feedback on AI combat behavior

[ilteroi]

I've done this to keep a coastal city that I can't properly defend because his navy has destroyed mine, slowly wearing down his navy and waiting for him to peace out (on a turn where I contol the city). It's incredibly easy to do if it can't get to your mounted units.

epic discussion going on here. it's just hard to resist the cheesy move ...

so i would propose: AI should not capture a city if it's obvious they will lose it again (not easy tell). capturing a city always ends the turn (have to impose some order etc). captured cities do not get any extra health (in fact i would like them to start at zero).

 

[Legen]

capturing a city always ends the turn (have to impose some order etc)

This is problematic in many ways.

First, it hinders anyone trying to move units into the city's surroundings to protect it from any resistance left. This is especially true on later eras amphibious assaults, where the captured city can act as a port and allow any embarked unit to retain some movement points upon disembarking.

Second, we may also have plans on what to do with the city immediately after capture, like moving its great works to other cities, using an inquisitor and moving air units to it (especially with Imperialism's finisher).

Third, I sometimes am sieging two or more cities at once and can capture two in the same turn. I see no reason why one division capturing Berlin should prevent the other from liberating that formerly allied city-state that Bismarck annexed.

Fourth, it could lead to another cheesy tactic in which one city could be kept from being conquered by having another flipping sides.

There are other things, like how it would work if you have other mandatory things to do, like voting on the World Congress, or if you planned to peace out immediately after capturing the city. Overall, I think this proposal is problematic.

captured cities do not get any extra health (in fact i would like them to start at zero).

That means any unit garrisoned in it gets blapped by any attacker due to the city having 5-11 CS. A riflemen attacking it would deal 300 damage going straight to the garrisoned unit. It's already tough as it is, but the city's recovered health allows for the garrison to live a bit.

 

[LifeOfBrian]

Both ending the turn and starting the health at zero seem too much to me.

 

[ilteroi]

Both ending the turn and starting the health at zero seem too much to me.

i didn"t say i want zero hp. but i'm thinking about a change to the garrison logic. right now the city is attacked and the garrison takes a share of the damage. which can be much larger than if the garrison is attacked directly. how about doing it the other way around, attacking the garrison first?

This is problematic in many ways.

First, it hinders anyone trying to move units into the city's surroundings to protect it from any resistance left. This is especially true on later eras amphibious assaults, where the captured city can act as a port and allow any embarked unit to retain some movement points upon disembarking.

Second, we may also have plans on what to do with the city immediately after capture, like moving its great works to other cities, using an inquisitor and moving air units to it (especially with Imperialism's finisher).

Third, I sometimes am sieging two or more cities at once and can capture two in the same turn. I see no reason why one division capturing Berlin should prevent the other from liberating that formerly allied city-state that Bismarck annexed.

Fourth, it could lead to another cheesy tactic in which one city could be kept from being conquered by having another flipping sides.

There are other things, like how it would work if you have other mandatory things to do, like voting on the World Congress, or if you planned to peace out immediately after capturing the city. Overall, I think this proposal is problematic.



That means any unit garrisoned in it gets blapped by any attacker due to the city having 5-11 CS. A riflemen attacking it would deal 300 damage going straight to the garrisoned unit. It's already tough as it is, but the city's recovered health allows for the garrison to live a bit.

i don't understand the disembarkation argument ... only the first unit to enter the city has to spend all movement points. all following friendly units can just move through.

i think there is a misunderstanding here regarding ending the capturing unit's turn vs ending the capturing player's turn.

let's just pretend that an enemy city is "rough terrain" which stops fast units. in return you get extra hitpoints for the city.

 

[Emperorman1]

What if melee attacks would be against the garrison, but ranged attacks against the city? Of course if the city is at zero health or there is no garrison, then the attack is directed against the other. Would this work?

 

[Moi Magnus]

how about doing it the other way around, attacking the garrison first?

Would bonuses against cities count against the garrison? To be honest both Yes and No answers come with problems: Yes -> units in garrison are easier to kill than outside the city, No -> bonus against cities are mostly useless.
Moreover, depending on how it works, that could lead to the reverse problem: sometimes, adding a very weak garrison could make dealing damages to the city easier rather than more difficult.

 

[LifeOfBrian]

Ilteroi, I like your idea of first having to deal with the garrison before the city itself!

 

[Stalker0]

I don’t think I understand what “attacking the garrison first” actually means.

 

[Omen of Peace]

I like the proposal that capturing a city ends the unit's turn!
I've also been guilty of letting the AI retake a contested city to destroy the unit they leave there. It doesn't happen often, but when it does it's guaranteed kill every turn.

If "attacking the garrison first" means you first do normal attacks against the garrison until it dies, and only then you start attacking the city HP... then congrats, I hate it.
Because garrisons would tend to evaporate.

 

[civplayer33]

I don’t think I understand what “attacking the garrison first” actually means.

I think he means the garrisoned unit is treated as if it were outside the city and when something attacks the city it attacks the unit without damaging the city until the unit is dead and only then is the city damaged...I don't think it's a good idea tbh; Moi Magnus points the most obvious problems out:

Would bonuses against cities count against the garrison? To be honest both Yes and No answers come with problems: Yes -> units in garrison are easier to kill than outside the city, No -> bonus against cities are mostly useless.
Moreover, depending on how it works, that could lead to the reverse problem: sometimes, adding a very weak garrison could make dealing damages to the city easier rather than more difficult.

The other thing about "ending the unit's turn" is just a rehash of the proposal about mounted units not being able to move out of the city after taking it, just in a more generalized form, and has all the problems I already mentioned earlier.

The issue that is being blown up to enormous proportions here is not that big that it requires dramatic changes. If anything should change it's the AI not taking cities repeatedly that it cannot hold. How about adding a subroutine to AI tactical logic that does this:
IF (city that AI took last turn was retaken and AI unit inside was destroyed and this happened the turn before as well already)
THEN
make sure that AI controls city surroundings (with ZOC fields should be fine, doesn't have to surround the city literally with units) before taking it again; until then kill the enemy units only (and also teach AI not to shoot more at the city than it or the garrison has health, which I've seen a lot a while back...don't know if this has been fixed).

So basically when the condition is met the AI should focus on the enemy around the city until it can control every tile around the city (either by a unit standing there or a unit's ZOC field restricting access to that tile).

 

[Stalker0]

I think he means the garrisoned unit is treated as if it were outside the city and when something attacks the city it attacks the unit without damaging the city until the unit is dead and only then is the city damaged...I don't think it's a good idea tbh; Moi Magnus points the most obvious problems out:

Assuming that the garrison no longer shares damage with the city, than this makes city taking much easier. The garrison is a big reason cities last as long as they do. Once the garrison goes down, my city damage almost doubles (shares no more damage, and the CS goes down).

Now, the idea that melee units use this mechanic and ranged units don't is very interesting, it might give a very nice niche to melee units in city taking. There city hit bonuses may need to be nerfed to compensate but I would be fine with that. But it would give a strong benefit to a more combined arms strategy, melee to soak up the garrison and then siege to really take the city.

 

[065tdsa]

This idea for a new garrison system sounds like it'll only kind of solve one problem and suddenly create 20 new ones. If you want the cities to be harder to capture, give them more HP rather than take it away and make up a whole new system that will probably require a re-balance of city bonuses among other things?

Automatically ending the unit's turn is just what people had proposed previously, and I have to say I'm still on the side of giving the AI slightly more logic. I'm not an expert on the subject but it is already pretty decent, maybe to a fault, at avoiding dangerous situations. Surely it can be made to not dive into somewhere it's bound to die just because that somewhere is a city? The AI capturing a city it'll lose again next turn can be a bad move even if one of my units is inside it, so long as I have more than it does or simply if the unit in there is less valuable to me than killing its unit.

 

[crdvis16]

I think trying to teach the AI to not take cities if it can't hold them will probably have unintended consequences. There are times the AI probably should take the city anyway even if it might not hold it.

I still think just forcing the loss of all movement points after capping the city is the easiest solution so that at the very least the human is hurt just as much by a city repeatedly changing hands as the AI. Despite all the hyperbole from civplayer33 I don't think it's a change that has any real consequence and it seems kind of logical to be honest- if crossing a river drains all of a mounted unit's movement then surely taking a city (killing the population, breaking their fortifications, etc) would also use up your movement?

However, changing the way a garrison's health/CS works when attacking a city might be a more complete solution and I share Stalker's interest in giving melee a role to play in city siege. I'd propose this:

1) Ranged/Siege units damage the city first based on the city's CS and benefiting from the normal city attack promotions. If the city has been reduced to 0 HP then they damage the garrison based on the garrison's CS and using unit attack promotions.

2) Melee units damage the garrison first based on the garrison's CS and using unit attack promotions. If the garrison is killed or there is no garrison then they damage the city using city attack promotions.

3) Garrisons enjoy a fortification bonus while inside the city of X% (maybe go with 50% like a citadel?).

4) Cities enjoy a CS buff from their garrison the same as they currently do.

With that implementation, capping a city with melee mounted and then moving your melee mounted out would just deprive you of a much better chance at holding the city. If you do leave the city capper in the city then your opponent has to be able to kill that unit 'fairly' like any other unit- no easy free kills just because that unit is in a weakened city.

No more killing a full health garrison for free because it happens to be inside a 0 HP city that you take. No more insta killing a garrison with siege attacks because that garrison is using the city's low CS in the calculation.

 

[Rekk]

It really is odd that a full health garrison is killed no matter what it's health was when the city is captured. This combined with the fact that the garrison doesn't increase city strength ver much and a city without defensive buildings make capture extremely easy, means that killing the garrison much easier than if it had been outside the city.

I think this is an oddity worth looking into, regardless of anything else going on.

 

[065tdsa]

The garrison system looks like a very good system, full health units being killed because the city has no health was always somewhat counter intuitive. But I can't even imagine what it will do to balance. It also does absolutely nothing to prevent you from leaving the city unguarded so the AI can capture it and allow you to kill their unit.

Sure, the unit is harder to kill now, but if I captured the city in the first place I have enough units there to destroy one of the AI's after they've come to the place I was aiming all my siege at. So ultimately, change the system and the fact stands that capturing a city you can't hold is still a bad idea, because it leaves the unit there in a situation where it is out of position and will die.

In comparison forcing the loss of movement points just means nobody should ever capture a city they won't be able to hold for at least one turn with any kind of unit. Why I guess is what a lot of people want, and may not be without merit.

 

[civplayer33]

I think trying to teach the AI to not take cities if it can't hold them will probably have unintended consequences. There are times the AI probably should take the city anyway even if it might not hold it.

Every new AI mechanism can have unintended consequences; doesn't mean we shouldn't improve the AI and this is one area where it can be improved. That's also why I suggested checking first whether the recapture has happened previously, e.g. x times in the last y turns, so the AI can stay more aggressive but it will "learn" as the war goes on, if that city is currently in too bad a position to be taken and then only take it if the surrounding area is secured.

I still think just forcing the loss of all movement points after capping the city is the easiest solution so that at the very least the human is hurt just as much by a city repeatedly changing hands as the AI. Despite all the hyperbole from civplayer33 I don't think it's a change that has any real consequence and it seems kind of logical to be honest- if crossing a river drains all of a mounted unit's movement then surely taking a city (killing the population, breaking their fortifications, etc) would also use up your movement?

Me wanting to play the game the way I like is now hyperbole? Are you kidding me? You need to respect that not everyone plays the game the way you do; not everyone uses exploits or makes the same strategies.

That said, I actually don't mind your new proposal; I don't think it's absolutely necessary but it would add more strategic depth, as @Stalker0 already pointed out, and it improves certain scenarios, though I'd probably boost the defense bonus to 80 or even 100%.

However, changing the way a garrison's health/CS works when attacking a city might be a more complete solution and I share Stalker's interest in giving melee a role to play in city siege. I'd propose this:

1) Ranged/Siege units damage the city first based on the city's CS and benefiting from the normal city attack promotions. If the city has been reduced to 0 HP then they damage the garrison based on the garrison's CS and using unit attack promotions.

2) Melee units damage the garrison first based on the garrison's CS and using unit attack promotions. If the garrison is killed or there is no garrison then they damage the city using city attack promotions.

3) Garrisons enjoy a fortification bonus while inside the city of X% (maybe go with 50% like a citadel?).

4) Cities enjoy a CS buff from their garrison the same as they currently do.

With that implementation, capping a city with melee mounted and then moving your melee mounted out would just deprive you of a much better chance at holding the city. If you do leave the city capper in the city then your opponent has to be able to kill that unit 'fairly' like any other unit- no easy free kills just because that unit is in a weakened city.

No more killing a full health garrison for free because it happens to be inside a 0 HP city that you take. No more insta killing a garrison with siege attacks because that garrison is using the city's low CS in the calculation.

 

[065tdsa]

A large defensive bonus on a city and the ability to switch in a new full health garrison unit every turn on my home ground would give me a huge boost to defense, maybe insurmountable with good geography. The bonus needs to be properly chosen so that the unit can resist one turn against casual attack but against determined attack it must die within one turn, or else the defender will slot in a new one fresh off the production line and make their city impossible to take.

 

[civplayer33]

A large defensive bonus on a city and the ability to switch in a new full health garrison unit every turn on my home ground would give me a huge boost to defense, maybe insurmountable with good geography.

The defense bonus is for the garrisoned unit, not the city. Catapults and other ranged units still attack the city only, which retains the current garrison bonus (and after they bring it down to 0 HP they start attacking the unit only; melee units, on the other hand, will attack the garrisoned unit only, which receives a significant defense bonus. If it didn't receive such a bonus it would die far too quickly. Units in Medieval already have 2 promotions minimum, usually, so those 80% wouldn't even mean 80%, but usually less (depending on what other promotions they have) since it modifies the base strength only, not the strength value that is the result of the other promotions.

 

[065tdsa]

The defense bonus is for the garrisoned unit, not the city. Catapults and other ranged units still attack the city only, which retains the current garrison bonus (and after they bring it down to 0 HP they start attacking the unit only; melee units, on the other hand, will attack the garrisoned unit only, which receives a significant defense bonus. If it didn't receive such a bonus it would die far too quickly. Units in Medieval already have 2 promotions minimum, usually, so those 80% wouldn't even mean 80%, but usually less (depending on what other promotions they have) since it modifies the base strength only, not the strength value that is the result of the other promotions.

My thought was that if the garrison unit is too strong to kill in the turn it goes into the city, you can swap it out for a new one with a full HP bar and handily stall out the conquest of your city for a long time, potentially indefinitely if you're near your unit production areas. But having continued to think of it, that may not necessarily be a bad thing? Making cutting off reinforcements by land and sea more important to taking a city may actually be a good thing.

 

[ilteroi]

lots of interesting ideas. i like the split roles for ranged/melee but i fear it is too complex and invasive.

how about this:

* a city's CS is always at least as high as the garrison's
* damage to the garrison is limited to the amount it would take in a direct attack
* any excess damage is forgotten (not sure about this one)
* you can"t capture a city with a garrison (not sure about that one either)