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

Create an account  

 
AI

Lots of 'magic spirit' efforts, land linking, several 100-250 mana cost spells like pestilence/famine, iron skin, great lizard, etc. I don't think a single disjunction.

Freya is often smarter, focusing on colossus.

Edit: oh no! She cast 'nature's wrath'

Reply

Disjunction should be frequent, unless they are allied with the owner of Suppress Magic.

Magic Spirit is not disabled - However if the AI does know any good, high priority summons, the loss on those is marginal (and if one gets through and melds a node, it might be worth it).

Land Linking is not on the disabled list - I rather have the AI fall back to do this, than have everything disabled in which case they'll.., what will they do...I think, cast "None", skipping the entire turn - which is worse.
The problem is buffs are ALL cheap - but the question is not "do I cast any buff", as that is already rolled, but "do I cast this buff, or the other".
However, we can, if we want, set a rule of "Buff spell category priority = zero if any suppress magic is on". I feel that's a bit drastic though.


Currently these are listed in the "buff" category (as they offer by far the lowest cost, at least relevant effect in lategame) :

mov word ptr [0D388h],0 ; Resist Elements
mov word ptr [0D392h],0 ; Resist Magic
mov word ptr [0D3A0h],0 ; Holy Weapon
mov word ptr [0D3A4h],0 ; Holy Armor
mov word ptr [0D3A6h],0 ; Heroism
mov word ptr [0D3A8h],0 ; True Sight
mov word ptr [0D3B0h],0 ; Cloak of Fear
mov word ptr [0D3B6h],0 ; Waterwalking
mov word ptr [0D3C4h],0 ; Planar Travel


This is the curse priority :

mov word ptr [0D388h],0 ; Blizzard
mov word ptr [0D38Ch],0 ; Spell Blast
mov word ptr [0D38Eh],0 ; Stasis
mov word ptr [0D392h],0 ; Corruption
mov word ptr [0D396h],0 ; Raise Volcano
mov ax,100
mov bx,0D39Ch ; Call the void +100
call addto
mov word ptr [0D3A2h],0 ; Black Wind
mov word ptr [0D3A4h],0 ; Drain Power

Again, in this category the best the AI can do is Call the Void (cost 500), but all other "second best" options are 100-300 mana. They do cripple the human player enough though to be worth it, even if the success rate is low - which is not the case for buffs.

Finally Lizard, let's see, creatures :

mov word ptr [0D388h],0 ; War Bears
mov word ptr [0D3C2h],0 ; Nagas
mov word ptr [0D38Ah],0 ; Sprites
mov word ptr [0D38Ch],0 ; Giant Spiders
mov word ptr [0D39Ah],0 ; Floating Island
mov word ptr [0D3A2h],0 ; Hell Hounds
mov word ptr [0D3A4h],0 ; Gargoyles
mov word ptr [0D3AAh],0 ; Fire Giants
mov word ptr [0D3BCh],0 ; Skeletons
mov word ptr [0D3BEh],0 ; Ghouls
mov word ptr [0D3CEh],0 ; Magic Spirit
mov word ptr [0D3DCh],0 ; Cockatrices

shl word ptr [0D392h],1 ; Gorgons x2
shl word ptr [0D394h],1 ; Behemoth x2
shl word ptr [0D396h],2 ; Colossus x4
shl word ptr [0D398h],2 ; Wyrm x4
shl word ptr [0D39Eh],1 ; Djinn x2
shl word ptr [0D3A0h],2 ; Sky Drake x4
shl word ptr [0D3B0h],1 ; Hydra x2
shl word ptr [0D3B2h],2 ; Great Drake x4
shl word ptr [0D3B8h],2 ; Archangel x4
shl word ptr [0D3C4h],1 ; Wraiths x2
shl word ptr [0D3C8h],2 ; Death Knights x2
shl word ptr [0D3CAh],2 ; Demon Lord x4

So lizards are not disabled, but not prioritized either - if the AI casts it a lot that means they don't have any of the listed, expensive summons, otherwise they'd be picked at least 4-10 times more than lizards.

...so yeah, the question. Do we want the AI to never cast a buff if there is a Suppress Magic, or at least do it at a massively reduced chance? (While not optimal, some buffs, like magic immunity, regeneration, invulnerability, etc are worth having to do multiple times to get it work once on a hero - not for anything else though.)
Reply

Decided to go ahead with the difficulty change right now.

These are the existing values :

Pop Growth 75 100 120 140 150
Casting discount 100 80 66 50 40
Production 75 120 140 166 200
Gold 100 100 140 180 250
Power 75 130 160 200 250
Research 75 100 140 160 200
Food 100 100 100 100 100
Maintenance 100 60 50 40 30

And this is the proposed new table :

Pop Growth 75 100 120 130 140
Casting discount 100 80 66 50 40
Production 75 100 125 150 175
Gold 100 100 140 180 225
Power 75 120 160 200 240
Research 75 100 140 160 200
Food 100 100 100 100 100
Maintenance 100 60 50 40 30

Furthermore, as the goal of "Impossible" is not to be impossible, and also because it might mislead people into thinking it's the same difficulty as in vanilla MoM when it's very far from it, I think this level needs to be renamed to one of the following :

Lunatic
Nightmare
Phantasm
Champion
Kamikaze
etc...
Reply

I'd like to reiterate that I think overland casting skill is the single biggest culprit in the difficulty right now. Changing everything else is nice, but I don't think changing everything else will have as big an impact as simply changing overland casting.

The ability to do one turn disjunctions is probably the single worst thing about high difficulty AI, as it effectively removes globals from the human spell list.
Reply

(May 10th, 2017, 06:21)Nelphine Wrote: The ability to do one turn disjunctions is probably the single worst thing about high difficulty AI, as it effectively removes globals from the human spell list.

I'm still waiting for your suggestion on disjunction priorities. This is not a problem than can or should be addressed through reducing overland casting skill, which would make the AI worse at summoning and buffing.

I'd like to point out I was successful at using an all-chaos global enchantment strategy before. Sure, I needed to recast Meteor Storm quite often but it did work. (might not work without Specialist though. That dispel resistance is important. )
Reply

And for every success story, there is the bad story, where the human spends 15 turns casting and recasting a global, and never has it in play longer than the turn it completes being cast.

And yes I'm aware that reducing overland casting will reduce summons and buffs. I don't mean get rid of the bonus altogether. But spells are more important for the AI than all the things a city does.
I'm also aware you want better numbers for priorities. The '1 turn dusjunction' isn't a problem because the human can't cast spell blast - in theory most humans don't have spell blast - the problem is that means the AI has so much casting skill that it would need to cast disjunction less than 1% of the time for it not to be regularly cast within 2 turns of a global being cast; which is an unrealistically low priority. And that means the AI simply needs to cast less spells, which means a lower overland bonus.
Reply

I would like to see overland bonus:

Current
Casting discount 100 80 66 50 40

New
Casting discount 100 90 75 65 55
Reply

Note, I've managed to figure out how to play impossible again. But, the game is just brutally fast. 1408, I have over 20 cities (never built a settler), 20 nodes; I'm not in first place. I have over 1000 power production and I could have 200+ casting skill, but I had to slow down for a while because I'm in too many battles and need too much mana.

I feel like the higher difficulties just throw away the timeline the game is aimed for. I can't imagine not winning (or losing) before rares happen (and very rares, forget it), just because there are so many units around.

Basically, you end up playing by making the AI do all the work, while you focus entirely on attacking so that you can reap the rewards of the higher production rates. (And, in a strange turn of affairs, this means that having your capital on a crappy island in the middle of nowhere is actually much more important than having a good starting tile with good follow up city spots. By being on a crappy island, the chance of your capital being chosen as the main action continent by an AI is much lower. It also means that, in the long term, long term races are bad, because you can never remotely keep up with the AI, so the only long term strategy is that you must have a strong attack force before they can shut you down so you can steal a long term empire. Defeating an AI by 1403 is almost mandatory, and my personal goal is to aim for 2 defeated AI by 1404. This game I only got one, because the second one had a single city on Myrror due to Astral Gate, but that's almost the same thing as defeat.)

This also means that once an AI is weakened (say down to under 5 cities), all of their expansion (because they still build settlers like mad) ends up just being more power for the human. It doesn't help the AI at all.

To actually keep the difficulty, I think AI need to have checks on when they use settlers. I have ideas for what these should be, but I'd like to hear other thoughts first.
Reply

Ok, I was slightly wrong. The myrran wizard has efreets in 1409; and I have life force (researched) cast in Feb 1410. So very rares will come into play as the game is finishing on impossible. Obviously it depends on exact treasure - I got lots of nifty items this game, instead of books, retorts, or very rare spells. (My opponent got 2 books out of treasure, and I got 1 death book.)
Reply

For some more specifics; impossible AI seems to start seeing widespread amplifying towers in about 1406. (Not everywhere, because they're obviously still expanding like crazy at that point, but, that's definitely when I've started noticing that I get to steal them regularly.)
Reply



Forum Jump: