At 10:23 PM 3/31/2004 +0200, William Saar wrote:
> I have just gotten final approval on my thesis on quest systems
> for MMORPGs. It's available from:
While there's little new to me (or, probably, to MUD-Dev), I thought
you explained problem well. I'd like to see your quest system in
action. Has it been used in a published game?
On Morrowind...
I worked on Morrowind and can add a little insight to that section.
The lead designer, Ken Rolston, divided players into categories
similar to the ones you mention. "Pilgrims" are mostly non-violent
explorer or story-driven players. "Conans" just want fights and
loot. It's not much different from more popular player
classifications. It's been a few years since I worked for BethSoft,
so I don't remember the whole system. You might want to contact him
about it.
I wrote all the "faction" quests--the Guilds and Houses and Temple
and Imperial Legion and Morag Tong. The original intent (which was
lost in the 100+ hours/week nightmare of game development) was to
have at least two paths, one "good" and one "evil" in every
faction. Likewise, my original goal was 50 quests per faction. Alas,
50 quests for each of 12 factions was not an achievable goal for 1
person in 3 months... I could write, implement, and do a first-run
test for 3-4 quests on a good day, which was 3-4 times faster than
everyone else, but nowhere near the 6-7 I'd have to do each day to
meet the milestone. The original design also called for a conflict
in each faction that would force the player to choose between, for
instance, the Fighter's Guild or the Imperial Legion. This wasn't
fully implemented because the TES:CS made handling interdependencies
(and multiple solutions) even more difficult than other quest
systems I've used, and because I wasn't overly fond of the idea so I
didn't place a high priority on it.
Alot of the dialogue in Morrowind is uncomfortable because of the
need to use (or avoid) specific words for the hyperlinks. This was
an ongoing problem, and I tried to make keywords as specific as
possible. However, I was not the only person writing quests, and I'd
frequently have to re-do my work when another designer (or
artist... or programmer... or intern...) added a common word to the
pool. I think a better system, for a purely scripted game, would use
a manual hyperlink each time one is needed (and a single file for
each quest instead of adding them to a common database with no
"ownership" locks). So the designer would have to right click on a
word in the editor and add a link manually needed instead of them
being automatically created every time a particular word is
used. This would be easier on the localization/translation teams as
well.
Personally, I prefer branched dialogue for scripted content. Which
could also be used for dynamic content... An algorithm could select
4 or 5 appropriate player responses from pools, possibly even
tracking the most common tone and choices of the player to customize
content. Based on the simple "find treasure, meet beggar" example,
you may be using a similar system already. But I'm getting even
further off-topic here.
The Hlormar Wine-Sot quest was written by an intern and wasn't on
the test list until relatively late in beta. In fact, I didn't even
know the quest existed until beta, which was bad since I was the one
who was more or less "in charge" of quests and books and
dialogue... Being uninformed was partly my fault. I gained a
reputation for rejecting quests because of "silliness" (or just
workload as I couldn't test and fix every quest that other
developers wanted me to do). There is a limit to the number of "man
without pants" quests I want to see in a game... Perhaps this is a
minority opinion. In any case, I remember fixing a few things about
the Wine-Sot quest, but I'm not surprised I didn't catch them all.
World-State Generated Quests
Some of the problems you mention about being unable to change the
game world could be solved if the world itself is less static. In
such a game, quests could be generated by the current state of the
world. Are Orcs moving into the areas around Three Pines? Then the
Mayor of Three Pines will give you a silver coin for every Orc ear
you bring him.
This would be most interesting if "camps" were not static--if mobs
could move from zone to zone and even fight each other for land. A
true population model would be a disaster (basically, a tragedy of
the commons--though this could be used educationally... think "Fish
Trawler 2004"). However, I believe a population model that has upper
and lower limits on population density is feasible. Even if you
didn't want to use a full population model, you could do theme-based
spawning in dungeons. So that while players are in a dungeon, the
mobs are all related by theme. If the current Theme is Necromancer's
Lair, the players will encounter skeletons and zombies. If the
current Theme is Orc Lair, the players will encounter several kinds
of Orcs. Each time the dungeon has no players in it, or whenever the
dungeon is "cleared" by players, there is a chance that it will
change themes. The current Theme of a dungeon could be used to
generate quests.
The main problems with this are time and ego. What happens if the
players get a quest to kill Bob the Wicked Necromancer, and they
arrive just as UberGuild001 finishes rolling for his loot? This
could be solved by instancing zones, but that would be more like
missions in Anarchy Online, and not really a dynamic system. Ego
comes into the picture because developers think they can build
better content than a random number generator. However, my own
experience with players is that they won't notice any of the
nitpicky details you put into a zone or quest (for instance, all the
houses I built for Morrowind are for left-handed residents... all
the pot handles face left, the place settings are backwards,
etc... I did it more as a signature than something I'd expect anyone
to notice, and AFAIK, no one has). If your game is focusing on
quantity more than quality (such as Morrowind and many MMORPGs), a
random number generator can create more content with less work.
Dice Play God
I've always had a weakness for randomness in games. Although most
attempts at dynamic content have been infamous failures, I blame the
implementation instead of the idea. The main problem with most
random content is that the choices available to the random number
generator are too small. Arena and Daggerfall, for instance, took
quests from a very small pool, added some random elements (names,
basically--the name of the dungeon and the person or prize). Once
you got to the dungeon, it was composed of large, pre-made blocks
stuck haphazardly together. This system quickly became dull and
repetitive. After doing twenty quests, I recognized all the quests
and dungeon blocks.
But would it still be repetitive if the quest generator could choose
from five hundred quests instead of five? Or if the quest generator
created quests from event-level elements (as your system appears to
do)? Or if the dungeon builder worked at the
wall/floor/ceiling/furniture level instead of multi-room blocks?
Color me optimistic.