AI defensive targetting overland

[Seravy]

if someone has a city within 4 of it's fortress, do you REALLY want peace with that person?

By all means yes, it means they are most likely strong enough to destroy me and I want to avoid war with them at all costs. They had to conquer my second largest city somehow...
Especially true for the AI - due to the resource advantage, peace benefits them more, unless said human already controls double-triple the territory but then the AI lost no matter what.

Unless we are talking about turn 30 and it's an outpost, that's a different story. I don't think I ever had the change to plant an outpost that close to an AI capital though.

these attacks would need to be an exception to normal treaty rules

There is no such thing. Even if the AI orders the attack to happen, it'll be automatically canceled by the treaty - and I'm not going to remove the safety mechanism that prevents accidental attacks for this feature.

that can reach the human stack this turn

The AI has no way of knowing that, unless the distance is literally 1 tile.

[Nelphine]

I'd accept 1 tile. But if you're not willing to do the treaty part, then this is all moot. Fortress strikes will just continue to be the simplest way to kill an AI.

[Seravy]

The treaty safety mechanism absolutely has to stay - otherwise attacks from engineers, idle ships, or random units would happen again.
But even without it, there is no way to attack and not lose the treaty - if the attack is not cancelled, it automatically applies a "break treaty" effect and penalty.
I also know no way to actually have an exception anyway - at the time the AI orders units to do it, it would need to somehow store "this stack is the one that can ignore treaty rules this turn" somehow, somewhere, that can be picked up by the movement procedure. The only way I know to do that is changing the normal "move and attack" order to a new one - but then half the game's code has to be rewritten to recognize the new code as another "movement" type order. That's not doable.

[Arnuz]

Do you guys call "fortress strike" only the backstab? I prepare for a fortress strike also in the case of war. It's easier to conquer the other cities with the wizard banished because he uses up all his mana quickly, and then they are just like neutrals. Plus, that gives spells.

So, avoiding the diplomacy issues, in case of war Nelphine's ideas are still valuable, and I agree that all the troops surrounding the capital are meaningless right now. Having the possibility to evaluate an attack not only stack versus stack but also group of stacks versus stack would be priceless. I use multiple stacks as well in fortress strikes, because then I can do them earlier: I will normally send in a bunch of random crap first and take out the ranged troops. Maybe the AI could do the same? Go in with low level troops first?

This should probably be only for higher difficulties though... And no idea on how difficult it would be to implement, it sounds pretty hard to do.

[Seravy]

Well...my greatest concern with this is - and it hasn't been mentioned yet, wonder why - idk how others do it but I attack from the sea over half the time (during late game where this discussion is relevant) - capitals are usually close enough to a shore one way or another.
At which point the whole thing falls apart : nothing can attack the stack because it's on sea. (ships aren't that strong to matter, land units can't go there, transported units can't fight if in ships,and the entirely flying doomstack will attack anyway if strong enough and shouldn't if not, as there is only one of it.)

Obviously, this assumes my units can move over water, but almost every mid-late game summoned creature can, and likely my heroes will, too. And normal units do nothing against a late game fortress anyway. So ultimately, the only case when I attack from land is if there is no sea nearby, or if I play Nature and can't do it (no, water walking is not safe, if dispelled, the units drown. It's safe if no enemy is sorcery and the stack is too strong to attack though so I guess even Nature can invade from the sea fairly often...)

So...for the insane amount of work doing this would be, it'd still have a "miss" rate of over 75%. Not worth it...

(and this might not even be intentional tactic from the player - if the AI is on another continent, which they usually are, the player has to come from the sea anyway.)

[Nelphine]

Three things:

I didn't realize just how the code between movement and diplomacy was linked. I unfortunately expected something more modular - that's my fault, especially given the age of the game. So, I (vastly) underestimated how much work it would be to have an exception to the treaty. I still think its important, but I'm not convinced its worth the amount of work required.

Second: a fortress strike for me is simply when I hit the fortress without bothering with any of the intervening cities or troops. Peace or war doesn't matter.

Third: I forget I have been playing a lot on the higher landmass sizes. Obviously this would do nothing on tiny, but tiny I consider to have its own set of problems anyway. But I hear your point about sea, as it would be important in a fair number of small-huge games anyway. I don't think it would be 75%, but it woyld exist. Two counter arguments, neither of which is super compelling: it would force the human to come by sea, which could at least slow them down in a fair number of cases. Also, this would apply to intercontinental units such as ships and water walkers, so it would still have a chance of helping albeit not as much. As previously mentioned, combat spells are far more important than anything and all this is aiming to do is cause a little attrition before the human doomstack hits the fortress. 7 triremes might not do much on their own, but they have enough HP to allow the AI to at least use several combat spells.


TLDR: seravy you have raised good arguments why this doesn't need to be done. I'm completely OK if you don't do it. I think its sad, because there regularly are dozens of stacks (not just units, full stacks) lying around near the AI fortress, which really don't do anything except falsely add to the AI army strength graph.

[Seravy]

I think its sad, because there regularly are dozens of stacks (not just units, full stacks) lying around near the AI fortress, which really don't do anything except falsely add to the AI army strength graph.

In theory these stacks should board ships and get moved to an enemy continent. Does that not happen? (Yes, they might get produced at a high rate so some will linger around waiting for a boat but ultimately they should get their turn to leave the continent in the end.)
Oh, I think I know! Yes, this is entirely your fault :P
If you block all towers then the Myrran AI will have no place to use their troops at since the only enemy (you) is not on their plane. Let them through and suddenly they'll be useful.
Arcanus AI shouldn't have this happen unless it's so early in the game they have no ships yet.

...or it might be the Huge land. Maybe the smaller number of ships can't keep up with the larger amount of units produced - The AI does go for fewer ships on larger land, but this number increases over time so...attacking too early can also be the reason. Or maybe another AI just killed all their transports :D

[Nelphine]

I find it's usually things like the 4 cities closest to the fortress are producing more troops/get more troops summoned to them, than the average in the empire; but other cities still build/summon troops. So, the ships can keep up with the other cities, but they can't keep up with them and the troops near the fortress. I see it on fair-huge games. One classic example for an Arcanus wizard is a chaos wizard who takes a while to get any summons past hell hound (usually this means they miss gargoyle, and probably also miss fire giant). When I get to them, say 1406-1407? There can be upward of 8 stacks of hell hounds (65+ hell hounds) near his fortress. And they just sort of sit there - the AI has a fair presence on other continents and islands, it just looks like the hell hounds are low priority to transport, and end up never getting transported.

However, the tower blocking thing also is a real thing; any way to make towers a valid target (treat them like a city or something)? I know you specifically stopped this for good reasons (I think it had to do with blocking their own units from going through so they ended up not travelling to the other plane at all); is there any way to at least make that tile a target, but more like a 'go here for transport' target? I doubt it, cuz you'll need to have multiple stacks going to the same target, but, it would be a better place for huge piles of troops to mill around and do nothing.

Actually another reason I might have this problem - what happens if the AI isn't at war with anyone (5 way wizard pact/peace treaty/alliance). What gets set as the enemy continent?

[Seravy]

Actually another reason I might have this problem - what happens if the AI isn't at war with anyone (5 way wizard pact/peace treaty/alliance). What gets set as the enemy continent?

Let me see... these are the possible continent types :

1 - AI fully owns - AI has a city on the continent and enemy presence is below 10%
2 - AI partially owns, at least 10% enemy troops present
3 - Not owned - no AI cities in continent.
4 - Allied - Only allied cities on continent
5 - No target - AI had a full stack on continent but failed to attack anything with it - everything was stronger
6 - No target - There literally wasn't any potential targets on the entire continent.

Out of these, anything marked 2, 3 or 4 can be randomly selected. If the AI fully colonized the entire plane, it might select nothing at all.

[Arnuz]

I add to Nelphine's observations that the AI won't hit the close nodes and lairs despite having these stacks wandering aimlessly, because of strategic value, and the feeling that it's not effective is quite strong: you arrive with your doomstack and you see all of this stuff and nodes and lairs still around.. For undead creation that's priceless, but even just for the loot it's good.

Maybe a solution would be to consider garrisons as stacks and re-evaluate the target of them as well each turn? Then you could aim for towers too as the garrison would move, to then be reformed with wandering troops?

Now that I think of it, considering groups of stacks vs stacks, despite the sea limitation, would allow the AI to also go for stronger nodes than the strategic combat allows, simulating the users' trick of whittling the node down. Strategic seems to prevent that somewhat if I get it right as it minimises losses, this might need to be changed.