Dollar auction and total war: AI overcommits to minor wars

[Careful Plum]

I`m sure I understand his argument pretty well, it boils down to two points, I have addressed both in fact. For AI to commit "necessary" amount of forces, would be to commit enough forces to defeat all enemy army, reliably in worst case possible, and any kind of mercenary or alliance shenanigans a player can use. That one was pretty clear, and nobody seemed to mind it. I really wouldn`t mind such calculation done by AI, but it totally should include tricks and exploits in mind and should ensure there is zero chance of war going wrong even in worst case scenario.

His second point was about AI over-committing to defence of a "worthless" province, that player wanted to conquer, but it was impossible, because AI would actually commit way more than player thinks AI "should", and thus player "has" to also commit huge resources to that war, making the war "unreasonably" expensive, and thus, impossible for player to wage with any hope of gain. How for his small realm such province would be of different "value" than for huge HRE realm, and how it shouldn`t be as committed to defending it with it`s full force, and overall, HRE should just commit a tiny amount of it`s forces, and after they are defeated by player`s way larger overcommitment, HRE should just give up and give player the province.

It`s your interpretation of his complains that I find strange and missing the point, quite frankly. You seem to be the only one who found any point about how AI should just stand down and give up if it is weaker.

It's true that my interpretation is a little different from the OP's, but the underlining problem is the same: the AI treats every war the exact same way. It will always raise all of its levies and always fight until exactly 100% warscore have been reached (well, I guess, in some cases there will be a white peace if the war ahs been going on for a long time).

The OP complained that the AI is expending too much of its strength to keeping a worthless county. You said that keeping a county is always worth it. This may be true, but it's beside the point: the AI doesn't expend its strength because keeping a county is worth it, but because it will always expend all of its strength to winning a war. That this might be the right move in a war about a county is coincidental. Try giving the AI a CB where neither the winner nor the loser will get anything, no matter the outcome, and the AI will still fight to the death.

 

[mursolini]

It's true that my interpretation is a little different from the OP's, but the underlining problem is the same: the AI treats every war the exact same way. It will always raise all of its levies and always fight until exactly 100% warscore have been reached (well, I guess, in some cases there will be a white peace if the war ahs been going on for a long

The OP complained that the AI is expending too much of its strength to keeping a worthless county. You said that keeping a county is always worth it. This may be true, but it's beside the point: the AI doesn't expend its strength because keeping a county is worth it, but because it will always expend all of its strength to winning a war. That this might be the right move in a war about a county is coincidental. Try giving the AI a CB where neither the winner nor the loser will get anything, no matter the outcome, and the AI will still fight to the death.

It is true that AI is treating wars same. But, the OP doesn't make any point about how AI makes things that make it weaker, OP claims that since the war to defend a county would cost AI more money than it will gain from it, AI should be scaling it's war capacity commited towards the size of the income from said county, and give it up if there is a cost overrun.

 

[raven63]

In game theory there's this thing called Dollar auction where two players, competing over a single dollar, pit insane amounts of resources against each other in a futile effort to cut their losses. CK2 AI doesn't get it.

I started a small test game as King Krasimir IV of Croatia. Before unpausing, I granted myself a claim on the Holy Roman Empire using the console, and the lousy single-holding tribal county of Kola to the Kaiser of HRE.

Spoiler: Pic

View attachment 175180

I declared war to see how HRE would respond. As expected, they take claimants to their throne seriously - by the end of the year they have well over ten thousand troops marching against fair Croatia (I had only about four thousand or so troops available, so that's a bit of an overreaction, but I digress). Fair enough, eh?

Spoiler: Pic

View attachment 175181

But what happens if I repeat the experiment, but claim the near worthless county of Kola instead?

Spoiler: Pic

View attachment 175182

The AI clearly goes in for a total war again, despite the two titles fought over being of clearly unequal value. In the first war it's reasonable to commit a lot of troops since a lot is hanging on it - in the latter war the Kaiser is lucky to ever get the money spent on maintaining that levy back from his little Kolan exclave, one that he can't even defend against raiders.

Granted, this is an extreme example, but resembles a very common situation in-game - trying to usurp a single barony from a large enemy. Because the AI commits irrationally large armies to wars over small parcels of land, the player must commit such forces too, making wars against large targets unappealing because of wrong reasons. It's a dollar auction - one where the AI makes a stupid choice to spend lots of resources to defend something of limited value, forcing the player to play even more stupidly (=use an even more excessive army) to gain the prize - or accept the status quo.

Totally agree. +1.

 

[MichaelStakely]

That total war behavior combined with automatic call-to-arms was wonderful if you had an alliance with a large AI power.

In a recent Irish game, I had a daughter that I wanted to marry off, went to the marriage screen, and one of the many sons of the Basileus popped up. I pressed a claim on Tyrconnell and the Basileus dutifully spent months on end shipping the entire Byzantine Army over in batches of a few thousand at a time. The second I realized that was going on, I marched my troops home to avoid losing troops, dissolved my levies, and declared Wars to Make Tributaries on every other province in Ireland. One marriage wound up getting me a county and four decades worth of tribute from eight different tribes.

 

[Strangedane]

Could you imagine if say the English only committed part of their resources to winning any war in history, based purely on the size of their opponent. If you have superior forces, you use them.

Falklands.
All their Indian Colonies.
Even the god damn independence war they didn't commit all their resources to.

I'd love for you to find even a single example where all of Englands reserves have been raised AND mobilized.

 

[Careful Plum]

It is true that AI is treating wars same. But, the OP doesn't make any point about how AI makes things that make it weaker, OP claims that since the war to defend a county would cost AI more money than it will gain from it, AI should be scaling it's war capacity commited towards the size of the income from said county, and give it up if there is a cost overrun.

That is absolutely right. Do you really disagree with that? That is to say, if it WERE true that there was a cost overrun (in gold, soldiers, prestige etc.), shouldn't the AI do exactly that? Should it spend more on defeinding the county than it will gain from it, including of course ooprtunity costs? It seems to me the only difference between the two of you whether that overrun exists in this specific case. Maybe sying "you misunderstand" was a wrong formulation.

And at that point, is it really so much more to say that, given the AI will win the war, it should also not spend more than necessary? And given that it will lose, it should also not spend more than necessary before losing? These all seem natural conclusions from the above, but again, maybe my original formulation was a bit flawed.

 

[Strangedane]

What makes me really sad is seeing the AI throw everything into a tributary war when the agressor is more than 70 years old.

 

[psychoak]

This calls for crime theory.

A somewhat less than well liked, but unequivocal truth, is that crime prevention in most modern countries is ass backwards. Law enforcement has no will to track down and prosecute low level offenders, petty crime is ignored because it's petty. The manpower deemed necessary is far in excess of the benefits, far better to put more time into the rapists and murderers. Then we get the broken window theory.

Replace a window in an abandoned building for no reason, or leave it broken and come back to find every window broken. Municipalities that crack down on petty crime see far lower crime rates across the board, as opposed to those that don't. An excellent comparison would be New York City, versus Chicago. One is a very safe big city, the other compares more closely with war torn third world countries.

If the AI were to respond proportionately, you could take large empires one county at a time, the only time you'd have a real war is when two large empires go at it with an invasion CB or something. This would be an embarrassingly easy to exploit AI that literally couldn't defend itself from attacks. Even if it took a couple hundred years to pay for the invasions, you'd be getting that income, and the AI would be losing it, in a game that does indeed last centuries.

 

[PK_AZ]

Some way around would be to change the way vassal levies works (in defensive wars)
1) you dont raise your vassal levies, you call your vassals to war (similar to tribal mechanic)
2) vassals called that way are "minor participants", they will mobilise small percent of their levies (I think 10% is OK) and give them to you. Generally speaking vassals will not care about that war - they already give you troops, so their obligations are fulfilled.
3) you can call only vassals that are near the contested counties. I would suggest creating geographical area bigger than de iure duchy, something like de iure kingdom, but without drift (lets call it "territory"). You can call vassal only if his capital and any of contested counties is in the same or neighbourning territories.
4) vassals can offer to join war as normal participant. AI will do that only if their land is contested, their ally/friend land is contested, they are your ally/friend, they are aggressor rivals or in exchange for a favor.

This way HRE cannot mobilise all it's forces for some distant county. On the other hand, it still can mobilise when HRE title is at stake (because formally in war for main title every county in realm is contested).
Of course the way you mobilise forces for aggressive wars also needs to be changed. Personally I'd like to see something between Stellaris war declarations (when you are in alliance_ and CKII-style fractions. But I digress.

 

[StarSword]

It's true that my interpretation is a little different from the OP's, but the underlining problem is the same: the AI treats every war the exact same way. It will always raise all of its levies and always fight until exactly 100% warscore have been reached (well, I guess, in some cases there will be a white peace if the war ahs been going on for a long time).

The OP complained that the AI is expending too much of its strength to keeping a worthless county. You said that keeping a county is always worth it. This may be true, but it's beside the point: the AI doesn't expend its strength because keeping a county is worth it, but because it will always expend all of its strength to winning a war. That this might be the right move in a war about a county is coincidental. Try giving the AI a CB where neither the winner nor the loser will get anything, no matter the outcome, and the AI will still fight to the death.

This was kind of exacerbated by the automatic call-to-arms. I had the Byzantine Empire raise its entire army to help me punch out a single rebellious duke on the other side of the continent. I need them for the Umayyads, not for minor spats with vassals. Thank God they're dumping that in the next patch.

 

[mursolini]

That is absolutely right. Do you really disagree with that? That is to say, if it WERE true that there was a cost overrun (in gold, soldiers, prestige etc.), shouldn't the AI do exactly that? Should it spend more on defeinding the county than it will gain from it, including of course ooprtunity costs? It seems to me the only difference between the two of you whether that overrun exists in this specific case. Maybe sying "you misunderstand" was a wrong formulation.

And at that point, is it really so much more to say that, given the AI will win the war, it should also not spend more than necessary? And given that it will lose, it should also not spend more than necessary before losing? These all seem natural conclusions from the above, but again, maybe my original formulation was a bit flawed.

Absolutely, I do disagree, that is my entire point.
Why? I'm sure I tried my best to explain. There is value in keeping up your income and soldier base untouched, even if it may seem like a bad idea short term. There is value in draining your enemy's resouces, if it would take you less time to recover. There is value in draining enemy resources since it sets him up for a backstab.

It is often far better to start fighting at full strength right away, rather than slowly give up half of your kingdom without much fight.

I'm in favor of AI considering opportunity cost, but it is very complicated, AI really need to understand how the situation would develop if both enemies get drained. Who has better reserves, whose realm can handle larger stress, who is more lukely to get backstabbed.

No, I'm not in favor of formulas that establish how much AI should lose before giving up on war. And, I'm not in favor of AI giving up on wars in which it is objectivly stronger just because AI lost a few battles. Eve just making AI recognise wars where it only wants to defend and would peace out on white peace, and woyldn't waste troops on sieges would be a good start.

 

[Aries666]

In game theory there's this thing called Dollar auction where two players, competing over a single dollar, pit insane amounts of resources against each other in a futile effort to cut their losses. CK2 AI doesn't get it.

I started a small test game as King Krasimir IV of Croatia. Before unpausing, I granted myself a claim on the Holy Roman Empire using the console, and the lousy single-holding tribal county of Kola to the Kaiser of HRE.

Spoiler: Pic

View attachment 175180

I declared war to see how HRE would respond. As expected, they take claimants to their throne seriously - by the end of the year they have well over ten thousand troops marching against fair Croatia (I had only about four thousand or so troops available, so that's a bit of an overreaction, but I digress). Fair enough, eh?

Spoiler: Pic

View attachment 175181

But what happens if I repeat the experiment, but claim the near worthless county of Kola instead?

Spoiler: Pic

View attachment 175182

The AI clearly goes in for a total war again, despite the two titles fought over being of clearly unequal value. In the first war it's reasonable to commit a lot of troops since a lot is hanging on it - in the latter war the Kaiser is lucky to ever get the money spent on maintaining that levy back from his little Kolan exclave, one that he can't even defend against raiders.

Granted, this is an extreme example, but resembles a very common situation in-game - trying to usurp a single barony from a large enemy. Because the AI commits irrationally large armies to wars over small parcels of land, the player must commit such forces too, making wars against large targets unappealing because of wrong reasons. It's a dollar auction - one where the AI makes a stupid choice to spend lots of resources to defend something of limited value, forcing the player to play even more stupidly (=use an even more excessive army) to gain the prize - or accept the status quo.

What would you prefer in its place, people chipping individuals counties off the HRE, it's only one county what's the point in defending it? Presumably this continues until there are only two counties left in the HRE before the AI considers it worth defending seeing as it's now 50% of its remaining territory. Another extreme example but you get my point, that is that it is probably easier and preferable to have an AI that will always defend with full force than have one that has to make an arbitrary decision about what is/isn't worth defending and then considering a level of commitment for that defense.

 

[Aries666]

That's both faulty logic and not relevant. First of all, the territory could be of more value to me than it was for the AI realm - these things aren't absolutes. Also, you're assuming I'm like the AI - that I would gladly pay hundreds of gold in levy maintenance and risk my realm's security for a single worthless holding. I'm not.

As for your question "why shouldn't the AI produce more than enough to crush you", well, because "enough" is enough. Every soldier I deploy to fight costs me money and is away from potential other conflicts. Having a substantial advantage is good because it reduces casualties, but "overkilling" is inefficient.



No I don't - see my pictures, I don't have a single soldier deployed. I would like to capture small foreign holdings in my realm, but the would AI make a stupid choice that forces me to make stupid choice too - so status quo, letting them keep that holding, is the only option that doesn't drive the player insane.

Is it good gameplay, though? Or realistic? Neither. It's both annoying for the player, and unrealistic - real-life warfare is carried out using reasonable investments, because overkills cost money, men and stability.




Circular, faulty logic. As a player I can choose how much troops I spend. The AI always sends it all, see my pictures. I didn't deploy a single troop. You can't justify AI behavior as a fair reaction to a player overcommitting, because I didn't deploy a single soldier.



The Croatia thing is just an example. My argument boils down to this: the AI defends a worthless piece of junk county with equal resources it uses to defends its whole Empire title. That means it's overcommitting its resources on the first part.

You can't really create a scenario, concede that it is stupid and you wouldn't actually do because doing so would cause you to over-commit and you still wouldn't realistically win, then criticise the AI (for responding to you scenario that shouldn't happen) by similarly over-committing.

 

[Aries666]

mursolini, I think you misunderstand the argument here. It is a fact that the AI will ALWAYS commit the entirety of its ressources to a war. Doesn't matter if it's overkill or "underkill". It will happily march its army against a totally superior army to be slaughtered as well (it will try to avoid the enemy army, granted, but there's no reason to even bother raising those troops, and as I'm sure you know another player at least will eventually be able to catch that useless army anyway). There's also a point in the game where your army will be so much superior to the enemy, that more soldiers will NOT result in less of your own troops being killed.

It's not about whether the AI will defend its worthless provinces or not. The point is that it will commit more than is necessary to defend it, i.e., it will, after the war, have less than it would need to have to achieve the same result.

And, again, similarily, if the AI is attacked by an enemy it has no hope of winning against, it will still happily try to defend it, with the only difference being that it will loose the county AS WELL as its army.

This also runs into the same issue as the AI over-committing. You say the AI shouldn't defend when it is outmatched, what are the thresholds for that, when exactly should the AI just roll over and die for you. What you have essentially said is that once the player has the largest army the AI should simply not bother contesting.

 

[n00bypl4y3r]

Doesn't matter. You declared war on them, obviously they will respond in kind.

Could you imagine if say the English only committed part of their resources to winning any war in history, based purely on the size of their opponent. If you have superior forces, you use them.

Additionally your comparison to the dollar auction game theory isn't accurate, since there isn't a third party pocketing the proceeds. You are aggressively trying to take something from the other contestant.

No, the English structured armies exactly for this. An Anglo-Saxon kingdom had access to numerous layers of fighters, so if they only needed a few hundred, the King's Gesith/housecarls could deal with it. Larger, the thegns, their retainers and earls could deal with the problem. If they needed a large force to deal with a massive invasion like William and Harald, the fyrd, militias of trained freemen, mercenaries and peasants when necessary, could be called up. Even after England was invaded by two powers not all the manpower was raised, and just the power that was, was enough to win at Stamford Bridge and almost win at Hastings for awhile. After, there were still enough troops to put up considerable resistance at London, as well as other areas.

And of course, individual fyrd militias could be used whenever as well, they didn't need to be united into a common force.

 

[Helios Panoptes]

Paradox games always have this problem. Just try playing the American Revolution in EU3, King George III will personally lead the entire British Armed Forces to kick your head in.

It's part of why small states can't survive like they did historically, but I'm not sure how to fix it..

 

[alanschu]

I'm in favor of AI considering opportunity cost, but it is very complicated, AI really need to understand how the situation would develop if both enemies get drained. Who has better reserves, whose realm can handle larger stress, who is more lukely to get backstabbed.

This really can't be understated. I'd love to have AI that could properly play like another human player would but it's suuuper complicated (especially quickly).

 

[Thrake]

Didn't you read any of my post? They are making a huge mistake paying for that many troops to defend a practically worthless piece of clay way up North. If you disagree with that, you're pretty much objectively wrong. Not every small territorial clash was fought on a Hundred Years' War scale.

I disagree. Vassals pay maintenance, and gold in CK2 is of limited use. Lands are worth much more than gold.

Furthermore, unraised troops are wasted troops. More raised troops means shorter wars and faster going back to business. Less losses too which means being less vulnerable to vultures. You are similarly reacting disproportionately declaring a massive war over an useless piece of land. The AI barely defends itself; YOU are overracting.

I only raise part of my troops in offensive wars when it's easy wars like annexing a count I can do it just with my demesne to avoid raised levies penalty (the AI could be improved in that reguard) but really defensive war according to game rules ought to be total war or it is suboptimal play.

 

[es333]

If someone considerably bigger than me declares war on me for a single county, I just look at his size, look at my levies, look at his size again and then instant surrender. I don't fight wars that are almost certainly lost although sometimes I just say "fuck it" and go for it.

I once started as some Norse duchy in Ukraine and needed a port, so I declared a war on Byzantium for the county they have in Crimea, I got it and raised massive amounts of tribal armies to keep it. Byzantium kept marching its death stacks onto me and also invading from the sea, but Odin must have been watching over Crimea as I won battles against 4 times bigger stacks and finally invaded Byzantium to get the warscore to 100% for a single county.

The point is. Byzantium completely depleted its army to fight for a measly county in Crimea and the consequences of doing that meant civil war and invasions from its neighbors. Sometimes giving away the county is the best option.

 

[alanschu]

It'd be valid for the AI to be less stubborn with its war score.

(I think it can sometimes instantly surrender with the Tributary CB?).

This isn't really related to how much force the AI commits if it decides to actually attack though.