AI

[Seravy]

One more thing I discovered in a game I played today about strategic combat.
The number of ranged turns were set to 3, but that's ridiculously low. Yes it sounds about right (units fire at each other for 4-8 turns in normal combat) but truth it each strategic combat basically rolls 1-10 for each army and they do that many percentage of damage of their relevant attack rating. An army generally has about as much "health" as "attack" so a battle of average units lasts until about 100% damage is done. Numbers aren't exact ofc, some armies have more health, others have more attack, but anyway. A turn does 5.5% on average. So 3 turns are 16.5% out of the 100% of the combat. In other words melee decides the outcome pretty much on its own since it is about 6 times more relevant than ranged.

In the game I did, there was an army of 9 sprites defending, and the attacked has 2 units, a dark elf cavalry and a naga. And the attacker won. I peeked into the internal numbers, the sprites army had some 934h ranged attack vs the 320h health and 9Eh ranged of the attacker. In other words they had a 10x superiority in ranged attack power, and the enemy had a pretty low health, only 1/3 of the attack power. They should have lost with these numbers. Their melee was however superior, about twice as good as the sprites, and the sprites had barely more health than the 2 other units since yes, they are that fragile. So in a melee combat the 2 units would have had a reasonable chance to win based on the numbers.
This clearly indicated the 3 ranged turns are not enough, since 10x ranged superiority was less decisive than a 1.5x superiority in melee. (the two wizards had roughly equal magic power too)

With 10 ranged turns, I got the expected outcome of the sprites winning, and when sending my army of strong melee units against a stack of bowmen and shamans, the melee units were able to win without taking much damage so it's not excessive, tanky melee units can still win against weaker ranged units easily.
I don't know what the ideal number is, since that's subjective. The higher we set it, the more advantage ranged armies get over melee armies, and this entirely depends on what we want.

[namad]

I guess there's no way to use movement scores to determine it? 3 ranged turns against cavalry is about right but against nagas it should be more like 6 and against dwarves it should be more like 20....

but, uh, spirites only have 4 ammo? there's no way to use ammo in the math is there? if a unit has 4 ammo it should get 4 ranged turns?

Although also, does this mean flight does nothing in strategic combat? because nagas should never be able to harm sprites ever in any situation? Also it explains why in my current game minotaurs keep attacking my doom drakes forcing me to boringly hit end turn 25 times.

[Seravy]

I guess there's no way to use movement scores to determine it? 3 ranged turns against cavalry is about right but against nagas it should be more like 6 and against dwarves it should be more like 20....

but, uh, spirites only have 4 ammo? there's no way to use ammo in the math is there? if a unit has 4 ammo it should get 4 ranged turns?

Although also, does this mean flight does nothing in strategic combat? because nagas should never be able to harm sprites ever in any situation? Also it explains why in my current game minotaurs keep attacking my doom drakes forcing me to boringly hit end turn 25 times.

There are no units, no movement and no ammo.

there are 2 armies, an attacker with a total melee, ranged and durability, and the same 3 numbers for the defending army.
Nothing has effect unless it gets converted into the attack, the ranged, or the durability.

figures affect all 3 stats. health and defense affects durability. melee attack affects melee, ranged attack affects ranged. attack flags affect the appropriate attack type. first strike affects melee. poison, gaze, thrown or breath adds to melee on top of the normal value.
Every other ability or stat has no influence at all.

At the end the lost durability % is dealt as actual damage to the winning army divided randomly between units.

[namad]

Is it possible to do something weird like count movement as melee attack rating? or count ammunition as ranged attack rating?
it's probably not possible, but if those are the only three numbers compared ideally you'd do something like set melee attack damage=melee attack damage+movement-2 and ranged attack damage to something like ranged attack damage+ammunition-4.

I guess all that means that one minotaur probably kills 9 sprites even though in a real battle it'd never harm them at all? eh? I guess that explains why berserker's always wreck face against sky drakes and great drakes and such.

[Seravy]

I think I'll be going with this list :

Nature
1. Sprites
2. Water Walking
3. War Bears
4. Web
5. Call Centaurs
----------------------------------------this many books guarantees all 10 commons for research
6. Fairy Dust (worth having as a starting spell, very effective against early game units)
7. Nature's Eye (research the rest faster)
8. Resist Elements (Magicians are not likely in the early game, sprites and shamans too weak to need the spell)
9. Earth to Mud
10. Earth Lore

Sorcery
1. Confusion (human player will want to eliminate sorcery in early game while they are weak. This makes that tactic way less effective)
2. Floating Island
3. Nagas
4. Focus Magic (first two picks do not help in overland military power, this does)
5. Psionic Blast (Confusion is way better.)
----------------------------------------this many books guarantees all 10 commons for research
6. Resist Magic (Can help against shatter, black sleep, confusion, Life Drain, all other threats are uncommon+ but there is a chance the AI does not pick it in time for research so might as well get it earlier)
7. Guardian Wind (can potentially make a difference in the early game against bowmen/slinger rush, way less useful later)
8. Phantom Warriors (AI uses this poorly but it can something kill a strong hero by luck)
9. AEther Sparks (low cost can be decent for early game damage. Nice if it hits caster heroes.)
10. Counter Magic (AI needs higher than starting skill to even consider using this, not important to get early)

Chaos
1. Fire Bolt (most cost efficient nuke in the game)
2. Wall of Fire (amazing protection against early attacks from the player)
3. Hell Hounds
4. Fire Elemental
5. Corruption (disable that lucky early mithril/gem/adamant that wins the game on its own)
----------------------------------------this many books guarantees all 10 commons for research
6. Shatter (can be very good in early game)
7. Warp Wood (stop bowmen/slinger rush)
8. Eldritch Weapon (can counter weapon immunity by chance)
9. Warp Creature
10. Disrupt

Life
1. Heroism (extreme power in early game, especially on heroes)
2. Just Cause (Get more heroes to buff by the above, also good economic benefit)
3. Healing (so the super-powerful early hero cannot be killed)
4. Heavenly Light
5. Endurance (very amazing buff that makes a difference even in late game)
----------------------------------------this many books guarantees all 10 commons for research
6. Holy Armor (combo with 1 and 5, most relevance in the early game, and adds military power)
7. Holy Weapon (bonus to hit now adds to military power, counters early weapon immunity)
8. Star Fires (can kill scary enemy fantastic units)
9. Guardian Spirit (10 book life at least should have that one summoned creature until they produce more units)
10. Bless

Death
1. Ghouls
2. Wraith Form (water walking is essential, weapon immunity is amazing)
3. Black Sleep
4. Summon Zombie (strongest common combat summon and creates undead!)
5. Darkness (very strong with ghouls and undead which the AI is guaranteed to have)
----------------------------------------this many books guarantees all 10 commons for research
6. Life Drain (nice combat spell for early game)
7. Skeletons (not essential but helps having more targets for darkness to get cast)
8. Weakness (not very good but is cheapest, might be able to cast when not enough skill for anything else. Early game units have low enough stats for this to really hurt them)
9. Cloak of Fear
10. Mana Leak (not an early game spell, can research later)

[namad]

Any ai-mage with exactly 2 picks in blue/green/black and all the rest in white/red who starts on an island might as well be removed from the game. Roughly. I've honestly never seen an island start not cripple the AI, even if they get off it reasonably fast it still harms them a little.

On the flip side any wizard with 2picks in blue/green/or black and 3picks (or more) in another blue/green/black will get a free awesome spell instead of a useless duplicate get me off this island spell.

Even if a wizard doesn't start on an island they can do interesting things with water walking settlers over to some coal or gold on an island.

I'm happy with it like this, although I'll probably stick to playing large landmass instead of fair. I'm not sure having it the other way would change my mind though.

[Seravy]

I think the AI needs to somehow learn to escort their settlers with a few units so they're not that vulnerable, and the new city doesn't stay empty for the first 10-15 turns. No idea how, though.

[namad]

I think the AI needs to somehow learn to escort their settlers with a few units so they're not that vulnerable, and the new city doesn't stay empty for the first 10-15 turns. No idea how, though.

Well if you snipe that free unguarded settler/outpost you gain almost nothing, they lose something they have many of, and then you go to war with them and they'll counterattack.

Usually by the time they're at hamlet the have a single guard, also not sure how you'd fix it.

[Seravy]

I'm wondering, should the AI prioritize settlers when casting Wraith Form overland, like it does for Water Walking, or should it stay a generic buff that goes on the strongest military unit (since it's actually quite relevant as a combat buff, granting weapon immunity, poison immunity, magical weapons AND immunity to a lot of nature magic, plus it helps actual armies go across water which is not all that much worse?)

[Tiltowait]

Keep it as a combat spell. By the time the computer gets it, there will be swarms of settlers everywhere.