AI defensive targetting overland

[Nelphine]

As per my latest post in the ghouls thread, I believe that the AI's current biggest problem is the vulnerability to fortress strikes.

If an AI sends a doomstack towards the human lands, the human has two options.  One: They can build a counter doomstack strong enough to attack the AI doomstack before it reaches its target.  Two: The human can dump a doomstack in to garrison whatever the AI is targetting.  This second option is a little problematical because the AI doomstack may switch targets at that point, which may result in the human requiring the equivalent of a doomstack in all of their garrisons.

However, on the reverse side, if the human sends a doomstack towards the AI, especially to do what we refer to as a fortress strike, then generally the AI has one huge problem with this.  The human will usually do so with a stack powerful enough to kill the AI fortress by itself - however, this generally has the side effect of meaning the human doomstack has enough overland army strength that nothing short of the AI capital garrison will even consider attacking the human doomstack.  This means that even though there may literally be dozens of AI stacks of units nearby, none of them will attack the human doomstack until it is weakened.  This means the human can literally wait and pick the time to attack the fortress so that the human knows they will succeed.


I'd like to add a new targetting/diplomacy feature to the AI:

If there is a human stack within 6 squares of the AI capital, it should be targetted by all stacks of 7-9 units the AI has within reach REGARDLESS of the comparative overland strength; and the AI should actively try to build stacks of 9 near its capital (which it already does to a large extent). Basically, this is to give the AI the chance to use combat spells to at least weaken the human doomstack.  Good chance it will do nothing, but right now, leaving the human doomstack to wander around unopposed just leads the AI being banished. City garrisons (especially the fortress garrison) would be exempt from this rule to attack the human stacks.

The human should get a warning if it leaves a stack within 8 squares of the AI capital.  If at least 5 turns go by with no human units within 8 squares of the AI capital, the warning will be reset.

The warning should basically be 'I don't care how friendly we are, even if I'm peaceful and we're allies. Get away from my fortress or I'll destroy your offending troops.'

The warning itself will have no diplomatic penalty. If the human leaves, hurray, diplomacy continues without notice. (Given the 8 square distance, I think this is required, as this will cover a big chunk of the map.) Since presumably it will be the AI attacking the human and losing in most cases, I don't THINK there would be any diplomatic penalties there either? So the human is free to stay near the enemy fortress and be attacked, without actually changing the diplomacy of the game.

This will have a secondary side effect, that the AI will claim all cities and nodes within 6 squares of its capital as it's own. If there are enemy cities in that range, that could lead to potential wars, but I don't think this is a problem. If something is that close to a fortress, the AI SHOULD consider it theirs, in order to protect themselves.

Edit: I'm aware this leads to potential abuse with the human sending lots of small stacks near the enemy capital to use up mana. But since the human can already do that, I don't think it's an issue. And if the AI attacks and wins a battle, I believe that has major diplomacy repercussions, so the human would pay for it anyway.

[Nelphine]

I think there are two other things we may want to do. As far as I can tell, the AI cannot disembark from a ship, and then attack something, on the same turn. Given this, either we need to teach the AI to be able to do this, or more likely, we should make disembarking use up all of the remaining movement of the units that are disembarking.

Similarly, using astral gates should use all of a units movement. Potentially planar travel as well, but that leads to some issues if the player happens to click the 'plane' button while selecting a planar travel unit. Maybe a pop up whenever a planar travel (or astral gate) unit is selected and the human clicks 'plane' asking if they really want the unit to switch planes, which would include a warning that all remaining movement will be used up?

[Seravy]

So...

are you saying the change to the banishment rules was inadequate to fix the problem?


I was writing a longer essay for a reply but then I realized there is no need.
This idea is impossible. The AI's own cities - the first 3-4 they built at least, are in the area you mention. So if the human player conquers any of those, they'll be permanently unable to make peace with that AI. Now, that sounds bad by itself, but consider that on a "fair" landmass game, this can easily be half that AI's empire. On Tiny, it probably is all of it.

The other problems are
-Not having a free diplomacy slot for a new message type, especially warning
-The AI can't attack without first breaking the treaty with the human player
-The AI most likely endlessly wasting not just mana but troops - A doomstack that can wipe out everything in the first turn will act like a black hole and suck in all the AI troops from the entire continent, killing them for free without risk.
-The human can put their doomstack 7 tiles from the capital and when they want to attack, reach it in 2 turns before the warnings turn into attacks anyway.

Also, I'm not going to break ships and planar travel for balance reasons, don't even think about that. I'm glad they at least function mostly correctly.

[Nelphine]

Ships I think is worth it - it's a difference between AI and players that doesn't have any apparent reason for the difference (by apparent, i mean in the eyes of a new player, which means the difference won't be noticed.) Currently, the AI is simply weaker. They get off the ship, and then do nothing. Which means the player ALWAYS has a turn to react to a naval invasion, whereas the AI does not get that - the human can attack directly from sea to land (which this fix won't change), but they can also move from ship to land, to another tile on land to attack something further inland. If the AI can't handle it due to their code, why not simply put the same limit on humans?

Planar travel - I'm not worried about it. But, my argument basically is the same as with ships - why should humans be able to do something the AI can't? (of course, the AI can target enemy units in a way the human can't, so I guess there are at least trade offs. I'd just rather see it the same across the board.)

Banishing.

You and I have different ideas of the problems of banishing since the start. My problem has ALWAYS been the overland spell use. Your fix does nothing for that. Which means, aside from banishment lengths, which we have good discussions going on, and as far as I know, you are still considering have a special exception for spell of return cast time, all your fixes have done nothing to change my playstyle.

I waltz in, fortress strike, and the AI is dead. It might take a while, but without overland spells, they no longer threaten me.

My fix above is aimed at the fact that the AI has coded limits on what it considers a valid target for it's stacks. I literally regularly will just dump a doomstack down next to every AI fortress and put them on patrol, then forget about them until that AI declares war on me. They declare war on me, maybe they attack some of my cities, maybe i lose some, maybe I lose some nodes. Then my doomstack fortress strikes them. Then it's over.

Or in closer games, my doomstack moves toward the enemy fortress; maybe we're at peace, maybe we're at war, it doesn't matter. I can take however long i need to get to the enemy fortress, and in some games, i've gone past literally hundreds of units, until I can attack the fortress itself. Then I kill the fortress. Game over. All of those units that the AI builds literally mean nothing, because they have to be allowed to attack in the first place, and in 95%+ of all cases, nothing the AI has will attack ultra-elite bezerkers with a few buffs. It doesn't matter how the actual combat will go, the army strength tells the AI they can't attack in the first place.

So as long as my capital has one such stack, and I have enough other stacks to actually win the combats to kill the enemy fortress, it doesn't matter if I lose every other city on the map; they won't attack my important stacks.


However, we all know that actual combat isn't limited by overland army strength. If the AI could attack those doomstacks with much weaker stacks, I could easily lose them, or at least be weakened enough, that I couldn't take on the fortress.


So you're right that in many cases the AI would be throwing away stack after stack for virtually no return. But right now, all of those stacks of units literally do nothing anyway - so this would be at least a little bit better.

[Seravy]

If the AI can't handle it due to their code, why not simply put the same limit on humans?

Because it's not just ships. The AI can't do two consecutive moves with the same units in any way.
For example, they can't merge two units into a stack and then move the stack. They can't attack a node directly from inside a city. They can't leave ships if the target tile has an enemy stack even if they would want to attack it. They can't leave a city, move units through and then put the garrison back. They can't even swap the garrison units. And so on. Changing movement rules to match these restrictions would make the game completely unfun and picking only one of them feels...not very useful at all.
Also, automatic pathfinding will move units through ships if that's shorter - if the unit would lose 2 turns doing that it would be really annoying.

But, my argument basically is the same as with ships - why should humans be able to do something the AI can't?

Because the humans can't see the other side and know in advance if there is water or not, and I still haven't been able to fix planar travel allowing to move into water nor will I ever be able to. So changing planes means a 50% chance to drown the unit in your suggestion.

[MrBiscuits]

An 8 square radius on a tiny map can be half the usable land. You can start with two fortresses that close together.

[Nelphine]

OK, I am convinced about ships and planar travel - automatic pathing, and chance of drowning are both way to convincing. Those ideas are dropped.

MrBiscuits: I'm OK with the tiny problem. That just puts you into immediate war. So be it. Tiny has a lot of strangeness anyway. (Although 8 squares was only for the warning, and there's no room for a new warning anyway, so only the 6 square radius matters. I don't think I've ever seen fortresses within 6 even on tiny.)

[Seravy]

OK, I am convinced about ships and planar travel - automatic pathing, and chance of drowning are both way to convincing. Those ideas are dropped.

MrBiscuits: I'm OK with the tiny problem. That just puts you into immediate war. So be it. Tiny has a lot of strangeness anyway. (Although 8 squares was only for the warning, and there's no room for a new warning anyway, so only the 6 square radius matters. I don't think I've ever seen fortresses within 6 even on tiny.)

Fortress minimal radius is 10 but if that fails to generate, it drops by 1 until success. In theory it can go down to 6 but unlikely.

I find the inability to make peace for the rest of the game if I conquer any of an AI's early cities (that get built in a range of 4 typically) unacceptable though. Likewise the thought of the AI throwing away its entire army on the whole continent to a "trap" of 9 ranged units and some decent combat spells to ensure I finish all the battles the first turn (or anything equivalent - a solo Sky Drake and Call Lightning could do the same). That costs the AI more than getting banished - without idle troops on the home continent, there will be nothing to ferry across and attack the player with. It's equivalent to losing the game unless the AI owns at least another large continent.

[Nelphine]

Ah, you may be misintepreting one thing. Only stacks that are actively in range of the human doomstack would attack. So they won't be sending things from the whole continent. Only things that are in the immediate viscinity. Generally, the AI has large clouds of troops around their fortress; that's all I'm trying to use. And yes I've already acknowledged that the AI is liable to lose a lot of troops to this. But those clouds of troops currently literally don't do anything to prevent the actual fortress strike, and because they're right next to the AI fortress, they aren't in a position to threaten your empire after the fortress strike - so those troops end up being meaningless. And when that's one of the highest concentrations of AI troops, they should not be meaningless.

And think about it from the AI point of view - if someone has a city within 4 of it's fortress, do you REALLY want peace with that person? They literally have a foothold to just attack your own personal fortress and banish you. I know if an AI settles next to my capital, well, I'm going to take that city sooner rather than later. (Also remember, you could still have peace with them; these attacks would need to be an exception to normal treaty rules, where they can attack regardless of treaty status. So the rest of your empire you could have peace with them, unless they attack and conquer that city.)

[Nelphine]

The basic idea is that at the very start of AI targetting, before anything else is done, if the human has any units within 6 squares of the AI fortress, then any AI stacks of 7-9 that are not currently in a city, that can reach the human stack this turn, will attack the human. After that, the AI would do all normal targetting.