Dungeon Generation

The generator in 1979's DMG begins thusly:

From the start, it's plain that the purpose here is not to provide content for self-play, but for the DM to use the generator to spontaneously create a dungeon for other players.  It recommends amending results to fit the space available.

When I first encountered this in 1979, my immediate interest was to use the generator to create a dungeon that I could adventure through.  In those days, I did not care much about believability.  If one room had orcs and the next kobalds, and the next after that skeletons, I didn't care.  My concern was to run myself through combats, collect treasure, see where the halls and stairs took me, and do so as often as I liked, for as long as I liked, as there was no reason to explain any of the illogic to anyone.

I have in the past ascribed much of my confidence in running combats to the considerable number of battles that I ran through this generator.  Repetition helped me remember the damage done by every weapon, equate the armour with armour class, test the use of spells in game play, read the lists of spells when choosing them for my own characters, test combat strategies, group dynamics and compare many types of monsters against characters of different levels and abilities.

I recognise that many feel that such activities would be boring, or beneath them, or even a little bit shameful, in that it speaks to an obsessiveness with the game that might be uncomfortable.  I was certainly obsessed with the game in those days, when I did not write about it as now, but rather set up endless physical miniatures on a giant dungeon made of taped together graphpaper sheets, drawing the rooms by hand and then running those battles, keeping track with pencil and paper.  This being before 1985, I did not have a graphically-capable computer for this sort of work, there were no video games that could come close to this kind of combat simulation ... and as I remember it, I had lots and lots of time.

A considerable drawback that I found in the DMG generator as written was the completely useless empty spaces it provided.  The first table, "Periodic Checks," provides only a 1 in 20 chance of encountering anything alive:

There is an equal chance of finding a trick or trap.  There's a mere 15% chance of finding a room.  The rest of it is 75% of empty hallways, stairs, side passages, turning passages and the occasional door.

The room content's table is no better:

Of that 15% of chambers or rooms that occur, only 25% have a monster in them.  25% of 15% is 3.75% ... so that only an 8.75% of total rolls in this dungeon have something that can be fought ... and including the "monster only" result on the table above with the wandering monster on the periodic check table,  there's a 74.3% chance (in the whole dungeon) that any monster that IS discovered won't have any treasure.  Though yes, there is a 1 in 20 chance of treasure being found without a monster.  That increases the chance within the dungeon of generating a treasure result to a total of 2.5%.  That is, 1 in 40 rolls we make using the generator will produce a treasure.

No wonder the generator never developed any popularity.  It comes nowhere near the abundance of monsters and treasure that the company's modules provided!  Worse, because it's random, it's easy to fall into a slump where roll after roll, followed by drawing out the dungeon, can mean no meaningful result for actual hours of physical mapping.  I finally gave up as a kid and just rolled a d8 on the chamber/room table, adding 12 to the result.

I think I started with a d10, adding 10, but then I starting thinking, "Fuck it.  What do I want empty rooms for?"

This doesn't represent a failed concept, merely a very poorly rendered one.  Though I've admitted that I couldn't construct a generator myself ... but in all honesty, this hasn't been because I couldn't just move numbers around and improve the DMG version easily, but because every time I've considered this project, I want it to be much, much more than this.

There aren't enough results.  Granted, though, there are only so many kinds of tunnels, only so many kinds of rooms.  There are only so many things to put in rooms, for that matter.  If the reader doesn't believe me, suggest one addition to Table 5, above, that would be practical to roll every 20 times one encounters a room.  Go ahead.

After a lot of thought on this, I've come to the conclusion that the "periodic check/random room" design is the fundamental problem.  Halls and rooms are not randomly built by engineers who make humanoid lairs.  The placement of rooms within those lairs isn't random.  There are right and wrong places to put storerooms, common rooms, private chambers, treasure rooms.  They can't be scattered willy-nilly everywhere.  Every tunnel requires ungawdly amounts of excavation, shoring up, removal of materials, lighting and maintenance.  And the result of an unneeded hallway is tremendously inconvenient where livability is concerned.  Ask yourself if your home would be improved by an unnecessary 20 foot hallway between the bathroom and your bedroom, or your workspace and your kitchen.

By the way, that's a thing that's needed in 1 of every 20 rooms.  A bathroom.  There are none in this generator.  I see no reason why orcs or hobgoblins would be happy without one.

If the generator's going to be of value, it's got to account for the way we build habitats.  Everything else in this post is going to be specifically about humanoid habitats.  While some randomness does exist (my place is different from your place), the functionality of the way these rooms fit together must be accounted for.  We can't have a 180 foot hallway because we rolled a 1 or 2 three times in a row.

But that creates a problem ... not one of the generator itself, but of the possibility of the generator working the way it did for me 43 years ago.  I can fashion a set of generated layouts, and assign random numbers to them, but it means that as a player, I know what the next 3 or 5 or ten rooms are going to be, before opening any doors.  That spoils the "self-play" ideal I've been considering.  Which I admit is somewhat disappointing.

I've been trying to crack this problem.  The front door of a house is easy enough to generate; we've got to get in.  The front hall is pretty obvious; it's the first room inside the house.  I've created tables for both these spaces.  After that, things get complicated.

In the original dungeon generator, you enter a room, you roll 7-12 orcs (all combatants), then you fight them.  But in any believable arrangement of people dwelling together, when the hell would 7-12 of them be gathered in the same random place?  It might be dinner time, but then there'd be dependents as well.  It might be a common room, but again, dependents. It might be the chieftain's hall, but what, all 7-12 of them are always there, all the time, on the off chance that a bunch of adventurers will show up?

And, point in fact, if they are part of a clan, does 7-12 orcs represent the whole number of combatants?  Or are there another 7-12 somewhere else?  And if so, how many rooms have 1-3, 2-5, 1-6 and just 1 by his or her self?  What is the TOTAL number of orcs, combatants and defendents, in this specific lair?

I've decided that it's a first level dungeon, and that the number of orc "soldiers" and dependents equals about 16 to 28.  From there, we generate "logical" rooms that designate some number of these (each room has a spread of humanoids), until all the appropriate rooms AND humanoids are accounted for, whereupon the "dungeon" dead-ends.

OR, some random result along the way provides a threshold to another kind of dungeon, or stairs down introduces us to more of these same humanoids on "level 2," where double the number of orcs we found on the first level dwell (48-84 total).  Another flight of stairs (or other access) downwards increases this number again ... or perhaps allows the incorporation of the orc's masters, ogres or some such.  It's too early to deal with that problem.

The one at hand, presently, is how to generate those logical rooms?  My first thinking is to not assign the total number of humanoids at all, but to assign room types which then provide a final total of monsters when all the rooms are generated.  Of course, this means that somehow we have to decide when this common room is the "last" room, and there isn't a mess or a hall or a treasure room still left to be found.

Seriously, look at the tables.  If there are only 3 hobgoblin warriors and 7 dependents in the whole "dungeon," there would probably be some treasure, but does a treasure room make sense?  A gathering hall?  A shrine?  A chieftain's private apartment?  But how do you know which rooms don't make sense if you don't know, ahead of time, how many hobgoblins there are? 

Which is fine, if we're generating the dungeon for players ... but how do we generate the dungeon for ourselves, to give ourselves the pleasure of being surprised?

Working on that.