As a French person I feel like it's my duty to explain strikes to you. - AdrienIer

Create an account  

 
AI vs global enchantments

As some of the global enchantments clearly have a major effect on what spells you should or should not cast on the overland map, there is no way around it, the AI needs to know this.

Divine Order was already implemented a long time ago, making AI players cast more enchantment type spells and fewer summoning.

There is no good tactic to get around Evil Omens - everything costs more equally.
Herb Mastery makes Nature's Cures worthless but the AI never uses it anyway so it doesn't matter.
The AI already does not cast Chaos Channels during their own Doom Mastery, or Raise Volcano during Armageddon or Corruption during Great Wasting. It also already considers Detect Magic when using Spell Blast. It casts more summoning spells if it has Survival Instinct, too.

Zombie Mastery, Aura of Majesty, Time Stop, Enlightenment, Life Force, Crusade, Charm of Life have no effects on spellcasting.

This leaves Holy Arms, which should disable Holy Weapon, and Meteor Storm, Suppress Magic, Nature's Wrath, Chaos Surge and Eternal Night.

I've already implemented the following effects :

Quote:-The AI will cast the following unit enchantments at a higher priority if Meteor Storms are in effect : Elemental Armor, Bless, Resist Elements, Iron Skin, Regeneration, Magic Immunity, Invulnerability.
-The AI will cast more Chaos Channels if Chaos Surge is in effect.
-The AI will almost never cast the following spells if Suppress Magic is in effect and another spell of the same category is available : Resist Elements, Resist Magic, Holy Armor, Heroism, True Sight, Cloak of Fear, Waterwalking, Planar Travel, Blizzard, random Spell Blast, Stasis, Corruption, Raise Volcano, Black Wind, Drain Power.
-The AI will not cast Call the Void while Nature's Wrath is in effect. The AI will cast it at a higher priority if Suppress Magic is in effect.
-The AI will never summon the following creatures if Meteor Storm is in effect : War Bears, Nagas, Sprites, Giant Spiders, Hell Hounds, Unicorns, Skeletons, Ghouls, and will summon Cockatrices half as often.
-The AI will summon Chaos creatures at a higher priority if Chaos Surge is in effect, with greater focus on Hydras, Chimeras and Gargoyles.
-The AI will not summon the following creatures during Suppress Magic : War Bears, Nagas, Sprites, Giant Spiders, Floating Island, Hell Hounds, Gargoyles, Fire Giants, Skeletons, Ghouls, Magic Spirits, Cockatrices.
-The AI will priorize summoning these creatures higher during Suppress Magic : Gorgons, Behemoths, Colossus, Great Wyrms, Djinn, Sky Drake, Hydra, Great Drake, Archangel, Wraiths, Death Knights, Demon Lords.
-The AI will not cast the following spells during a Nature's Wrath : Storm Giant, Chaos Spawn, Angel, Wraiths, Summon Champion, Enchant Item, Create Artifact.
-The AI will cast the following spells during a Nature's Wrath at a lower priority : Djinn, Sky Drake, Hydra, Great Drake, Efreet, Archangel, Death Knights, Demon Lord, Resurrection.
Note that all of these are for the priority of the spell within its own category (buff, curse, summon etc). None of these spells change the priority of any of the main categories, only the spell selected from the category itself ones it was chosen. Also note that some categories redirect to others if no spell is available from them, summoning and city buffs in particular. Other categories do not so the chance of spells cannot be reduced to zero in them, as that results in the AI failing to cast any spell for the turn if the category is picked.
I'm not sure if implementing checks for Eternal Night is particularly useful, if it is the AI's own, it's likely to spam Death Wish often enough anyway and Black Wind doesn't have a smart targeting. Summoning more Death creatures might be a good idea but 1/1 isn't all that much of a boost to really matter. I rather be lazy and not bother with this one.
The AI also does not case Summoning Circle if the player has Suppress Magic in effect. Otherwise it doesn't matter because it costs nothing even if it gets countered, but the message spam is annoying.

If I missed anything important the AI should consider, or if you think any of the above are bad decisions, let me know. (Note that this is for overland spells only!)
Reply

I don't think the AI should ignore corruption and raise volcano just because of the very rare versions. Targeting ore is always always important to chaos.
Reply

Id also like to take a look at disjunction. It's too effective at removing globals. Cast a global for x, opponent (who usually has more casting skill or else they aren't a threat and aren't who the global targeted snywau) casts disjunction for x, global is gone. So the person who cast the global spends more time, and doesn't get much out if it; so the primary purpose of globals ends up being to use up your opponents skill, but since they have more to start you are hurting yourself more than them.

especially when you consider four possible opponents, this seems flawed.
Reply

I actually was wrong about that one. It's only Raise Volcano and Armageddon. Corruption can still be used at any time which is enough to hit ores.


Quote:disjunction for x
It requires 1.5*x for 100% success and the AI intentionally never goes for it that cost. It always uses Disjunction with the slider to the half. That's 1125 power which is not enough for a 100% chance against most spells, more like 50-75%. Only very cheap spells are going to get removed at a 100% chance but those have a low priority to get targeted.
If I reduced Disjunction effectiveness it would be useless for the human player, it already is rarely a good choice.
The AI's priority to cast a Disjunction is intentionally set pretty low (except for some really nasty enchantments) to avoid this problem on their side.
The priorities are listed in this post, feel free to post suggestions on which needs to be changed in the AI thread.
http://www.realmsbeyond.net/forums/showt...#pid598778
The numbers are the amount of chance the AI picks "Disjunction type" for casting a spell, while the typical sum of priorities to select from is about 300-600. So a spell with "10" listed will trigger a disjunction 1/30- 1/60th of time from an AI whenever it casts a new spell. Also, allied enchantments do not count and are never targeted with Disjunction.
If you still think Disjunction is overpowered, reducing the casting cost is an option, as it causes the AI to fail more, however that also means the human player cannot dispel stuff that matters at a 100% chance. I think committing over 1000 mana is not a decision one should make based on random chance in a strategy game.
Reply

Sigh agreed. The main problem is the disparity on overland casting skill on higher difficulties, so even when it's random, the AI still feels like it's spamming it when it can cast thousands of mana per turn (which isn't uncommon by the time the human gets very rare globals.) I'll read that thread and try to see if I can come up with coherent suggestions.
Reply

I think the other changes in response to globals are good ones.
Reply

OK read the link. Remember reading it now; makes more sense with context. I think my problem is experience in games where the AI can cast 1500+ worth of spells per turn and most of the AI are allied. With 2500+ worth of spells per turn being cast, those dusjunctions feel far more common than the 5-10% per cast they get rated, since the AI checks which spell to cast so many times per turn. I think I simply have to chalk it up to a poor experience pool on my part; your reasoning, both there and here, is still sound.

I think the only way to get around it is to artificially reduce skill/mana compared to research. Due to the high pressure and intensity of the mod, doing things as fast as possible is top priority - which leads to very Rares being late enough in the game (whether 1415 on impossible, or 1425+ on lower difficulty) that people have a huge casting skill compared to the cost of those very Rares (Even if it's lower than the cost of the very rare it's still going to be a significant portion, which makes an early disjunction reasonably likely, especially with multiple opponents.)

And I don't think changing the pace of the game would be a good idea.

So, I will accept disjunction being cast more often than I think fun as a natural result of the game that has been designed.
Reply



Forum Jump: