistic touch and game world doesnt feel like worth roleplaying in, if u care for actual genuine role-playing experience that is, if u have ur own definition of what RPing means, thats cool, but u gotta respect whats the base definition too, from which things start/are based on.

My point is that me caring for A.I. being a least a bit more than minimally competent, makes game more roleplayable in the long run, not only cause things u do seem and feel more rewarding and interesting as a result, but also cause theres a history sense of what u did before in the game during ur scenario, that seem like meaningful even after u have failed in ur plans/agendas some time later, that credibility of meaningfulness and history need to be enchanted/cared about more and Im confident in saying that improving A.I. is one big part of what would do the trick.

Also please, with all due respect to the DEVS, considering the amount DLCs they sell daily whether on sale or not, just off CK2, they should be able to spare some people or even get some more to work on the game more extensively at least on the A.I. aspect, if they care about the game that is of course.

 

[bloodyglasspuppet]

I wanted to write a response to the "AI being good doesn't matter at all"-posts in this thread, but Malprau said everything i wanted to say in his last post. So i leave it with an "Agree"...

 

[Pankila]

I wanted to write a response to the "AI being good doesn't matter at all"-posts in this thread, but Malprau said everything i wanted to say in his last post. So i leave it with an "Agree"...

Thanks for sharing this

 

[Karlington]

And your statistcal proof that shows this? People are basing it purely on perception of viewing some outcomes. And human nature tends to reinforce bad outcomes more than good ones. So we perceive bad results more often when hard numbers would say they are much more mixed.

I'm well aware of the issues regarding both cognitive biases and statistics that you mention. I studied logic, research methods, and statistics (with courses in both general statistics and research statistics) at university, but thanks anyway.

I must confess to not having designed a proper statistical study with a large number of samples here. I think that would be a pretty strong demand in an informal discussion about a computer game, heh.

I haven't really seen any solid statistical evidence supporting the other point either, by the way. So even with the same data some people will perceive bad results more often, and others will perceive good results more often. Neither side is inherently any more liable to biases.

That being said, we do have it documented that the AI uses five values to influence its behavior: rationality, zeal, greed, honor, and ambition. We know that AI with lower rationality does behave more irrationally.

(The Wiki's trait page lists each trait's effect on all these values, if any, for those who are interested.)

On top of this it is documented that certain traits influence the AI by making it more or less likely to take certain actions. Some actions are even possible only to those with certain traits, others only to those without certain traits. This is easily verifiable by going into almost any game file.

By the way, let's not forget that this is a highly subjective question to begin with. We all presumably agree that the AI should behave irrationally at least sometimes and rationally at least sometimes. But where do we draw the line between enough and not enough of one or the other? And are the values that govern this set correctly or not? And so on.

So even if someone really sat down and designed a proper research study, in the end we'd be debating both the operationalization and the actual outcomes of the study.

 

[Less2]

To those who think AI is fine, land your heirs and see what happens.

I land my own family all the time. I make new family members specifically to keep up with the land I conquer. It's generally alright. The biggest issue I have is their education of future generations and them getting sick (not using a good physician with the well-tested choices) and getting injured in battle.

 

[Karlington]

I land my own family all the time. I make new family members specifically to keep up with the land I conquer. It's generally alright. The biggest issue I have is their education of future generations and them getting sick (not using a good physician with the well-tested choices) and getting injured in battle.

So just their preparation of future generations and staying alive in the current generation, no issues other than that?

Jokes aside, I find that you will usually have a better heir if you keep him unlanded than if you land him. On the other hand landed heirs as rulers have more access to focuses and events, so that tends to be a positive.

Personally I try to make my heirs mercenary captains (the kind that is available to nomads and those with certain bloodlines, not the regular ones). Tends to produce the best heirs in my opinion, though there are a few unfortunate events for them.

Same risks there as you mentioned, though. I guess it's all about the benefit analysis.

PS. Has anyone had much experience with making your heirs the captains of "regular" mercenary companies? Any good effects on your children? I have never created one.

 

[Less2]

So just their preparation of future generations and staying alive in the current generation, no issues other than that?

Jokes aside, I find that you will usually have a better heir if you keep him unlanded than if you land him. On the other hand landed heirs as rulers have more access to focuses and events, so that tends to be a positive.

I mean, it's a roleplaying thing. It's managable, every other AI has to deal with it, and its arguably more realistic to not have your entire family living into their 60s with perfect health (which is what players generally do by managing focuses and treatment). So I'm not too troubled by it.

 

[WomboCombo]

I don't get this idea where a game HAS to be challenging to be fun.

A challenging game isn't always fun.
But an easy game is always boring.

 

[Poliut]

But it's fun to have a character that's flawed?

It can sometimes be problematic when you're kinda forced into either breaking character upon taking a character over or trying to figure out what the AI was trying to do with your character and maintain his/her personality. Flawed characters are one thing, but having to do a 180 upon player control is another.

 

[Karlington]

There were two positions stated. One was the AI was just random and had no logic. I think your statement here helps to prove that position is factually wrong. The other position was related to your comment:

It is definitely wrong. While the AI is random*, the possible outcomes and the chance of each of them will vary strongly with both traits, AI behavior values, and the situation the AI is in.

"This does NOT mean that he AI shouldn't make poor choices! It means that the ratio is too skewed as it is right now."

Setting aside that neither of us actually counted "poor choices" let alone come to an agreement on what exactly is a poor choice, you comment at least shows there is some rational behind it. Therefore, if there are in fact a higher ratio of poor to good choices then wouldn't that be based on a higher ratio of the underlying trait (rationality) that is causing it?

So wouldn't a good place to start is looking at how this value gets set and then checking that ratio?

That would definitely be a good place to start, yes. The issue for us, as players, is that the results of the AI behavioral values appear to be hardcoded into the engine. So while we can read "ai_rationality = 30" in the trait definition for Genius and "ai_rationality = -50" for Lunatic, and even change them, the results of those changes are not really clear to us. I believe only Paradox knows for sure what behaviors are influenced by these values, how much, and in which way.

*I'm not being dogmatic when I say the AI is random. Almost certainly a highly rational AI character with 700 men is never going to declare war on the Byzantine Empire or Abbasid Caliphate at the height of their power. Maybe a Possessed Imbecile Lunatic with Rabies, would though. But since as I mentioned above we don't really know the exact effects of the values, while we can probably rule the first one out I have no idea if the second would happen or not.

A challenging game isn't always fun.
But an easy game is always boring.

Is that how you think? It's an interesting point of view, where the pleasure of gaming mainly (I'm not saying exclusively) seems to be coming from the feeling of satisfaction people get when they achieve something challenging. It's probably the most common type of game, going back all the way to Spacewar!, Asteroids, and the other early games.

I'd say there appears to be a trend of growing appreciation for a different type of game, though. Witness the exploding market for "visual novels," or episodic "story" games like those from Telltale Games.

One huge benefit of a grand strategy game like CK2 is that both kinds of enjoyment easily can co-exist within the same game.