Kiting enemies or "How not to do AI"

[Tamerlane]

Yea i get what you're saying and how it could apply to the game. I guess I wasn't really factoring in that you were considering all party members to be lightning fast or otherwise unreachable.

"Lightning fast" doesn't really enter in to it. With two characters of equal speed, one catching the other is a ****ing nightmare. Especially if attack animations require standing still and there's no way to knockdown (though knockdowns are A Thing in PE, at least) or apply a speed debuff.

[Nerei]

But really, kiting is the really the lack of AI, because 'intelligent' things quickly realize they aren't going anywhere by letting someone kite them. In any online game where you can kite intelligent human players, they tend to take cover and force a situation, retreat, or simply try to preserve their life against an unbeatable foe. All of these responses should be programmed into a game where kiting might be a possible strategy. At the very least, you would hope enemies would have a "survival instinct" of some kind which would imply their animal-intelligence. Not that I personally like it when all my foes cower and run in fear when I fight them - that's usually boring - but they should be more intelligent than an animal with rabies (who have suffered the degeneration of their ****ing brain).

 

I'd argue that kiting is actually the end result of RT implementations. The AI has to compete with player input, game state checks, some degree of sound/graphics processing* for a slice of each second. Each unit representing it's own AI.

 

AI requires complicated algorithms to achieve good results, complicated algorithms require time and horsepower. So the AI in any RT system is going to end up with deficiencies simply because it's severely resource limited. Tossing some more processors at it helps, but ultimately, you end up bound by the amount of processing that can occur in around 1 second. Which isn't nearly as much as people think it is with AI, since AI generally consists of NP-complete problems.

 

*While GPU's and hardware sound chips handle most of the processing on their own, the CPU still needs to process the triggers for the graphics at the bare minimum and depending on the implementation may process part of the job as well, older GPU's and on-board sound often require CPU resources during processing.

 

So what you are saying is that the computers we have today cannot handle the AI effectively in realtime due to a lack of basic processing power?!

If that was the case I really wonder how the fossil computers that ran games like Dune or C&C (or Baldurs gate for that matter) could handle anything then. The laptop I am typing this on has a CPU that at the basic level is something like 100 times stronger than what I originally ran those games on!

In many cases when I run RTS games on my desktop they do not even all my system resources. Claiming that AI deficiencies is then down to my CPU not having the power to run those scripts effectively seems a bit strange.

 

http://en.wikipedia....ions_per_second

Yes I know MIPS is not an entirely accurate measurement of the power of the computer, but it really does paint an interesting picture of the increase in processor power over the years does it not?

 

 

The limitation for the AI is in the programming, not in any way related to the actual power of the CPU (it might have been at some point though, like back in the 1990's). Making an AI that can do as complex things as a skilled human being is technically just about impossible. That is the limitation.

 

 

Personally I do not expect the AI to be great mainly due to it not really being possible to make an AI that can use 100 different abilities or tactics effectively in a multitude of situations. No scripting the events does not count.

[Hormalakh]

Personally I do not expect the AI to be great mainly due to it not really being possible to make an AI that can use 100 different abilities or tactics effectively in a multitude of situations. No scripting the events does not count.

 

The AI might not be "human" level, but we can do a little better than "IF (see X enemy) THEN (attack X enemy). " Throwing in a few more possible actions would be nice. Allowing enemies to communicate with each other can do a lot to make combat more interesting.

[Umberlin]

It's true you can do more with AI than we typically see in games, as a general, and not just RPGs, as shown by the very few and rare games that have exceptional and challenging AI. Better challenging AI than things that add tedium, as per increasing as enemies damage/health pool, as if that's in any way challenging and not just a more tedious version of the same thing all over again.

[BetrayTheWorld]

Could someone define kiting?

 

Doing damage to enemy while staying away from his attack range.

 

Yup, this is kiting in a nutshell. But anyone who read the thread knows that the OP meant "pulling". Either way, both tactics are silly when dealing with an intelligent enemy. The only way kiting should work is against a beast with sub-par intelligence, and even in that case, I'd think said beast would realize that they weren't catching up to you, and were only being hurt more, so they'd likely run and hide. All natural beasts have the "fight or flight" system of behavior. And even when dealing with apex predators, nature has given them a strong sense of self-preservation. It may not be their first instinct to run, but it's still an instinct in the right circumstances.

 

Point is, kiting shouldn't work very well.

 

Pulling, on the other hand, should only work if done through circumstances that are believeable. If you cause a ruckus attacking someone, they should shout out for help. If you pull them away from their group by making a small sound that they go to investigate, then quickly dispatch them within the time-frame of a round, this sort of pulling is acceptable and realistic.

[Nerei]

Personally I do not expect the AI to be great mainly due to it not really being possible to make an AI that can use 100 different abilities or tactics effectively in a multitude of situations. No scripting the events does not count.

 

The AI might not be "human" level, but we can do a little better than "IF (see X enemy) THEN (attack X enemy). " Throwing in a few more possible actions would be nice. Allowing enemies to communicate with each other can do a lot to make combat more interesting.

 

My comment about the great AI is really more directed at the talk of kiting (and other more complex situations the AI might have to handle). You can add a lot of fancy mechanics, but if the AI fails to handle them properly (which it likely will) it is just more likely to fail completely.

Take the IE games, they actually had a morale check. However it usually just resulted it enemies running a bit away and then standing idly waiting to be put to the sword. For complete hilarity you could have the lone kobold end his morale failure check and charge the 6 bloodstained adventurers that had just butchered his 20 friends.

 

The IE games have fairly lousy AI, improving on it is certainly possible and naturally should be done to the best of Obsidians abilities.

I agree that the tactic of pulling 1 enemy away from the group of 3 and killing it alone like you could in BG was silly, cheap and borderline an exploit (and as mentioned fairly easy to solve given that modders could do it). You can add all the cheap ways of killing Drizzt to the list of examples of bad AI exploits. However changing that does not make the AI a tactical genius. It pretty much just means it has advanced past the "too stupid to live" point.

 

Making the AI handle things like kiting properly (or any other complex strategy for that matter) is something else completely, it is far more complex that pulling.

If the AI cannot handle the situations presented to it properly I would never label it as better than "adequate". That is really why I say I do not expect it to be great (and why I in general never really expect complex games to have a great AI).

[Hormalakh]

My comment about the great AI is really more directed at the talk of kiting (and other more complex situations the AI might have to handle). You can add a lot of fancy mechanics, but if the AI fails to handle them properly (which it likely will) it is just more likely to fail completely.

Take the IE games, they actually had a morale check. However it usually just resulted it enemies running a bit away and then standing idly waiting to be put to the sword. For complete hilarity you could have the lone kobold end his morale failure check and charge the 6 bloodstained adventurers that had just butchered his 20 friends.

 

The IE games have fairly lousy AI, improving on it is certainly possible and naturally should be done to the best of Obsidians abilities.

I agree that the tactic of pulling 1 enemy away from the group of 3 and killing it alone like you could in BG was silly, cheap and borderline an exploit (and as mentioned fairly easy to solve given that modders could do it). You can add all the cheap ways of killing Drizzt to the list of examples of bad AI exploits. However changing that does not make the AI a tactical genius. It pretty much just means it has advanced past the "too stupid to live" point.

 

Making the AI handle things like kiting properly (or any other complex strategy for that matter) is something else completely, it is far more complex that pulling.

If the AI cannot handle the situations presented to it properly I would never label it as better than "adequate". That is really why I say I do not expect it to be great (and why I in general never really expect complex games to have a great AI).

 

You know, I'd agree with you if I hadn't seen mods to these IE games make the AI incredibly better. All these tactics mods to the Baldur's Gate games tells me that it's possible to be done. Perhaps Black Isle/Obsidian thought that we couldn't handle better AI back in the day? Well since this is a niche market, I'd hope that they work on a stronger AI this time.

 

The problem with a lot of these scripts is that they were mods and so not built directly into the engine. That probably made them so much less efficient coding. I really hope that OEI takes a look at this once again and considers whether any changes would occur. At the end of the day the devs know more about this than we do.

Edited November 15, 2012 by Hormalakh

[Tethros]

I think that it's entirely possible to create a balanced AI system where this strategy works against some enemies and not against others. Highly social or hive minded creatures should react to you attacking one of their group, whereas solitary creatures might ignore it or even just run away.

 

After all, as "cheap" as it seems, guerilla warfare is a legitimate tactic and, ever increasingly, the norm in real-world combat.

[wanderon]

As far as kiting is concerned it essentially only works against a foe that is not faster moving than the party members and has no ranged attack in their arsenal - ( and that the party has enough ranged firepower to take it down)

 

Should the game then be designed so that all enemies either have both a ranged and melee attack or else are always faster than a (hasted?) party member just to take kiting off the table?

 

I would vote no on that.

[Sacred_Path]

Should the game then be designed so that all enemies either have both a ranged and melee attack or else are always faster than a (hasted?) party member just to take kiting off the table?

 

I would vote no on that.

 

I would say YES PLZ with cherries on top.

 

Why should your party members be the only beings clever enough to pack both melee and ranged weapons?

[Tamerlane]

As far as kiting is concerned it essentially only works against a foe that is not faster moving than the party members and has no ranged attack in their arsenal - ( and that the party has enough ranged firepower to take it down)

 

Should the game then be designed so that all enemies either have both a ranged and melee attack or else are always faster than a (hasted?) party member just to take kiting off the table?

 

I would vote no on that.

Well, no, of course not.

 

But neither should we see every damn creature with the same movement rate.

 

And I still want to see knockdowns and speed debuffs aplenty.

[Hormalakh]

Some of it relates to what the enemy AI is programmed to react to (and how) and I'm not seeing where discovering this and then designing tactics to exploit a weakness is a bad thing.

 

IIRC NWN2 had many enemies reacting by attacking any foe that caused them damage. This lead to scenarios where you could play tag with one by having two different archers attacking from different sides - each time an arrow would hit the enemy it would turn and run towards the archer that hit it but before he got there he would be hit from the other side and change directions and go for the other one until he died without ever reaching either one.

 

Your party members often did the same thing IIRC unless you took control of them and forced them to attack a single foe.

 

Now I'm no programmer but it seems to me that no matter what sort of AI gets programmed for the enemy that as a functional thinking human (well on a good day maybe) I am eventually going to see that enemy A seems to always act in a certain way under certain circumstances which means he may have a weakness to tactic B or C.

 

The question is when I figure this out am I abusing the AI or just paying attention and forming good tactics to counter a weakness I have seen in the way this enemy operates?

 

The AI is never going to be so robust that you can't find a solution but isn't that the whole point of the game - to find solutions to defeat your enemies in order to stay alive and complete the story?

then you make a "dynamic" AI, that has some sort of timer. If he doesn't reach an enemy in like 10 seconds, he switches strategies. He starts just following one person. I mean at the point where you've got 6 people on 1 guy, you can do all sorts of tactics to kill him. He's dead anyway. He might just run away which could be another strategy.

Edited November 15, 2012 by Hormalakh

[Hormalakh]

As far as kiting is concerned it essentially only works against a foe that is not faster moving than the party members and has no ranged attack in their arsenal - ( and that the party has enough ranged firepower to take it down)

 

Should the game then be designed so that all enemies either have both a ranged and melee attack or else are always faster than a (hasted?) party member just to take kiting off the table?

 

I would vote no on that.

I would also vote no. But then you should also worry about fast creatures like we worry about creatures with range. It would be a nice mix up of AI that makes the game fun. We can always think of goblins as weak cratures not because they're low-level, but because their AI sucks. Then hobgoblins are both higher-level and more tactical. that would be fun! Mix it up!

[The Guilty Party]

I dislike Kiting and Pulling. They feel awkward and game-y and are generally tedious.

 

I'm okay with reasonable dividing-their-forces behavior. If there's some patrolling guard, making some little noise to grab him around the corner and then bashing his head in: good planning. Inching forward on the map to reveal one monster in a group of monsters so that only it responds when you shoot it? Lame.

[Ravenfeeder]

I find pulling to be a bit cheesy, but still find myself doing it sometimes. I'd be very happy with enemies in earshot running to join a fight, even opening their own doors to do so (unless locked - which a sneaking rogue could do).

 

I loathe and despise kiting and hope it is not a viable tactic. In DA2 the best antagonist had the worst end as you spent half an hour kiting round him. Awful, awful gameplay, the designers of that scene should hang their heads in shame.

[Jarmo]

Then again, when mongols pulled the western knights away from the infantry and then just retreated ahead all the while peppering the knights with arrows, the knights probably felt it's all very unfair, unsportsmanlike and gamey. Especially when the mongols would then much later when the knights were dead tired, turn around and make them completely dead.

 

Then proceeding to kill the infantry as well. Also very unfairly as the infantry had few ranged weapons and no way to reach the horsemen.

Annoying strategies, but usable on occasion. If the enemy is slower and has no ranged weapons, there's no reason to not use the easy way out.

 

If an arrow from a distance pulls the worgs in first and the slower orcs follow behind, should be workable to first deal with one group and the rest later.

As long as not everybody goes the exact same speed and the enemies are not blind to their friends getting hit 4 feet away, it should all work out.

 

And most intelligent enemies should have some kind of ranged capability. If only a couple of darts.... or ..rocks picked from the ground? Would that be possible.

[Hormalakh]

Yeah I don't have a problem with kiting or pulling as the exception. But not as the rule. Even the mongols were successful because this wasn't the rule in combat and because they have vast hordes of horsemen. Kiting isn't a cheese tactic in and of itself. It can just become a cheese tactic when it works in every scenario.

[Gatt9]

But really, kiting is the really the lack of AI, because 'intelligent' things quickly realize they aren't going anywhere by letting someone kite them. In any online game where you can kite intelligent human players, they tend to take cover and force a situation, retreat, or simply try to preserve their life against an unbeatable foe. All of these responses should be programmed into a game where kiting might be a possible strategy. At the very least, you would hope enemies would have a "survival instinct" of some kind which would imply their animal-intelligence. Not that I personally like it when all my foes cower and run in fear when I fight them - that's usually boring - but they should be more intelligent than an animal with rabies (who have suffered the degeneration of their ****ing brain).

 

I'd argue that kiting is actually the end result of RT implementations. The AI has to compete with player input, game state checks, some degree of sound/graphics processing* for a slice of each second. Each unit representing it's own AI.

 

AI requires complicated algorithms to achieve good results, complicated algorithms require time and horsepower. So the AI in any RT system is going to end up with deficiencies simply because it's severely resource limited. Tossing some more processors at it helps, but ultimately, you end up bound by the amount of processing that can occur in around 1 second. Which isn't nearly as much as people think it is with AI, since AI generally consists of NP-complete problems.

 

*While GPU's and hardware sound chips handle most of the processing on their own, the CPU still needs to process the triggers for the graphics at the bare minimum and depending on the implementation may process part of the job as well, older GPU's and on-board sound often require CPU resources during processing.

 

So what you are saying is that the computers we have today cannot handle the AI effectively in realtime due to a lack of basic processing power?!

If that was the case I really wonder how the fossil computers that ran games like Dune or C&C (or Baldurs gate for that matter) could handle anything then. The laptop I am typing this on has a CPU that at the basic level is something like 100 times stronger than what I originally ran those games on!

In many cases when I run RTS games on my desktop they do not even all my system resources. Claiming that AI deficiencies is then down to my CPU not having the power to run those scripts effectively seems a bit strange.

 

http://en.wikipedia....ions_per_second

Yes I know MIPS is not an entirely accurate measurement of the power of the computer, but it really does paint an interesting picture of the increase in processor power over the years does it not?

 

 

The limitation for the AI is in the programming, not in any way related to the actual power of the CPU (it might have been at some point though, like back in the 1990's). Making an AI that can do as complex things as a skilled human being is technically just about impossible. That is the limitation.

 

 

Personally I do not expect the AI to be great mainly due to it not really being possible to make an AI that can use 100 different abilities or tactics effectively in a multitude of situations. No scripting the events does not count.

 

The limitation for AI is inherent in the architecture not in the programming.

 

http://en.wikipedia.org/wiki/NP-complete

 

Anything you're going to want to do is going to fall prey to this issue in a RT system, because it's going to have to try and handle NP Complete problems in RT, or problems approaching NP Complete.

 

Shortest path is a prime example, it's NP Complete, and it's something any AI is going to need to handle.

 

http://en.wikipedia.org/wiki/Shortest_path_problem

 

Further, lets say you have two enemies (A and B) attacking two characters (C and D), you could have AC & BD, BC & AD, ABC & D, ABD & C. 4 possibilties.

 

Lets give you two enemies (A and B) attacking three characters (C, D, and E). You could have AC & BD & E, BC & AD & E, C & BD & AE, AC & D & BE, ..., it's a combinatrics problem. Which means that the number of possiblities just for attacking scales upwards at an extremely rapid pace as number of combatants increase, for just a simple attack. That doesn't even try to put any logic into it like assessing what the best combinantion is, or to try and assess the right time to use a special ability, or any other logic. If you give each enemy a bow and a sword, double the possiblities. Give them a special ability too and you'll triple the possibilities, and that's still without even trying to determine which is the most appropriate. It's not inconceivable to end up with AI potentially having hundreds of combinations it has to evaluate in less than a second, with each of those combinations introducing even more variables that have to be evaluated (Like if it has an AOE special ability and must then evaluate the position of all of it's allies and their likely future positions).

 

All of this is competing with the game handling the math in the background, any processing that must occur, etc, and it all must occur in a tiny enough timeslice such that the end user cannot detect it. This is all simply brutal on a computer, and it's why a RT game tends to lack any significant degree of tactics.

 

That's before we even discuss the processor cycles lost in the background to OS duties, background tasks running in the task bar, etc.

 

It's quite possible to develop incredible AI's that'll stomp humans, but it's not going to run in a few hundred milliseconds. Just processing shortest path is going to eat all of that time (And still not finish), which is why most games don't try. They instead try to move in a direct line to the Player, which yields kiting and snagging on environment.

[Continuum]

PLEASE do not make kiting or pulling MMO-style an essential (or particularly useful) mechanic. It's one of the worst things about MMORPGs - the toughest challenge in most MMO fights is not actually beating the enemy but rather in getting a small enough group of enemies far enough away from the others to push the same buttons 30 times to beat them.

 

Space enemies out appropriately, and design it so that they all react appropriately when they can realistically see and hear your party.

 

And yes, the IE games had some awful, awful, awful pathing. Pathing is probably the #1 reason that I often prefer turn-based control of a multi-character party. Even Dragon Age drove me nuts, as you either have to turn off friendly fire entirely or micromanage to a degree that it might as well be turn-based anyway if you want to use AoEs at all.

[dropko]

add enemy behaviour. not all enemies need to be "groupies".

some individual creatures, even when grouped-in-waiting, could be kited if they don't have herd instincts (or are stupid animals, like chikens). humanoids should always "call out" others. if not, others should "hear" them, unless some are mute and others are deaf, for example (either by nature or by spell).

[Hormalakh]

The limitation for AI is inherent in the architecture not in the programming.

 

http://en.wikipedia....iki/NP-complete

 

Anything you're going to want to do is going to fall prey to this issue in a RT system, because it's going to have to try and handle NP Complete problems in RT, or problems approaching NP Complete.

 

Shortest path is a prime example, it's NP Complete, and it's something any AI is going to need to handle.

 

http://en.wikipedia....st_path_problem

 

Further, lets say you have two enemies (A and B) attacking two characters (C and D), you could have AC & BD, BC & AD, ABC & D, ABD & C. 4 possibilties.

 

Lets give you two enemies (A and B) attacking three characters (C, D, and E). You could have AC & BD & E, BC & AD & E, C & BD & AE, AC & D & BE, ..., it's a combinatrics problem. Which means that the number of possiblities just for attacking scales upwards at an extremely rapid pace as number of combatants increase, for just a simple attack. That doesn't even try to put any logic into it like assessing what the best combinantion is, or to try and assess the right time to use a special ability, or any other logic. If you give each enemy a bow and a sword, double the possiblities. Give them a special ability too and you'll triple the possibilities, and that's still without even trying to determine which is the most appropriate. It's not inconceivable to end up with AI potentially having hundreds of combinations it has to evaluate in less than a second, with each of those combinations introducing even more variables that have to be evaluated (Like if it has an AOE special ability and must then evaluate the position of all of it's allies and their likely future positions).

 

All of this is competing with the game handling the math in the background, any processing that must occur, etc, and it all must occur in a tiny enough timeslice such that the end user cannot detect it. This is all simply brutal on a computer, and it's why a RT game tends to lack any significant degree of tactics.

 

That's before we even discuss the processor cycles lost in the background to OS duties, background tasks running in the task bar, etc.

 

It's quite possible to develop incredible AI's that'll stomp humans, but it's not going to run in a few hundred milliseconds. Just processing shortest path is going to eat all of that time (And still not finish), which is why most games don't try. They instead try to move in a direct line to the Player, which yields kiting and snagging on environment.

 

Did you actually read the links you sent out? Or are you a CS major in college trying to teach some of us who have degrees in these fields about NP complete problems? Your argument is void. NP-complete problems are not inherent to architectural limitations, but programming algorithm limitations. In many cases, these problems are still approximated and solved using "tricks of the trade." Go back and re-read your wikipedia article about common misconceptions.

 

---------------------

 

Having a "human-like AI" is not what I'm asking for. I'm asking for varied AI for different enemies that takes into account their creature habits (hordes should generally work together, no "pulling" of intelligent enemies, individual enemies should be smart enough to run away if being kited by 6 characters, etc). Your implementation of AI is only one of many different possible solutions. For a completely robust AI, yes it might be limited by architecture. But I don't want a completely intelligent AI, I want one that doesn't always fail to the same problem.

 

This shouldn't take up all the computer's processing power.

[mstark]

I believe the IE games never had any morale checks for going up against a larger group of foes? A thing or two could be taken from Warhammer here, which has very robust group morale systems. Almost to the point of annoyance, since the entire game almost becomes about failing morale checks, fleeing & regrouping.

 

It wouldn't have to be taken to that extreme, but such a system would certainly mean that unless a character has extraordinarily low Intelligence, or extremely high Leadership (though I suppose there won't be such a stat in PE), they would pretty much be guaranteed to fail morale checks when attempting to charge a superior enemy and end up scrambling away in a random direction (or less random, if there are friends nearby who will add to their morale check rolls).

 

A failed morale check would have a similar effect to fear or confuse spells.

 

Edit,

Come to think of it, I would love the PE devs to dig through the Warhammer and Warhammed 40k rule books and take some inspiration for their group vs. group mechanics . Can't think of a better worked out system for this.

Edited November 16, 2012 by mstark

[wanderon]

Should the game then be designed so that all enemies either have both a ranged and melee attack or else are always faster than a (hasted?) party member just to take kiting off the table?

 

I would vote no on that.

 

I would say YES PLZ with cherries on top.

 

Why should your party members be the only beings clever enough to pack both melee and ranged weapons?

 

Because not every enemy you typically meet in an RPG IS clever and/or fast and I don't see why they should be made to be more clever or fast just becuase some people think kiting should be taken entirely off the table. Thats overkill and totally unnecessary.

 

People may also wish to consider that just because something is annoying in an MMO doesn't mean it should be removed from a single player game where only the PLAYER needs to make the decision about what tactics he chooses to use and it has no impact on anyone else...

Edited November 16, 2012 by wanderon

[Nerei]

You know, I'd agree with you if I hadn't seen mods to these IE games make the AI incredibly better. All these tactics mods to the Baldur's Gate games tells me that it's possible to be done. Perhaps Black Isle/Obsidian thought that we couldn't handle better AI back in the day? Well since this is a niche market, I'd hope that they work on a stronger AI this time.

 

The problem with a lot of these scripts is that they were mods and so not built directly into the engine. That probably made them so much less efficient coding. I really hope that OEI takes a look at this once again and considers whether any changes would occur. At the end of the day the devs know more about this than we do.

 

To be fair most of the AI/tactics mods include quite a bit more than just AI updates, they also change and/or add abilities to the mobs. I remember some of them that among other things also said they “improved the AI” also changed the enemies to the point where the challenge most definitely did not just come from the new AI. With a few it was just brute force in the end that made them hard.

 

Naturally that is not to say there was not some fairly good updates among them, I remember a beholder mod that made them play a lot better, for one they actually tried to disable the casters instead of wasting it on the fighter.

It was still a "glory or death" charge which will always irritate me, but I guess I can live with that if they just fight effectively while they are alive.

 

I would definitely take the AI in those mods over the basic IE AI. I might even accept it, although not aways their methods of increased difficulty (improved Bodhi I am looking at you). My wishes for what an AI should be able to do is somewhat coloured by the number of grand strategy games I play I will admit that and I know I cannot have such an AI, but I am allowed to dream

 

The limitation for AI is inherent in the architecture not in the programming.

 

http://en.wikipedia....iki/NP-complete

 

Anything you're going to want to do is going to fall prey to this issue in a RT system, because it's going to have to try and handle NP Complete problems in RT, or problems approaching NP Complete.

 

Shortest path is a prime example, it's NP Complete, and it's something any AI is going to need to handle.

 

http://en.wikipedia....st_path_problem

 

Further, lets say you have two enemies (A and B) attacking two characters (C and D), you could have AC & BD, BC & AD, ABC & D, ABD & C. 4 possibilties.

 

Lets give you two enemies (A and B) attacking three characters (C, D, and E). You could have AC & BD & E, BC & AD & E, C & BD & AE, AC & D & BE, ..., it's a combinatrics problem. Which means that the number of possiblities just for attacking scales upwards at an extremely rapid pace as number of combatants increase, for just a simple attack. That doesn't even try to put any logic into it like assessing what the best combinantion is, or to try and assess the right time to use a special ability, or any other logic. If you give each enemy a bow and a sword, double the possiblities. Give them a special ability too and you'll triple the possibilities, and that's still without even trying to determine which is the most appropriate. It's not inconceivable to end up with AI potentially having hundreds of combinations it has to evaluate in less than a second, with each of those combinations introducing even more variables that have to be evaluated (Like if it has an AOE special ability and must then evaluate the position of all of it's allies and their likely future positions).

 

All of this is competing with the game handling the math in the background, any processing that must occur, etc, and it all must occur in a tiny enough timeslice such that the end user cannot detect it. This is all simply brutal on a computer, and it's why a RT game tends to lack any significant degree of tactics.

 

That's before we even discuss the processor cycles lost in the background to OS duties, background tasks running in the task bar, etc.

 

It's quite possible to develop incredible AI's that'll stomp humans, but it's not going to run in a few hundred milliseconds. Just processing shortest path is going to eat all of that time (And still not finish), which is why most games don't try. They instead try to move in a direct line to the Player, which yields kiting and snagging on environment.

 

I remember you making a similar post in another thread where your argument was also basically destroyed.

 

I am not really going to use too much time on this given that (as Hormalakh said) you should probably read the links you posted. It looks pretty bad when the misconceptions part of a page you link to hits you.

 

Your targeting ideas are also way too complex to be effectively programmed into a game (and in many ways with the AI we have today could be done just as effective with a simple random check and a few predetermined scripts). Remember the AI is not able to do logical reasoning, the programmer have to do all the work. Hence why the AI as a whole is fairly primitive.

It is possible to make an AI that use too many system resources, but it would either require bad programming or devoting way more resources to making it that any game developer would consider (again making it a programming issue). It would also still not be able to accomplish complex tasks effectively (and likely fail hard should it try).

[PsychoBlonde]

 

Oh, but on the subject of kiting, it does seem like a pretty cheap tactic . . .

 

Considering how cheap a lot of the enemies in the BG games were, I don't feel bad at all. Not even a little.

 

Friggin casters.