Darkfall "Released"

[Baldrake]

There are a couple of core issues that differentiate programming AI for MMOGs versus SP games.

The main technical issue is horsepower. Advanced AI techniques (planning-based) require a lot of computation and memory. Multiply that out by the thousands of mobs in the game and you run out of resources pretty quick. This leads us back to simple scripted or FSM AI that can be easily computed.

The main playability issue is that most MMO players don't want to deal with hard AI whose actions might be hard to predict.

[ashrik]

I remember reading articles written by AI developers who would note how hard it was to build tough and realistic A.I.s. Looks like Aventurine thew out the realistic part of the equation and turned their mob bots all into duck-and-weave kiters. I don't know who told you that making an AI for an MMO was automagically harder, because I've seen no proof of it. Monsters in our current blend of MMOs rely on little more than an aggro list and a cooldown list. Are we going to pretend that because the fellows in Greece are doing something different (read: more annoying) that it's more advanced?

Games like FEAR 2 and Killzone 2 are typically used to show off advanced AI- the bots make active use of their environment for cover, flanking attacks, etc etc. Say you walked into a room and all the soldiers immediately ran out of it. You chased them around a corner and they popped you in the face with a shotgun, and ran away to the next corner. You'd say that was a shit AI. Hard- yes, but shitty. Pop it into a platformer or beat-em-up? Still shitty. But put it in an MMO and it's advanced? Well fuck me sideways, I don't see why they wouldn't make you chase down trees or ore. We'd then be praising the advanced and immerssive crafting system instead. Let me know when the mobs can do something different than run away like an annoying ass and we'll talk. Hopefully we'll see something innovative like attack the healer always!

Playing a game where goblins acted like goblins, and not 12 year old Elf Rogues named SephigolasxX, has never broken "immersion" for me.

[Montague]

Playing a game where goblins acted like goblins, and not 12 year old Elf Rogues named SephigolasxX, has never broken "immersion" for me.

The videos remind me of D&D DM's that drop mindless undead that coordinate with military precision on level 1 players. Something tells me the developers of this game are going to have loads more fun than the players.

And yeah, it looks like a semester project at State University's Game Design 401 class.   

[Ingmar]

I remember reading articles written by AI developers who would note how hard it was to build tough and realistic A.I.s. Looks like Aventurine thew out the realistic part of the equation and turned their mob bots all into duck-and-weave kiters. I don't know who told you that making an AI for an MMO was automagically harder, because I've seen no proof of it. Monsters in our current blend of MMOs rely on little more than an aggro list and a cooldown list. Are we going to pretend that because the fellows in Greece are doing something different (read: more annoying) that it's more advanced?

Games like FEAR 2 and Killzone 2 are typically used to show off advanced AI- the bots make active use of their environment for cover, flanking attacks, etc etc. Say you walked into a room and all the soldiers immediately ran out of it. You chased them around a corner and they popped you in the face with a shotgun, and ran away to the next corner. You'd say that was a shit AI. Hard- yes, but shitty. Pop it into a platformer or beat-em-up? Still shitty. But put it in an MMO and it's advanced? Well fuck me sideways, I don't see why they wouldn't make you chase down trees or ore. We'd then be praising the advanced and immerssive crafting system instead. Let me know when the mobs can do something different than run away like an annoying ass and we'll talk. Hopefully we'll see something innovative like attack the healer always!

Playing a game where goblins acted like goblins, and not 12 year old Elf Rogues named SephigolasxX, has never broken "immersion" for me.

I don't think I would say that was a shitty AI. It also strikes me as how goblins *would* act. They're little weasely types, just the sort to run off and lead you into an ambush where they have the upper hand.

On the other hand you wouldn't want, say, an ogre or something similar to act the same way.

[ashrik]

My point is that it acts like something playing a video game, not a goblin- and one gaming the system weirdly at that. Because troops from country XYZ are notoriously jumpy, it doesn't make sense for bots to be jump-ducking when I'm playing Battlefield 2. It's hard, and cheaply so, not advanced. Or special.

I haven't heard any description of A.I. that was not like that Goblin, but hopefully not everything works like that. We'll see. Maybe that guy from earlier will be back to tell us how stutter-strafing bots are totally pushing the envelope as far as developments in artificial intelligence go.

But hey, maybe that's what DF players want. Or have at least told themselves they want, because AV put it in-game.

[Lantyssa]

Please play this card and I'll shred it up; because the majority of singleplayer games take advantage of the limited amount of space AI have to move in.

MMO AI programming is much harder, so some allowances have to be made.

Please to be giving your credentials.  I'm calling BS.

The only differences between any given AI, single-player or MMO, are the amount of CPU cycles you're willing to dedicate to them, the environment in which they exist (which varies incredibly from game to game), and the competency of their programmers.

AI might be hard, but there is nothing inherent to an MMO which makes it harder.

[Venkman]

I don't know who told you that making an AI for an MMO was automagically harder, because I've seen no proof of it. Monsters in our current blend of MMOs rely on little more than an aggro list and a cooldown list.

Err, your lack of proof is based on it not having been done.

Baldrake mentions the end user expectations as the second problem with making "smarter" AI (the first being computational cycles), but I actually think it's the first. These games are not designed for every single fight being a challenge of planning, trial, error, and luck. Players do not want this. Not with that big ass XP meter sitting there.

Let's say Blizzard flipped the magical switch where all mobs a) got smarter; and, b) operated within culture expectations. Now you've got much more aggressive social aggros, seeking cover, constantly moving out of your cone, etc etc. I can count on one finger the number of picoseconds between that and the end of Azeroth.

People want to talk about the technical challenge of making smarter AI and that's fine. We'll just have the same armchair programmer discussions we've had for years. But to actually deliver better AI is to fundamentally rethink the motivation of playing these games.

If you want a world populated by an intelligent NPC culture, you have to rethink the role players play in that world first. Right now we're just a bunch of sadists commiting mob-genocide on our path to the optimal template. Smart AI would break that right quick.

[UnSub]

AI might be hard, but there is nothing inherent to an MMO which makes it harder.

I'd say the amount of CPU time to operate the AI and transmit it to a large number of players at once makes things harder for MMOs.

Honestly, if mobs were 'smart', they'd form large gank parties, zerg rush all players (so that players wouldn't be able to solo) and every occupied cave / town / castle / hole in the ground would be fortified. The first act of every mob on encountering the players would be to raise the alarm so that every other friendly mob in the area would seek to swarm the players. Right now, the AI is set to aggressively loiter in a limited area and spends their time praying for the sweet release of death.

'Smart' mobs would be no fun, especially for the casual soloer.

[tmp]

If you want a world populated by an intelligent NPC culture, you have to rethink the role players play in that world first. Right now we're just a bunch of sadists commiting mob-genocide on our path to the optimal template. Smart AI would break that right quick.

Pretty much. This is mentioned often enough, but typically when a player says "i had so much fun playing the game" (in MMO at least) it's associated with "i was playing this broken template that let me two-shot the shit in droves". People don't play these games to spend half hour trying to stab one goblin, they are playing them to feel powerful. City of Heroes kind of player/mob power balance.

[Lantyssa]

I'd say the amount of CPU time to operate the AI and transmit it to a large number of players at once makes things harder for MMOs.

Running a combat script?  Not anymore than you need more CPU time for multiple people running around.  Even if you say it will scale with a little more overhead compared to a single-player gamer in their own world, you'll have a signficant overlap from players in the same area such that it might even be less processor intensive because the number of active NPCs is shared.

If you say well, there's this area of dozens of mobs where a single player game might have five, then it's a design issue.  Why do you need that many?  Historically because they've made the AI so dumb you scythe through them like you're harvesting a field of mobs.  Or spawn things dynamically.  We could probably have a twenty page discussion on How Things Are Done that would impact this in all kinds of ways.  That doesn't mean you need to be burning an excess amount of CPU time running their AI though.

If you want to take it to a level beyond RadiantAI where mobs live their lives and do things on their own, maybe.  You'll be scaling the server and running parallel processes anyways.  These things can be simulated though, and until a game lets NPCs grow, die, own land, clear dungeons, and otherwise act like players that kind of processing power still won't be necessary.

It's not harder.  MMO game design is just so far behind single-player games in a variety of areas that, surprise!, AI is one of them.

[patience]

It's not harder.  MMO game design is just so far behind single-player games in a variety of areas that, surprise!, AI is one of them.

It is harder. Unsub wittingly (happy?  ) exemplifies my point when he stated how he would design NPCs to be smarter.


NPCs should never be designed to follow the optimal strategy because in reality it may turn out not to be optimal.

But that doesn't automatically prove AI design in MMOs is inherently harder in single player games.

What does is the need to develop AI that can formulate multiple strategies because you'll be dealing with multiple people at once employing their own strategies to complete their objectives.

A second and third reason MMO design is harder has to do with addressing the playing habits of very different people. I haven't designed an MMO but I have given thought about doing that. Specifically to the AI part I was thinking of the requirements of creating a system where NPCs remember what players have done to them and what they themselves have done or seen in the past.

A lot of the problems with the system revolved around how the memory should be shared between the server and the client and how would I represent this past information in a way that doesn't bog down either machine because of the ongoing nature of MMOs. In a singleplayer game you don't have to worry about making this memory persistent because singleplayer games are not persistent.



A fourth problem revolves around certain types of MMOs.

If an MMO isn't designed to be a seamless world and everything is bounded by a map design philosophy similar to a Quake arena, well then an MMO doesn't need to be any different than a single player game, in terms of develooping AI and not be inherently harder. But many MMOs strive to create a seamless world because players prefer that even if the content is sort of on rails like a level based game.

NPCs to be really human, most would have to exhibit behavior that allows them to roam beyond their normal territory; because other NPCs or players are making things difficult for them and they have to do something to overcome these challenges. A singleplayer game can have a really large world like an MMO but in a singleplayer world you can expect your AI to have to deal with multiple AI and a very limited number of people. You could easily cheat on the issue and create an illusion that NPCs are free thinking agents because only a handful of people are roaming the world but in an MMO you can't use the same tricks to create that illusion.


A fifth reason is the one Unsub pointed out. You touched on why why it can be harder for an MMO so I see little need to elaborate.

People want to talk about the technical challenge of making smarter AI and that's fine. We'll just have the same armchair programmer discussions we've had for years. But to actually deliver better AI is to fundamentally rethink the motivation of playing these games.

You are doing this wrong. Motivations are varied and a market exists for anything. What matters is how large that market is for that specific type of mindset.

[UnSub]

It's not harder.  MMO game design is just so far behind single-player games in a variety of areas that, surprise!, AI is one of them.

It is harder. Unsub unwittingly exemplifies my point when he stated how he would design NPCs to be samrter.

I'd like to think I did it wittingly. 

[tmp]

If you say well, there's this area of dozens of mobs where a single player game might have five, then it's a design issue.  Why do you need that many?  Historically because they've made the AI so dumb you scythe through them like you're harvesting a field of mobs.

(..)

MMO game design is just so far behind single-player games in a variety of areas that, surprise!, AI is one of them.

MMO mobs are of the galaga/r-type/whathaveyou variety -- they pose 'challenge' (if you can call it that) with their numbers, not individual AI prowess. Yes, it is a design decision. But to say this somehow put them behind single-player games because *some* single player games do it different... well, that's just dumb imo. Because it ignores both the single-player games which uses the same 'many mobs' approach you get in MMO games, *and* the MMO games with more advanced AI behaviours. The mobs that run when hurt, call help when outnumbered, cc and kite players taking advantage of their range, ambush them etc... are not new to the MMOs. They aren't popular though because lot of the players don't seem to like them.

[Fordel]

I think people over think what it takes to make AI "smart". Mobs don't need tactical matrices or whatever.

They just need to be able to attack the healer first. 

[patience]

Brilliant Level 10 Healers with speed boosts aavoiding aggroed mobs while level 100 players attack without penalty will show how smart NPCs can be.

You are simply replacing tank and spank with run and gank.

[DLRiley]

Mobs carry snares 

[slog]

These AI replies I'm reading are idiotic.

MOBS in a PvP MMO are walking bags of water that you pop with your sword to get the resources out.  Having actual smart AI would be retarded.  They are supposed to be farmed like lambs to the slaughter.

[Zzulo]

http://se.youtube.com/watch?v=3lOnddAJL_o&eurl=http://s1.zetaboards.com/betaleaks/topic/1153222/1/

more ingame footage!

Apparently beta leaks are everywhere now.


Oh and they really need to fire the guy who does the animations.

[Signe]

Yikes.  That was awful and so were the others, but that one was the worst I think.  Everything looks like ass, especially the ass!  I think it has something to do with whole pineapples.

[Sheepherder]

I'd say the amount of CPU time to operate the AI and transmit it to a large number of players at once makes things harder for MMOs.

The transmission is already done with the zergling AI that most modern MMO's have, because the only thing you really need to transmit is where the mobs are and what they are doing.  The additional overhead would only occur at the server.  At this point it is probably important to note that the largest bottleneck you will find in an existing game with a less primitive AI and a seamless world (Oblivion) is usually the GPU.

Note: Oblivion has extreme problems with rendering sound efficiently, most people who complain about extreme combat lag in Oblivion can solve it by disabling footsteps.

[Lantyssa]

MMO mobs are of the galaga/r-type/whathaveyou variety -- they pose 'challenge' (if you can call it that) with their numbers, not individual AI prowess. Yes, it is a design decision. But to say this somehow put them behind single-player games because *some* single player games do it different... well, that's just dumb imo. Because it ignores both the single-player games which uses the same 'many mobs' approach you get in MMO games, *and* the MMO games with more advanced AI behaviours. The mobs that run when hurt, call help when outnumbered, cc and kite players taking advantage of their range, ambush them etc... are not new to the MMOs. They aren't popular though because lot of the players don't seem to like them.

I've run into very few mobs that impressed me with their AI, much less game worlds.

Yes, it is a design decision.  Just like many DIKU standards have become the accepted norm.  If it weren't an across the board observation it might be dumb to claim, however it's standard operating procedure.  MMO players are so used to candy pinatas that of course they don't like decent AI.  It's new and alien from their perspective.

I'd also love to know what MMOs had these super-AI mobs you mention.  I've seen some of it faked, but not consistant, intelligent behavior in any MMO.  The only one I can think of that gave me a few 'Whoa!' moments was Ryzom.

[Mrbloodworth]

I loved Ryzoms AI, also, LOTRO got an AI update recently, quite impressive.

Anyway, perhaps we should get back to Darkfall.

[Signe]

Anyway, perhaps we should get back to Darkfall.

Nooo!  This is F13.  That would be contrary.

[Jayce]

Also, why are they wasting great AI on a PvP game?  The point of the mobs/xp/grind aspect of a PvP game is that they're a resource, not some great challenge to overcome.  They should be as challenging as harvesting wheat unless they are the point of the game.

4. PvE isn't just a resource to some pvpers. They want the challenge of fighting PvE mobs ramped up by developing strategies to protect themselves from PKers and other PvPers interested in the mob. These people also demand for better AI from mobs to ensure it's not just pvpers offering a challenge.
5. Even people who view mobs as just another resource to harvest can't stand how mind numbing brain dead AI is, because it makes resource collection too boring. (See Eve Online forums) If they loved braindead mobs they would have no problem mining or gathering but they prefer to avoid that too because it's not a combat scenario.

My point is that a game should have a core competency and focus on it.  Most can't get even one correct, so focusing on doing two stupendously well (which we're being asked to beleive - the best PVP MMOG and also incredible AI) is not only a recipe for disaster but probably not what you want anyway.

Second, your #5 point doesn't stand.  I play EVE and the last thing I want is intelligent AI.  There are plenty of (relatively) intelligent players out there, and they drop better loot.

Lastly, I don't think there will ever be an AI system that people don't figure out how to game within a week (I wish there was a Colbert emote here).

[Fordel]

Mobs carry snares 

[tmp]

Yes, it is a design decision.  Just like many DIKU standards have become the accepted norm.  If it weren't an across the board observation it might be dumb to claim, however it's standard operating procedure.  MMO players are so used to candy pinatas that of course they don't like decent AI.  It's new and alien from their perspective.

Well uhm, duh. Again, this is like making a complaint that swarms of mobs in the arcade shooter don't all just focus and blow up the player no matter how well said player tries to maneuver their craft rather than fly by in pre-defined patterns, and that making them this type of 'smart' would be in some way a desirable step forward that's woefully lacking.

But --also again-- this simplification to 'standard operating procedure' conveniently ignores the cases where this standard operating procedure is not followed. Using this very kind of reasoning i could claim here that single player games are inferior because they fail to provide the thrill of multiplayer competition. Which is of course a batshit crazy thing to say, but i'm simply ignoring the games which do come with some form of multiplayer. After all the multiplayer is not part of the standard operating procedure for single player games; and what's good for the goose...

I'd also love to know what MMOs had these super-AI mobs you mention.  I've seen some of it faked, but not consistant, intelligent behavior in any MMO.  The only one I can think of that gave me a few 'Whoa!' moments was Ryzom.

The behaviours i mentioned come from LotRO, but that's by no means the first MMO to have them. I would be also wary of making difference between 'faked' and 'consistant, intelligent' in a game -- unless you have the insight into the code you will never know for sure what is a well-made fake and what isn't, thus trying to create a division here along lines of "MMOs just fake it" (with the unspoken 'and the real games don't') is at best a stretch.

Not to mention, like others pointed out already the "consistant, intelligent behaviour" from NPCs would basically require the NPCs to make sure the deck is always stacked in its favour -- never attack the player if the victory is not guaranteed, and get the heck out if it's likely the player might "kill" them. The ages-old basics from Sun Tzu. Anything less is not intelligent, it's just bit more intricately arranged smoke and mirrors that's still supposed to make the player feel good about themselves in the end.

[WayAbvPar]

Apparently the upcoming EVE patch is going to upgrade NPC AI, at least in some spots. Should be interesting to see what they come up with.

[Lantyssa]

Not to mention, like others pointed out already the "consistant, intelligent behaviour" from NPCs would basically require the NPCs to make sure the deck is always stacked in its favour -- never attack the player if the victory is not guaranteed, and get the heck out if it's likely the player might "kill" them. The ages-old basics from Sun Tzu. Anything less is not intelligent, it's just bit more intricately arranged smoke and mirrors that's still supposed to make the player feel good about themselves in the end.

It wouldn't require always acting "intelligently".  It would require acting convincingly for a member of whatever species you are trying to portray.  Some things also have to be sacrificed for the sake of fun, and I'll concede many decisions are based on that.

I'm not talking completely out my ass here, either.  I've done mob procs on the MUD I helped program.  Not a huge number, and certainly not as advanced as is possible to get (I wasn't interested in re-writing how the game ran to allow better AI; what we had was sufficient in its context), but I have done it.  It can range from randomly picking its options to an intricate set of conditionals.  Usually my main concern was making a fight interesting rather than deadly.  Except for the procs on our avatars.  That warriormonk was nasty.

As for having insight into how they programmed something, when I play a game, I do see it in code.  I poke at it and try to figure out how all kinds of systems behave.  It is incredibly difficult for me to just play a game and not try and reverse engineer it.  It's a curse of being a code-monkey.

Regardless, the original point I care about and is getting lost, is that it is not any harder to program for an MMO than a single-player game.  When it is, that is simply because of other design decisions and (lack) of programmer or designer skill.  The toughest part isn't so much how it acts as it is balance, and that is a direct result of most MMOs being level-based systems.

[tmp]

It wouldn't require always acting "intelligently".  It would require acting convincingly for a member of whatever species you are trying to portray.  Some things also have to be sacrificed for the sake of fun, and I'll concede many decisions are based on that.

Well, i'd argue that i'd expect intelligent behaviour in order to be convinced, anything less it's a rather obvious bow towards the player's comfort and as such an immersion breaker. But yes, i hope we can agree that fun trumps the realism in games, and it just happens that intelligent enemies in overwhelming numbers are simply too hard to make fun... hence the genre tends to favour large numbers of dumb enemies over single, threatening targets..? (these being reserved for the "epic boss fights" or whatever) A design decision that just doesn't deserve to be picked on.

Regardless, the original point I care about and is getting lost, is that it is not any harder to program for an MMO than a single-player game.  When it is, that is simply because of other design decisions and (lack) of programmer or designer skill.

I have to disagree with this. It should be obvious why it is in fact harder to program the AI for a MMO rather than single player game -- the single player game has the AI fighting no more than one player. The AI in MMO has also to make decisions about target selection, in addition to the same problems it faces in single player game. This is extra factor which will make the code relatively more complicated. (yes, an obvious simplification that overlooks the games with AI-based squads supporting the player, but these are somewhat different kettle of fish as there is still only one player whose enjoyment out of the game really matters)

These aggro mechanics people love to hate and to abuse to their best benefit? Someone has to code them, and to make the overall end-effect of their working at least basically functionable. Single player games can (mostly) avoid this issue.

[kildorn]

Apparently the upcoming EVE patch is going to upgrade NPC AI, at least in some spots. Should be interesting to see what they come up with.

EVE doesn't have AI, or didn't last I checked. It simply aggro'd first proximity target or first attacker, then stays there forever.

[Murgos]

EVE doesn't have AI, or didn't last I checked. It simply aggro'd first proximity target or first attacker, then stays there forever.

That's still AI.  Also different rats do have different things they do when attacking.  Some come in close and others try and go to range and etc...

No, one has ever claimed it was brilliant but it is still some behavior.

[Lantyssa]

I have to disagree with this. It should be obvious why it is in fact harder to program the AI for a MMO rather than single player game -- the single player game has the AI fighting no more than one player. The AI in MMO has also to make decisions about target selection, in addition to the same problems it faces in single player game. This is extra factor which will make the code relatively more complicated. (yes, an obvious simplification that overlooks the games with AI-based squads supporting the player, but these are somewhat different kettle of fish as there is still only one player whose enjoyment out of the game really matters)

These aggro mechanics people love to hate and to abuse to their best benefit? Someone has to code them, and to make the overall end-effect of their working at least basically functionable. Single player games can (mostly) avoid this issue.

Ah, but we just got done agreeing most MMO AI is dumb.  Dumb AI is easy to program.

Compare FEAR 2 versus the standard mob in WoW, EQ2, CoX, etc.  That AI is a lot more complex and I can guarantee it was a lot harder to program.  What makes an AI more difficult to program is the complexity of its actions and the options available to it.  Fighting multiple opponents is more complex, but it's not that difficult.  It involves giving a priority list and actions to take against those types of targets.  With the crappy aggro mechanics most MMOs use, it doesn't even matter much because it's essentially fighting one target.  Others just happen to be beating on it.  Letting a mob pick an optimal target based on what's around it is trivial though:  if (is_mage($target)) then geek_the_mage($target);

I did think of an online game with really good AI:  Guild Wars.  It's almost all combat oriented, but it's one of the best I can think of.  It's pretty easy to replicate though as it just gives skills a priority based on existing conditions.  Things like use an interrupt if available and target is casting.  Cast the appropriate healing if needed.  Cast enchantments if in battle.  Use a corpse spell if bodies are available.

Realistic behaviors outside of combat are the most difficult to program.  Those require pathing and interaction with the world.  Things which are common to any game regardless of who plays them.  At the risk of repeating myself yet again, it really is not inherently any harder due to being single or multi-player.  Overall game design and platform plays a much larger role.

[tmp]

Ah, but we just got done agreeing most MMO AI is dumb.  Dumb AI is easy to program.

Or perhaps it's not as much that dumb is easy to program, but the combination of smart + handling multiple enemies winds up as too heavy combination and something has to give..? And since you can't get multiple players out of the multiplayer game (the all-popular instancting aside) it is the smarts that get cut.

Compare FEAR 2 versus the standard mob in WoW, EQ2, CoX, etc.  That AI is a lot more complex and I can guarantee it was a lot harder to program.  What makes an AI more difficult to program is the complexity of its actions and the options available to it.  Fighting multiple opponents is more complex, but it's not that difficult.

Define "that", though -- i'd say it is much easier to program the AI to duck out of line of fire of single opponent or to path their way around to get a jump on them or whatever, than it is to do the same when you have to take into account 5-20 people all with their own weapons, position, movement and desire to kill. So i'm really not surprised we are not seeing that in MMOs, because it'd involve entirely different level of complexity. Do it by simply copying the approach from single player game, and your players will just laugh at you as they game the AI to string the NPCs for easy kills, with no effective change to gameplay from the current popular MMO model (but with considerably more work put into it)

With the crappy aggro mechanics most MMOs use, it doesn't even matter much because it's essentially fighting one target.  Others just happen to be beating on it. Letting a mob pick an optimal target based on what's around it is trivial though:  if (is_mage($target)) then geek_the_mage($target);

That's in green, ain't it? Since your trivial approach is basically the equivalent of "stand there and shoot until one side dies", i.e. the very thing people mock about the MMO AI. And ironically enough that part of their AI you're trivializing can be quite a bit more complicated than that (in some cases; this will obviously vary, much like quality of NPCs in single player games can vary as well) I think it deserves as much credit as the single player NPC AI gets here -- when you look into it both are after all very similar problem, and boil down to assignment of 'correct' priorities to the list of options you have.

[Lantyssa]

You're failing to understand that the routines in geek_the_mage() could be simplistic or complex.  They are in turn highly dependent upon what is possible in the game world in question.

I can give a specific example and outline exactly how I'd write a generic combat script to handle single or multiple targets, but I'd need to know a lot about what's possible in that setting.  And whether the mob is limited by a class system or if it can have abilities above and beyond players.  Is it meant to be beaten one-on-one or should it be able to give even a group a run for its money?  How aggressive is it?  Does it value its 'life'?  Do we get into the more advanced pack behavior, and if so do we have differences in the members or are they all the same?  Give me some parameters and I'll show you just how simple it is to write.

[slog]

Here is my uber AI algorithm

1) Kill the healers
2) Kill the DPS
3) Laugh and run away from the tank.