Please sort out the AI

[bman654]

fyi I think most of the battle AI is written in lua scripts in the ses.kfs file if someone fills like improving it.

@littleduckie: What do you mean about the stonewall+trap? I've been trying to figure out how to use Trap and haven't had much success. And the stonewall that the rage spirit produces just sucks from my experience--one hit and it goes down. If you have some way to use it effectively I'd like to know

[Likethat]

Originally Posted by bman654

fyi I think most of the battle AI is written in lua scripts in the ses.kfs file if someone fills like improving it.

@littleduckie: What do you mean about the stonewall+trap? I've been trying to figure out how to use Trap and haven't had much success. And the stonewall that the rage spirit produces just sucks from my experience--one hit and it goes down. If you have some way to use it effectively I'd like to know

i think what he means is that you make the stonewall and use the spell trap to put in the cell in front of the wall

the spell trap works like this: use the spell on the selected cell and if ANYONE ( including your own) except for flying creatures steps on it.. they take damage and lose their turn.. so its actually pretty smart because then it will take 2 turns at least to destroy the wall.. hadn't thought of that myself

[Yaster]

Originally Posted by Likethat

the spell trap works like this: use the spell on the selected cell and if ANYONE ( including your own) except for flying creatures steps on it..

Actually it isn't working on soaring creature, not flying, you just to have predict where flying creature will land. Seems pretty hard to use but it isn't if you give it a thought. The key is to slow creature so it has limited options to choose. It is possible even too predict black dragons. Personally I use royal thorns as a bait, dragons really likes to fry them and their offspring.


But on topic: little story about some of the good stuff AI is capable of. At some point during some random fight I noticed a flaw, as I believed at first, in AI as it's dragon breath reached it's own vampires, nearly killing them of. Then after short interval there was the turn of huge stack of necromancers. In a moment flaw grew into brilliant tactical move, nearly succeeding in giving me a big headache. At least I hope it was tactical move and not big screw up, because you know, AI is machine, not some lucky human.

I believe that kind of tricky moves give player more than the lack of some clearly visible tendency to make bad moves. Good AI in video game is not the one with acts the best in current situation but the one works to enrich gameplay. Actually, as I see it, singleplayer games rely on providing player with source on with he could work out exploring and then utilize gathered knowledge in exploiting the system. I recently replayed Fallout and seeing it's walkthrough It's crazy how most of things can be exploited in some bizarre manner - "silently" killing people by giving them armed dynamite or putting it near them hidden in the bag, super stimpacks overdose etc. It's seems that It is one of the parts people enjoy most. And I really feels that this mechanism works great in KB:L. There is huge amount of stuff to discover. And the game don't patiently wait till you find a moment to search for them all by yourself but instead constantly gives you new and new challenges to make you search. And It gives huge backslash of enjoyment. The progression of difficulty is truly superb. And one of the parts to learn is AI, the way how it acts. The hypnotize is one of the stuff easier to spot, but I strongly believe there is whole new layer deeper. At least that mine impression and that's how I think good singleplayer AI should be like.

[waveform]

Amen to that.

(Following is entirely speculation on my behalf)
Also, AI has different priorities than a human player. AI focuses solely on dealing maximum damage whereas human player *probably* has other grand scale priorities. I might protect solely a particularly expensive troop or a troop that's tedious to replace. Hell, I tend to protect my Evil Beholders by any means in my current game only because the place to fetch reinforcement E.Beholders is so damn far. In King's Bounty the AI doesn't have to calculate the cost-effectiveness of its own troops. Maybe the AI attempts to deal maximum economical damage to the player by attempting to kill player troops that are rather expensive to replace. Maybe the random enemy heroes have individual AIs and spell and tactics preferences.

I could come up with a thousand "maybes" like these since I have no knowledge of the actual AI. The point is though that what seems like a dumb move on AIs behalf is actually the best choice available for the AI currently. We can't tell how the AI sees its own situation and what options does the AI think it has. And it might prefer particular tactics. But... we can only guess what the AI's priorities are. And they might change from battle to battle. Not to mention if there's randomness involved, for example if there are two seemingly equal options for the AI, will it choose one randomly or does it have some sort of priority list which it will then check and determine the course of action.

Aaaanyway, the Hypnosis spell... It's a classic spell from HOMM 3 at least and removing or changing that would be killing the spirit and the soul of HOMM 3. And we don't want that, now do we. In fact, it works wonderfully now since the ability to use that is tied to the leadership count. I suppose the next step in that spell is to tie the troop damage to how much it goes OVER the minimum leadership requirement.

[Zhuangzi]

The AI is pretty dumb. I can understand it just hitting the nearest unit for max damage, but the AI's use of spells is retarded at best. It is also really easy to exploit the AI with phantom/cannon fodder units.

But that's okay. Really excellent AI is probably impossible, and anyway it would make the game harder.

[phoenixreborn]

Killing the soul of Homm 3? OMG there's no behemoths. Heroes is dead!!

[wolfing]

About hypnosis, I don't think the problem is that the spell is too powerful or not, the problem is that the AI shouldn't attack its own hypnotized units, just like you probably wouldn't kill your own hypnotized units (you would dispel it, move away from it until spell ends or something like that)