FE 0.80: AI Test Game

from JoeUser Forums

Version 0.80 is an internal build of Fallen Enchantress.  This is basically a test of the AI on normal.

I’m playing as Relias, the sovereign of Altar, a Kingdom of Men.

My opponents are the Blood Elves of Magnar, The Mancers of Capitar, and the Wraiths led by Ceresa.

Note: I’m playing in the debugger so please bear that in mind before commenting too much on graphics as these are set to minimum.

An early battle:

This is the new tactical AI in action.  I’m not sure I like what it decided.  The wolves decided to bypass my strongest unit in order to target my weaker champion. But in doing so they were vulnerable to my main guy.

So let’s see what we can do next…

54,313 views 59 replies
Reply #1 from Elemental Forums Top


unless the 'weaker' guy has an injury or is damaged, would the wolves be intellegent enough to go after him?

Reply #2 from Elemental Forums Top

Screw the wolves. Think of it like you would expect a player to. That's the only way you'll ever approximate a challenging AI.

PS. The battlefields are way too small to worry about "leaving yourself open to attack". Deliver the most imaginable punch you can, as fast as you can. You'll never outmanoeuver a player.

PS2. To determine a target, you need to evaluate three variables: 1) is the unit dangerous to me? 2) is the unit valuable to the enemy?  3) how weak is the enemy to my attacks (ie - how fast can I kill it)

If the unit is dangerous to you, you'll need to remove it quickly. If the unit is valuable to the enemy, you gain a lot by killing it. If the enemy is particularly weak, ie easy to kill, you want to kill it even if it isn't dangerous or valuable.

Ex 1: A weak caster champion. High value, high danger, very weak. Kill it.

Ex 2: A path of the defender champion. High value, medium danger, very strong.

Ex 3: A path of the warrior champion. High value, high danger, pretty weak. 

Kill order: Ex1, Ex3, Ex2.

SoundPackager - Customize the Windows sound scheme
Reply #3 from Elemental Forums Top
Agree with Heavenfall. That's pretty much how I decide to fight my battles, by killing off the units that fall fast and provide most benefit to the adversary. Since there's no terrain and the battlefield is so fast to traverse, that's just about the only consideration in the battles.
Reply #4 from Elemental Forums Top

Quoting Heavenfall, reply 2
PS2. To determine a target, you need to evaluate three variables: 1) is the unit dangerous to me? 2) is the unit valuable to the enemy?  3) how weak is the enemy to my attacks (ie - how fast can I kill it)

Good point. But I think it's limited to intelligent units. Animals don't take into account unit's value to the enemy. I can only imagine that a wolf would attack the leader of enemy's "pack" (which is not always the strongest unit) to break the "pack".

Reply #5 from Stardock Forums Top

Here's a video of another battle:

 

http://www.screencast.com/t/J0mW8NyLRddm

Reply #6 from Elemental Forums Top

Looking better already...

 

A few comments though:

For the first battle, shouldn't wolves stopping next to your front champion try to land a hit on him so as to not waste an attack? (obviously this doesn't hold true if he has a counter attack ability).

 

For the second battle, shouldn't the archer try to get further away (unless it only has two moves)?

Reply #7 from Stardock Forums Top

Yep. Just did that actually (for first one).

I'm working on the second problem now. 

Reply #8 from Elemental Forums Top

Quoting Ihfacny, reply 6
For the second battle, shouldn't the archer try to get further away (unless it only has two moves)?

No. He would draw his enemy too far away from the second bandit. Am I right, is it intentional?

EDIT: 

Quoting Frogboy, reply 7
I'm working on the second problem now. 

OK, so it's not intentional.

Reply #9 from Elemental Forums Top

Gandhialf, you might be right, but the second bandit was still out of range after the archer moved back one square.

I suggest the archer should either move further away - or if the Archer can't get fully out of range anyway "retreat" around and forward to get still get some (1?) space to the champion and lure the Champion within range of the bandit.

Reply #10 from Elemental Forums Top

Except that it would be irrelevant in this case because the melee bandit did not get a chance to attack until after the archer died. 

Reply #11 from Elemental Forums Top

I think the wolves made the right call. They had no chance of killing your stronger character and no chance of winning the battle overall. Concentrating on the weakest foe to inflict max damage before losing is the right move. I think it would be nice if the AI made it so that unintelligent animals just attack the nearest threat, but wolves are pack predators that work by separating the weakest prey from the pack, so I think it was realistic here.

 

Reply #12 from Elemental Forums Top

I would like to suggest the importance of using tacitical OUTPUT as a key component to the tactical AI.

What I'm thinking of here, specifically, is a concept know as 'aggro' which I encountered in my brief MMO playing career.

Basically, this would take the output of what is currently occurring inside a tactical situation and cause the AI to strategically adapt to the circumstances based on the interactions of the strengths and weaknesses - or simply if some units are getting 'lucky' - of both parties.  Then, AI units would gain a certain amount of aggression towards a unit which could force a shift in focus during the course of a tactical battle.

This aggro rating could then be modified trough traits, spells, items etc to change the landscape of the battle.

This may already exist, or it may be impossible, or maybe it's not significantly different from what's been suggested already.  Just throwing it out there.

Reply #13 from Elemental Forums Top

Quoting mqpiffle, reply 12
AI units would gain a certain amount of aggression towards a unit which could force a shift in focus during the course of a tactical battle.

Do we really want an AI that is shifting aggro?  Do you shift aggro when attacking an enemy?  I bet you don't.  For beasts, it is an interesting thought, but for an AI that is having difficulty being challenging as it is, I say just stick to trying to kill me as efficiently as possible.

 

In the wolf fight, the wolves missed 3 chances to attack, 2 on Relias while trying to bypass him, and 1 on the back guy after moving into range.  Not that they stood a chance against Relias who one-shots all 3 of them.  They still could have done a bit more damage though.

Reply #14 from Stardock Forums Top

Here's another video test:

http://screencast.com/t/L5xcSXCZumUY

You probably recognize the monster group here.

Reply #15 from Stardock Forums Top

Here's another, what you didn't see in here was the AI cast focus on himself and then smack down my champion.

http://screencast.com/t/43xZZKj8HyQ

 

Reply #16 from JoeUser Forums Top

And yet another one:

http://screencast.com/t/oRNyg67tVJf

Notice that my units didn't just go charging at them -- they were backed up by archers so they protected the archers.

Reply #17 from Stardock Forums Top

Here's one where my side lost but one thing to notice is that they focus fire - when it makes sense (lots of calculations there on when it makes sense to focus on a particular unit versus overkill).

http://screencast.com/t/VfN1lRsg36Q

 

Reply #18 from Elemental Forums Top

Wow Vearwynn can't catch a break... Perhaps the AI, players anyway, should protect it's weak or damaged units more. Otherwise the  Ai is going to rack up a lot of injuries on it's weaker champions.

Reply #19 from Stardock Forums Top

Well, ideally, the system will eventually let me set up the PLACEMENT of units on the map at the start.

Here's another one:

http://screencast.com/t/fnP7JLAk

 

Reply #20 from Stardock Forums Top

Here's one with a TON of AI errors in it:

http://screencast.com/t/wrEdUDZj

This one I'll be going over a lot.

Reply #21 from Elemental Forums Top

That last one made me laugh. What was the AI thinking?

Reply #22 from Elemental Forums Top

Quoting Frogboy, reply 5
Here's a video of another battle:

 

http://www.screencast.com/t/J0mW8NyLRddm[/quote]

 

OK, the last bit of that battle, why did that bandit move to where the other guy could move in and attack first, he should keep his distance.

Reply #23 from Elemental Forums Top

Quoting mqpiffle, reply 12
I would like to suggest the importance of using tacitical OUTPUT as a key component to the tactical AI.

What I'm thinking of here, specifically, is a concept know as 'aggro' which I encountered in my brief MMO playing career.

Basically, this would take the output of what is currently occurring inside a tactical situation and cause the AI to strategically adapt to the circumstances based on the interactions of the strengths and weaknesses - or simply if some units are getting 'lucky' - of both parties.  Then, AI units would gain a certain amount of aggression towards a unit which could force a shift in focus during the course of a tactical battle.

This aggro rating could then be modified trough traits, spells, items etc to change the landscape of the battle.

This may already exist, or it may be impossible, or maybe it's not significantly different from what's been suggested already.  Just throwing it out there.

I've done extensive raid leading... Aggro is a way to get things to attack you without doing much damage. It protects your weaker units. No sane enemy would focus on the taunting fool in his face when someone else with a bow was pummeling with arrows. Ever. Its a game mechanic, pure and simple, that does not translate well to battlefields. If you were fighting me and you would send a tank in my face, I'd send something to keep it occupied and then smash the healer, CC and DPS.

Reply #24 from Elemental Forums Top

I do think simpler AI is appropiate for animals and stupid monsters, even on harder levels.

 

You're fighting wolves, not Wile E. Coyote and his family.

 

 

 

 

Reply #25 from Elemental Forums Top

Yea that last one ... it started out nice, but imho I think that Champion was just a bit too afraid of engaging in melee. He could have at least debuffed the ogres while the Archers wore em down. Then engaged in melee  once the ogres were "at the gates" so to speak.

 

Archers 'running away' is nice, as it did eventually keep 1 unit of archers alive, but it seems like they chose 'not' to attack on multiple turns, which doesn't make much sense imho.