New topic: AI and NPCs

[Caliban Tiresias Darklock]

I used to run an AD&D adventure at conventions which illustrated, in my
mind and several others, the exact failings of most online RPGs,
computer RPGs, and the vast majority of DMs. I'll lay out the general
structure of this adventure as background, but refrain from war stories
and specific instances.

The adventure was always run on each of three successive days, each time
in its entirety. Players were encouraged to register for more than one
session if they desired. Character generation rules were simple: using
only the Players Handbook, DM guide, and Unearthed Arcana (with the
advent of 2nd Edition AD&D, the Unearthed Arcana rulebook was dropped
from allowed references), you could take any one character and provide
him with one million experience points. If it was a multiclass
character, this was divided among classes; if a human dual-class, it
could be divided however you preferred between the two classes. This
determined your level. Hit points were to be the maximum for your class
and level, considering CON bonus and all. Stats were wide open with a
maximum combined stat total of 100 (e.g. no straight 18s). In addition,
each character could select from the DM's guide a total of 5000 XP worth
of magic items, but no single item could be worth over 1500 XP and no
items could grant wishes. Total player registration was at least six to
run and no more than twelve before applications were closed.

The title of the adventure was "They're Only Kobolds". Your opposition
was one hundred and twenty kobolds (combined hit dice: 60, in comparison
with combined hit dice of 72 to 144 for the party), completely detailed
on the game advertisement and in total concordance with the Monster
Manual guidelines on lair contents. The detail on the advertisement
listed not only all the monsters, but their complete statistics and all
treasure present in the lair. The final line of the advertisement was
"None of you are expected to survive". It was a challenge. We never had
trouble attracting (and wholesale slaughtering) full twelve person
parties on each day.

The way the game ran was somewhat complex. Two hotel rooms were used,
generally separated by a good distance and often on separate floors. I
was in one room with the party. My co-DM was in the other room, equipped
with one of a pair of walkie-talkies. I had the other. He controlled the
monsters. He had a map, and a list of his resources. The map was, in
fact, built by him before the convention -- his home, his control. He
placed all traps. He placed all creatures. He used the treasure listing
to equip his minions with appropriate magical items, and to buy supplies
with which he could build the traps he proposed. Everything was
accounted for. During the game, he would report on his actions, and I
would report on what he saw of the players. All dice rolls were made in
the room with the players, by me, to reassure the PCs that no cheating
was going on in combat. 

In the several years we spent running this adventure, before Scott moved
to N'awlins (some of you may have played it if you attended conventions
in the D.C. area around the mid to late eighties), no one ever survived.
(We had a couple parties actually quit, leave the dungeon, and go back
to their boring lives -- when you step into a dungeon and watch a series
of a half dozen kobolds systematically eliminate your two point men
before you even know what's happening, it tends to demoralise. I don't
count that as survival.)

The problem we were pointing out was that kobolds, weak as they are,
have an intelligence rating of 'average' in the Monster Manual. They
don't charge headlong into combat. They have actual tactics. They
operate as a group. And they're fiendishly efficient at guarding their
homes. Which brings me to the actual question, finally.

Has anyone managed to implement something resembling tactical intellect
in MUD mobs? I've been thinking about this sort of halfheartedly,
lately, and considering that this sort of thing isn't often seen. There
are several very simple tactical maneuvers which appear at first blush
to be rather easy to implement -- selection of targets, for example. In
most D&D style games, the creatures should concentrate on removing
single targets rather than spreading attacks: a wounded creature hits
just as hard as a healthy one, but a dead one doesn't hit at all. In
systems where spellcasting can be easily disrupted, such disruption
should attract at least one creature's efforts. Different methods would
fit different creatures, such as some who would target the weakest
creatures first to thin the numbers rapidly and then retreat, while
others might attack the strongest creatures first in order to simplify
the extermination of the entire group. 

There are additional possibilities, like the inclusion of raiding
parties: groups of creatures that go out and literally attack and
obliterate areas to loot and pillage them. This could actually create
rudimentary quests, as important families could on occasion attract
kidnap attempts and have an automatic response of posting a reward. 

Anyone thought much about this? It *looks* rather simple, but like I
said I haven't really gone in depth. What sort of efforts have other
people made in this regard?