Jump to content

Realism in AI


aigleborgne

Recommended Posts

Hello,

 

I have carefully read everything about your mod, and I like many things about it.

It was the mod I was looking for :(

 

But it comes too late for me, I decided to create my own mod last december (Enhanced creatures, found on Black Wyrm Lair, not released yet but close to)

 

While our mods are quite similar, they are still different in many ways. I just hope it will be possible to install both mod, with selective components to get the best of them ;)

 

A little note concerning your enhanced AI:

Do you think it is smart to attack "normal" target first (avoiding confused/slept/held...) ?

I was thinking it was before. But it is easier to kill confused/slept/held target first because player have no control over them, and huge bonus on tohit for held/slept targets. So generaly, those targets will drop very quickly (while others can just run away or be far more difficult to hit)

 

Cheers,

 

Arnaud

Link to comment
A little note concerning your enhanced AI:

Do you think it is smart to attack "normal" target first (avoiding confused/slept/held...) ?

I was thinking it was before. But it is easier to kill confused/slept/held target first because player have no control over them, and huge bonus on tohit for held/slept targets. So generaly, those targets will drop very quickly (while others can just run away or be far more difficult to hit)

 

It's an interesting question...

 

Here's my logic. If a group of creatures is thinking "how can we do the most damage possible to the party before being killed ourselves" then it makes a lot of sense to kill off the paralysed, sleeping etc. opponents. Having one or two of your party killed is really, really annoying in BG1, where you can't carry resurrection magic around - lots of people would just reload, in fact.

 

But most monsters aren't thinking, "how can we do the most damage possible to the party before being killed ourselves"? Realistically or otherwise, they're thinking, "how can we kill all of these adventurers?" And if you're aiming for an outright win, once you've killed off all the non-helpless opponents you can finish off the helpless ones very easily.

 

This all assumes that once someone becomes helpless they stay helpless. And obviously that's not reliably true. On the other hand, in BG1 it's relatively difficult to recover someone from many of the standard forms of helplessness (you might be carrying remove paralysis or remove fear, but there's not much you can do about sleep or charm short of dumping a dispel on the entire battlefield... which generally hurts the party as much as or more than the monsters.

 

That's the tactical argument. There's also an argument from realism: one of the main goals of SCS is to minimise the times when some opponent does something unrealistically stupid (hits himself with a fireball, for instance) and spoils suspension of disbelief. Fairly often in the late stages of battles in BG1 and BG2, I've found myself basically having won the day, but not quite managing to kill the last opponent before he kills off my still-sleeping priest. That's hopelessly unrealistic: that last opponent is probably doomed, obviously, but his least-doomed strategy is to fight the people actually still up and threatening him.

 

The remaining argument is more practical: as I said, being killed in BG1 is a real pain. SCS makes the battles a lot more lethal, which is the idea, but I think there's some advantage in the fact that if you win a fight, you're relatively less likely to have to trek all the way back to the Friendly Arm again...

Link to comment

Your arguments are valid and good.

But realism isn't a good factor. We try to improve the game, making it harder. I do it without using any cheats (not even autobuffing mages, well maybe just for long lasting buff like prot. from fire).

When I play through the game, I never reload when I loose one character and yes, it's boring to rez someone, carrying his equiment, going to the inn ....

 

My purpose is to make battles a challenge. My AI often plays the way I do.

When my cleric cast a command spell, I know my target will sleep for exactly one round, and it is an opportunity to attack it with great bonus. Paralyze usually last a long time and I not always attack this target.

 

Player make many choices depending on many parameters and it's obviously impossible to script.

Now suppose I have one fighter and 2 mages. I'm fightning a group of 6 enemies (fighters, mages, clerics...)

I can do this because I'm stronger in levels and wiser (even against a good script).

If their mages manage to hold, paralyze or confuse one of my character, I will probably loose. But they are going after the 2 others, I have 2 solutions:

- Fight, and probably loose

- Run, but make sure they follow me. When my 3rd is free, I can regroup and fight again.

 

One last word: remember that players who install AI mod are already good (they won't install it otherwise). And if they are good, they shouldn't die too often :(

 

In this example, targetting helpless character is better.

 

A little note concerning your enhanced AI:

Do you think it is smart to attack "normal" target first (avoiding confused/slept/held...) ?

I was thinking it was before. But it is easier to kill confused/slept/held target first because player have no control over them, and huge bonus on tohit for held/slept targets. So generaly, those targets will drop very quickly (while others can just run away or be far more difficult to hit)

 

It's an interesting question...

 

Here's my logic. If a group of creatures is thinking "how can we do the most damage possible to the party before being killed ourselves" then it makes a lot of sense to kill off the paralysed, sleeping etc. opponents. Having one or two of your party killed is really, really annoying in BG1, where you can't carry resurrection magic around - lots of people would just reload, in fact.

 

But most monsters aren't thinking, "how can we do the most damage possible to the party before being killed ourselves"? Realistically or otherwise, they're thinking, "how can we kill all of these adventurers?" And if you're aiming for an outright win, once you've killed off all the non-helpless opponents you can finish off the helpless ones very easily.

 

This all assumes that once someone becomes helpless they stay helpless. And obviously that's not reliably true. On the other hand, in BG1 it's relatively difficult to recover someone from many of the standard forms of helplessness (you might be carrying remove paralysis or remove fear, but there's not much you can do about sleep or charm short of dumping a dispel on the entire battlefield... which generally hurts the party as much as or more than the monsters.

 

That's the tactical argument. There's also an argument from realism: one of the main goals of SCS is to minimise the times when some opponent does something unrealistically stupid (hits himself with a fireball, for instance) and spoils suspension of disbelief. Fairly often in the late stages of battles in BG1 and BG2, I've found myself basically having won the day, but not quite managing to kill the last opponent before he kills off my still-sleeping priest. That's hopelessly unrealistic: that last opponent is probably doomed, obviously, but his least-doomed strategy is to fight the people actually still up and threatening him.

 

The remaining argument is more practical: as I said, being killed in BG1 is a real pain. SCS makes the battles a lot more lethal, which is the idea, but I think there's some advantage in the fact that if you win a fight, you're relatively less likely to have to trek all the way back to the Friendly Arm again...

Link to comment

I don't agree with the notion that AI enhancements should purely be intended as a difficulty increase. I'd place an immersive game experience above engineered challenges in terms of importance--in many cases the former *will* result in tougher enemies purely because the original AI is so awful, but I don't support aiming for everything to be as challenging as possible at the expense of believability.

 

For a specific example, take zombies. Under your proposal that AI should operate the way the player does and be intended primarily to make the game harder, we can easily assign zombies the ability to prioritize targetting of enemies, drink healing potions, shout to one another--any number of things. None of these make much sense; they're zombies. I don't think changes which are conceptually inappropriate enhance the game.

Link to comment

I agree with you. My own AI is assigned to intelligent people (beeing monsters like death knight, or humans/elves/...)

Non intelligent creatures will get an improved script, but just for some abilities like tracking their target...

 

As I said, I'm just trying to make a good AI, and a good AI is always wiser and harder to beat.

 

I don't agree with the notion that AI enhancements should purely be intended as a difficulty increase. I'd place an immersive game experience above engineered challenges in terms of importance--in many cases the former *will* result in tougher enemies purely because the original AI is so awful, but I don't support aiming for everything to be as challenging as possible at the expense of believability.

 

For a specific example, take zombies. Under your proposal that AI should operate the way the player does and be intended primarily to make the game harder, we can easily assign zombies the ability to prioritize targetting of enemies, drink healing potions, shout to one another--any number of things. None of these make much sense; they're zombies. I don't think changes which are conceptually inappropriate enhance the game.

Link to comment
I don't agree with the notion that AI enhancements should purely be intended as a difficulty increase. I'd place an immersive game experience above engineered challenges in terms of importance--in many cases the former *will* result in tougher enemies purely because the original AI is so awful, but I don't support aiming for everything to be as challenging as possible at the expense of believability.

 

I'd also agree. The philosophy of SCS was to make creatures perform as intelligently as possible given their own intelligence (which, for any creature with INT>4, means "as intelligently as possible full stop", since the IE scripting language is so limited). Beyond that level, if I wanted to make things more difficult I did it by adding more levels (to humanoid opponents) or sometimes more opponents.

 

But there's a caveat (and maybe this is where Sim and I disagree with Arnaud?) I prioritised "never doing something immersion-breakingly-stupid" over "doing the brightest thing as often as possible". For instance, SCS is very cautious about using Fireball and Lightning Bolt (it pretty much only uses it when a mage is alone and protected by MGI). That's often a tactically suboptimal choice; it's the price paid for monsters not electrocuting their own allies, which spoils immersion for me.

 

So in the case of the paralysis, while it's impossible (as Arnaud points out) to predict the best choice, it's possible to predict that avoiding the helpless is often the best choice and almost never an impossibly stupid choice. Even in the case of pursuing fleeing opponents, it's perfectly plausible that one might do this in the heat of battle, rather than coldly sitting back and slitting people's throats.

 

[edit: I forgot about Command; clearly you're right there. SCS was intended to have a loophole for that, but doesn't... I'll add it to the buglist.]

 

In addition to this, after a certain point exploits like getting monsters to run round and round after you counts as cheese. I had a deliberate design policy of not working too hard to block this sort of thing, because ultimately IE is stupider than you are, so if you want to cheese it you're going to succeed. I have absolutely no objection to people who enjoy cheesing battles, but it's not my chosen style!

 

I also don't fully agree with this:

One last word: remember that players who install AI mod are already good (they won't install it otherwise). And if they are good, they shouldn't die too often

 

This is a function of the difficulty of the encounter as much as the talent of the party. If you install a mod which replaces Tarnesh with ten glabrezus, I predict that you'll have a high casualty rate even if you are an ace tactician...

Link to comment

If I am in a real fight in real life and an enemy falls asleep I wont attack him while there are other enemys around that can actually do me harm. Its, IMHO, common sense to attack the ones that can still do you harm and leave the ones that cant till later.

 

My point of view.

 

ronin

Link to comment

Archived

This topic is now archived and is closed to further replies.

×
×
  • Create New...