AI

[Nelphine]

Do you know of a game that consistently successfully encourages the human to maintain garrisons?

I know for myself, I play the same way (no garrisons) in games like Civilization, StarCraft, world of Warcraft (for scenarios where you defend). Invariably, going on offense works better.

The basic problem is, if you can defend successfully with it, doesn't that mean you can attack? And if you have a choice between attacking and gaining territory, or not, then why would you ever choose not to attack?

And if you couldn't attack successfully, then how could you defend successfully? And if you can't defend successfully, then why build them?

A game like StarCraft encourages this in early game by giving you bunkers. Units that literally can't move and vastly improve your other units (one bunker more than triples defense AND improves offense capability of 4 units, and it doesn't cost as much as 3 units.) Compared to bunkers, city walls are pitiful. And even then, bunkers become obsolete by mid game and garrisons get dropped in favor of scouting and maximum offense.

Games like world of Warcraft try to encourage garrisons by putting defense and offense a huge stance apart. Then all units have (almost) a unified speed, and even the higher speedy it's have very calculated amounts of extra speed. Compared to this, wraithform warships or *shudder* a speed 8+ Jafar is insanely fast. Like 'give a wow character permanent sprint and blink on a 2 second cooldown and a warlock teleport circle with no cooldown' fast. 
And even then, wow raids inevitably become 'measure to the precise yard exactly how far you need to go, calculate to the second how long you need to go, and then spend the entire rest of the time on offense.'

Offense wins. If you're not on offense you're not winning. No matter how you go about it, any changes will just lead people to find better ways to go on offense. Without teaching the AI to go on offense (which I agree isn't really feasible), you can't get around that.

But it's a game. I think you should accept that people won't go on offense, and simply move on. Account for it, but don't try to stop it. Encourage garrisons, but don't try to make that a required game design.

[Bahgtru]

Just remember this. Whatever you do to nerf aggressive strategies, you're likely nerfing economic build-up strategies more, and this change would be the perfect example.

[Seravy]

Do you know of a game that consistently successfully encourages the human to maintain garrisons?

I do. That Civ 4 mod I tried. With so many rampaging barbarians, I not only needed garrisons but my garrisons couldn't even keep up with them - in that game the things you build are on the terrain tiles and barbarians destroy them. So while I was able to defend my cities, I still kept losing stuff.

...but this game, you don't need to defend non-city territory. So spawning more rampaging monsters could actually force garrisons, without having any effect on how offense or defense works against the other (AI) players and without causing actual losses to players who do build the garrisons.

So based on your question, I think the only solution is to drastically redesign the rampaging monster spawns - frequency, strength of stacks, etc. It has to be strong enough that you can't kill it using magic alone before the rare spell phase, but you can kill them without major losses if you did garrison the cities. It has to be frequent enough that ignoring it stops being an option. If the AI isn't doing the job, we have to rely on "divine intervention" from the game itself.

We do have the "bunkers" in Guardian, City walls, Walls of Fire, other city defense enchantments, and fortress lighting. Yes, they are far from being as powerful as you describe but buffing them is an option.
All of them rely on the presence of units and increase garrison effectiveness greatly, but they only have the intended effect if they can also deter players from attacking. This is the case sometimes, but probably not often enough - however we have to be careful not to overdo it and end up with a game where offense is not playable at all. (I've seen when Hadriex was playing Civ 4 that it took like 20 turns and an entire hour of playtime to conquer a single city. I don't want something like that, it's insane - I agree, offense is a way to win and needs to remain playable, it just shouldn't be )
Any suggestions on how to improve these that does help the early game without causing late game problems is welcome.

I do believe garrisons must be a required part of the game - the amount of resources generated by not having them is simply too much and I mean the maintenance, not the production cost, which isn't that great. And it works as intended for the late game - Even though I almost always have a doomstack to win, if my defense strategy is bad, I end up losing. Same can be seen in almost all of Hadriex's games - his doomstack works as intended but he loses cities at a faster rate than gaining.
So the problem is specific for the early game. We've cut down raiders and rampaging monsters in the early game because they were sometimes hurting the AI badly - but this had the opposite effect, the human needs less garrisons and thus gains more advantage than what the AI did by not losing some random outposts to them.

I think I'll start a new thread for this discussion now, as we seems to have decided this is not an AI issue.

[Seravy]

It seems the AI is explicitly set to "see" cities if a unit comes in scouting range, or a range of 3 as a feature from the original game. Do we want to change this and limit it to only the actual scouting range?

(and while at it, I might as well ask, are we happy with the large scouting range we have for ships?)

[Bahgtru]

For reference:

Found the procedure responsible for handling AI scouting. It does the following :

procedure(stack, scouting range)

for each non-neutral, non-dead, not own unit, if unit is in range of stack and if neither side is human, set contact between wizards.
for each non-neutral, not own city, if city is in scouting range or in a range of 3, mark the city scouted and if the owner isn't human, set contact.

So yes, what you reported is exactly what is in the code, this btw seems to be code from the original game, not an addition or change. Surprisingly, the "global" make contact checks exclude making contact between two AI players. Also surprisingly, the global one seems to consider invisibility. So you can use an invisible unit to scout without making contact, unless it's buggy, did anyone ever try to do that? It's unexpected - AI procedures don't have invisibility implemented, the AI can and will attack or target these units.

The AI only scouting procedure doesn't care about Invisibility either, not that it really matters, AI's won't have fully invisible stacks on their first contact, that's a very late game thing.
However, it has city scouting specified as "range or 3", intentionally. I can only guess, but probably because a human can also "see" there is a city by using the surveyor if they are in a range of 3. We should discuss if we want to remove this feature and limit the AI to the actual scouting range, in the AI thread.

It seems the AI is explicitly set to "see" cities if a unit comes in scouting range, or a range of 3 as a feature from the original game. Do we want to change this and limit it to only the actual scouting range?

(and while at it, I might as well ask, are we happy with the large scouting range we have for ships?)

The more you know... I haven't ever noticed the surveyor. Can you right click them?

For coherence, it's either one of:

1. get rid of the distance 3 scouting for the AIs and remove surveyor at distance 3
   
2. allow the player to also see cities when they're at distance 3, if it's already visible with the surveyor it means that it is possible to avoid proximity warnings in wizard pacts already - it just needs some dedication with f1.
   

I'd go for 2 because of the hassle with warnings. But I'd also be fine with 1.

Also for the WP warning hassle, I wouldn't reduce scouting range for triremes, perhaps the warship?

[Seravy]

I meant if you use the surveyor, it'll say "cities cannot be built in the range of 3 from other cities", if there is one. So you know there is one, even if you don't actually see where, you can usually figure out the exact tile. It's not actually scouted, you just know it's there. It would be difficult to extend that to actually scout more tiles.

[Bahgtru]

Ah! Nevermind that bit then. The surveyor is cute - it's exactly like playing minefield. although it gets old quickly.

I guess it's getting rid of AI scouting 3 then?

[Nelphine]

Oh that explains a lot. I thought it was scouting range or range 2, so that explains a few times I've been confused.

I'm fine with ship scouting range. 

Hard to say on the range 3 thing. Humans can tell there is a city there, but we can't tell who's city is. We also can't tell what terrain is around to nuke, which (as far as I can tell) is the only thing that scouting a city will actually do (well, all city curses).

The ai can independently target that city regardless of scouting with its stacks (confirm if this is correct please).  The ai doesn't damage wizard pacts by being near a city.

Therefore, since the human can't target a city with city curses even with the surveyor trick, I'd say yes, remove the range 3 scouting.

However, this is going to increase the discrepancy with the ai ability to target units overland regardless of scouting range. But that's a different kettle of fish.

[Seravy]

The ai can independently target that city regardless of scouting with its stacks (confirm if this is correct please).

Correct.

[myk002]

I also vote for removing the range 3 city scouting. It's not realistic for the AI to be able to curse a city it hasn't really seen.