Test games played

[Nelphine]

Finishing my halfling huge specialist sagemaster spellweaver game. I found conjuror and 1 sorcery book so far. It's Jan 1410 and I just completed summoning my first demon Lord.

I also got Pandora's box really early so I went ahead and actually used a hero to test it, and it feels strong. I've got hydra abd demons multiple times out of it, and it summons creatures after each player's turn. If I had a regenerating hero, I would call it completely overpowered. As it is, I can practically guarantee I'll win any battle, but the creatures don't necessarily keep my hero alive so i still run around with 8 other units (who all regenerate).

[Nelphine]

To go back to another discussions: in this death game, now (June 1410) that I control almost all of arcanus (I have 54 cities, 1 opponent is defeated, 1 opponent has 1 city, and 1 opponent has 4 cities), I only now have a larger army strength than the combined army strength of my arcanus opponents.

When I 'merely' had 40ish cities, and the strongest arcanus wizard had 11, he had about 30% more army strength than I did. I never use city garrisons; the AI does.

Don't get me wrong, I think this is how it should be; whomever is winning is likely to attack every single enemy city; on the other hand, whomever is losing is likely not to attack any given city. Therefore, once you're on offense, you can carry this through. However, human psychology kicks in - when they're on offense, they WANT to fight strong garrisons - so we give the AI cheating bonuses to have huge garrisons, without reducing how many doomstacks they'll build compared to how many the human will build. But, since the human doesn't have those bonuses, they can't keep up in doomstack production if they try for garrison construction. So, drop your garrisons.


On a much earlier discussion... demon lords without fire immunity are awfully fragile. It's incredible. You actually need to use stacks of them. They feel like nature or chaos very rares now (or even unbuffed sorcery). Which leaves archangels as a HUGE outlier in performance. I can reliably send 1 archangel against an empire, and expect to take out the entire empire with it, regardless of my opponents books (if I get it before the AI has rares, such as in this case). I don't THINK I can do that with any other very rare (except sorcery can use certain buffs to do this).

[Seravy]

Well, yeah, Life is good at this whole "not dying" business and Archangels are no exception. It's not like the Archangel itself is better than those creatures...but it has unlimited healing, invulnerability, high prayer and supreme light supporting it.

Don't get me wrong, I think this is how it should be; whomever is winning is likely to attack every single enemy city; on the other hand, whomever is losing is likely not to attack any given city. Therefore, once you're on offense, you can carry this through. However, human psychology kicks in - when they're on offense, they WANT to fight strong garrisons - so we give the AI cheating bonuses to have huge garrisons, without reducing how many doomstacks they'll build compared to how many the human will build.

Except that's totally not the design. We give the AI the bonuses because we want the AI to have enough stacks and units to force the human to stay on the defense longer - as getting into the war unprepared would cost more in lost cities than what the human's doomstack can conquer and hold : doomstacks are a limited, expensive resource, while masses of common troops is close to unlimited, and they still allow combat spellcasting, thus allow the AI to win through attrition if the player enters the fight without a large enough stock of resources - both in garrison troops and mana crystals. Basically the bonus is there to allow the AI to attack cities when they are losing, despite having garrisons.

But, since the human doesn't have those bonuses, they can't keep up in doomstack production if they try for garrison construction. So, drop your garrisons.

The design goal would be to drop the doomstacks until the garrisons are adequate since they can't possibly conquer faster than the hordes of AI troops backed by combat spellcasting can against your no garrison empire.

So no this definitely isn't working properly then. And I still don't really understand why. I should be seeing stacks of 9 nagas coming and razing all of the player's cities... On Lunatic the AI can summon 1 each turn so they should have like 3 full stacks by the time the player has the force to attack (nearby starting first AI not included).

[Nelphine]

Note that I wasn't playing lunatic for this game. I'm certainly not confident enough with death to play lunatic regularly.

[Nelphine]

And I think that's the problem. Given the choice of 'drop doomstacks, where you'll only need 1-3 such stacks, and each of those will individually be stronger than anything the AI has' and 'drop garrisons, where - in this case - I'll need 54 such stacks, and most likely a given AI doomstack will be stronger than whatever I decide to use as a garrison' I'm NEVER going to drop doomstacks. It makes no economic sense.

And obviously the question comes in earlier than when I have 54 cities; but even if I only have 5 cities, 5 garrisons strong enough to beat the AI doomstacks is STILL more expensive (and therefore takes longer, and therefore allows the AI to just build MORE doomstacks in the meantime) than the 1 doomstack I need at that stage of the game.

So human builds the doomstack fast enough that you can intercept the incoming AI doomstacks. And since (by definition) the human doomstack is still stronger than the AI doomstack, you eat the doomstack. Then you swing it around like a broom and eat the next AI doomstack.

Oh, and remember how your cities are defenseless? The AI definition of a doomstack is 'a bunch of units that are plenty strong enough to kill their target'. And they choose CITIES as their targets. So, they see my 8 spearmen and say 'yup, these 5 naga are an amazing doomstack'. And so those 5 naga run in. And my doomstack eats it, with almost no damage taken. And the next 5 naga. And the next 5. And each turn, my doomstack gets 1 (or 3 or whatever) tiles closer to the enemy city. And then I eat the enemy city. And then the process continues towards the next enemy city.

[Nelphine]

I think if the AI didn't send attacks based on the strength of the target, I would actually use stronger garrisons. But by having no garrisons, I do 2 things: I lure the AI into never attacking with anything terribly strong, and I avoid the problem of 'I have a garrison as strong as my doomstack. The AI thinks it can attack it. Crap, my doomstack can't fight it either then, particularly since I'll be on offense!'.

This brings me back to trying to make the AI have doomstacks based on the strongest units the human has, regardless of whether they're stacked together. Somehow we need to teach the AI to do what I'm doing - building a doomstack that isn't a helpful target for anything (it isn't guarding nodes or cities or towers, so it has no strategic value) AND it's stronger than anything the AI sends against me AND its secondary goal is specifically to eat enemy stacks that are stronger than what my city garrisons can handle (and since my city garrisons can't handle anything, the doomstack is designed to constantly eat enemy incoming attacks before they reach my cities) - which is exceptionally hard since it demands target switches almost every turn, with a long-term target that I'm still moving towards AND either a constant stream of replacements for casualties, or units that are exceptionally durable (regeneration is amazing for this, but so is ultra cheap bezerkers, where I have a trail of bezerkers behind my main doomstack, kind of like a logistics train - and now that requires us to actually teach the AI that units not in the same stack are actually part of the same stack, they just aren't there yet, and so what that the main stack already has 9 units?).

I don't think we can actually teach the AI this though.

[Seravy]

' I'm NEVER going to drop doomstacks. It makes no economic sense.

I'm not suggesting to not make a doomstack at all - but use it for lairs and nodes until the garrisons are strong enough to risk entering the new war - until you are sure you can hold the cities, one way or another.

Garrisons are made by cities - each of them can produce their own. Doomstacks are made of magic - casting skill. By the time you complete 2 good doomstacks, every city should have their 9 garrison units built up.

Even if you are buffing, the cost of the doomstack is still magic - the buffs themselves. And casting skill is not linear to increase - dropping your 54 garrisons will only get you a little bit more doomstacks, if any.

Of course there are cases when you need magic in your garrison - but you generally only need like one summoned unit (or city enchantment) per city, while the doomstack needs a full 9 or close to it. In an already ongoing war, the ability to defend 9 cities immediately instead of losing them in rapid succession is far better than the prospect of having a stack that will destroy one enemy city a turn (not conquer-holding it would require a garrison which you don't have) one year in the future.

Note garrisons don't need to beat real doomstacks (like 9 efreets) - the AI has a very limited quantity of those and losing 1-2 cities while the garrison holds them up and wears them down one unit at a time is acceptable. What they need to beat are those stacks of 9 stag beetles, halberdiers, manticores, griffons, paladins, etc. - and no, spearmen won't do it, no matter how good your combat spells are. You might be able to kill most of the stack but you still lose the city (or all the buildings in it).

Oh, and remember how your cities are defenseless? The AI definition of a doomstack is 'a bunch of units that are plenty strong enough to kill their target'. And they choose CITIES as their targets. So, they see my 8 spearmen and say 'yup, these 5 naga are an amazing doomstack'. And so those 5 naga run in. And my doomstack eats it, with almost no damage taken. And the next 5 naga. And the next 5.

Yeah but that shouldn't actually work. Those nagas won't be all coming from the same direction and won't be coming one stack at a time. And they'll keep coming as long as you are killing them instead of attacking the enemy. This is exactly the AI's strategy - force the player to be on defense as much as possible. Doesn't matter if it's in the city or outside, as long as it keeps the player's forces busy, it good enough.
But whatever, your doomstack kills all the nagas somehow, I don't get how you have a stack that moves 6-8 tiles each turn on water to do it but ok.
Meanwhile the AI's 15 stacks of normal units - halberdiers and wolf riders and such - conquers all your undefended cities anyway.

So I'd like to know two things - How could that doomstack be mobile enough to intercept all enemies, and how is it the AI doesn't send more units than what the doomstack can intercept? They have the bonus plus they use cheaper units - I mean their normal nondoom stacks - they should be sending like 4-5 stacks for each doomstack of yours.

This doesn't compute and the only real way I see how it can be possible is the human actually having MORE resources than the AI in the early game. The AI is spending thousands on mandatory stuff - fighter's guild everywhere, 1 new setter every 5-6 turns, 4 swordsmen, random economy buildings, etc - while the human gains a lot of extra resources (treasure) the AI does not, pours all into military and stacks enough economy multiplier retorts - military retorts count, you get more military force for the same amount of money, that is an economy multiplier - that they actually have a larger percentage bonus than the AI through them, for those very early turns only. (If this is indeed the case, the solution is to have the AI start with more stuff on higher difficulties but that has problems as well.)
This also would explain why none of the posted Lunatic wins are late wins, and why it's always an early strategy/rush win of some sort. There is no way an AI with only 150-200% resources can compete with someone who spends everything they have into their military, from starting picks to the last mana crystal - and the AI isn't allowed to do the same as that would make them lose if they don't fight the player early, and would make the player lose if they weren't playing the early rush strategy but end up in a war anyway - both undesirable outcomes. 

Somehow we need to teach the AI to do what I'm doing - building a doomstack that isn't a helpful target for anything

Impossible, the human can and will find a way to tear that stack apart with combat spells anyway. AI doomstacks are far less effective by definition, almost every doomstack is weak to at least something.
In worst case, if we succeed which is fortunately impossible, all but Chaos/Sorcery magic becomes unplayable, as those are the only two that can kill anything using combat spells, no matter what. But those are too slow and would lose anyway before reaching those spells.

Your doomstack isn't a target not because it can't be damaged by the AI (which is often the case for Life not so much otherwise) but because the AI doesn't know how to damage it. You don't attack into a stronger stack unless you're 100% sure you'll be doing damage and the AI isn't that smart to recognize that situation, nor will it ever be.

[Nelphine]

I mostly agree with your reasoning, except one important thing:

while the human gains a lot of extra resources (treasure) the AI does not, 

That's simply false. I don't think the early side has ever had less treasure than myself, particularly on lunatic. On lunatic I get almost no easy treasures.

Later in the mid game, I get more treasure than the AI, but that's far too late for the snowballing you describe.


The other big thing is this: how long does it take to 

Ugh my phone screwed up. Will need to continue in next post.

[Nelphine]

In the above post 'early side' should be 'early ai'.

Garrisons do matter. For me, for example, a garrison of 9 halberdiers simply loses. It doesn't hold anything. And probably you want alchemists guild and barracks (but reasonably, war college) for your halberdiers to even successfully not do anything.

So I take that 1250-1700 production and I build 8 spearmen. I then also build a shrine, a Parthenon, a builder's hall, a wizard's guild, a university (1160 production) and a cathedral to replace the war college.

That's FAR more magic for your military.

Do that in 6 cities, and for the same cost as your basic garrisons that can't fight 3 uncommon summons, I'm getting 132 power, 54 research, and 18 less unrest.

[Seravy]

Ugh Halberdiers are the worst garrison ever. I would never use them for that unless I already have a fighter's guild (and alchemist if not having Alchemy) in which case it only costs 540 production for a full stack. Levels I ignore - turns give experience as well as defensive battles so either they won't fight at all, or will get those levels from the battles. I might build the barracks if I expect the city to get attacked soon, but not the war college.

However, I mostly use something better than halberdiers and I usually distribute them from a few cities. So I usually don't need any buildings at all - I already have them, and the buildings required also give me economy - wizard's guild for magicians, parthenon for priests, forester's guild for longbowmen and horsebowmen, etc. or they come from cities that have no economy building options, or only minimal. Barbarians are obvious for berserkers, Gnolls have very limited economy buildings and they aren't that great - no wizard's guild or advanced buildings. Meanwhile they produce great units, and if needed, they can start doing so with only a stables (for wolf riders). Klackons have high production, growth and not so much buildings - you'll often have klackon cities already maxed out on economy buildings ready to produce beetles or halberdiers having nothing else to do anyway. In those "maxed economy" cities, the only thing you can be missing out on is Trade Goods which is not that great.

Still, the cost if converted to economy buildings seems like it's a lot - but it isn't. It won't give you a free building for eternity, it only allows you to get the building as many turns earlier as you'd take to build the garrison.
So for example a garrison of 9 Magicians is 1080 production - A wizard's guild and a cathedral approximately.
A lot, isn't it? Not really. Assuming the city is producing ~50 hammers, that takes 21 turns. So you missed out on 21 turns of wizard's guild and cathedral - 210 power and another 126 power but also the 126 maintenance. That's a net gain of 216 power = SP. At a skill of 100, that buys you 1 casting skill. Even assuming you used that  skill for another 100 turns (you did not, you are playing aggro the game is short) - it still didn't even summon half of a creature for your doomstack. Assuming a doomstack of 9 rares - 300 skill each - it's 27 cities garrisoned for one such stack, if and only if, you wait a whole hundred turns to utilize the skill. So the choice is "I defend my entire empire now, or I get an extra doomstack 100 turns in the future", in reality. (Ok, I forgot to include the 3 RP from the wizard's guild but it's not the amount that would make a huge difference...)

So yeah, garrisons are definitely worth it, IF you do it right. Build them in a city that already can do so. Even better if that city is already maxed on economy buildings. Don't bother with a war college - in 80 turns the units will be naturally veterans, or faster if they beat an enemy. Even if they aren't, it doesn't matter - About 50-75% of your strength will be your combat magic, so those levels aren't that significant. City Walls are worth many times more than the war college and cost much less anyway - 3 defense is far more than 1 level.