Building a dungeon as it's being explored?

[MonkeySage]

The last dungeon I designed, I planned out everything; exact number and type of enemies, precise location of traps, etc. I drew out a map with a precise shape, and events that happen when X happens.

I realized this was probably going to fall apart, and when my players completed it, I realized that my method was probably mistaken.

Most of my players loved the experience, but one of them complained that: The corridors were too narrow(it's a human mineshaft, i figured it wouldn't be as clean and spacious as a dwarven mine. The encounters were too few and far between, too predictable, and even the boss fights were too easy.(I went easy on them because they were level 1 and it's the first time I ran a dungeon for a level 1 party).

So this next dungeon, I wanna do things a little differently:
For starters, no map. I can take short notes and draw simple maps as my players go on, just for record keeping's sake, but before my players go down a corridor, it won't be drawn out.
Only plan my boss fights, and randomly generate all other encounters.
Randomly generate traps.

I may only make a brief outline before they go in, of the things they might find.

Bear in mind, this is the second dungeon i've ever created(i had others in mind for the future but haven't actually planned them yet in any detail). I'm pretty much a novice when it comes to this sort of thing.

I have a few major events in mind as well: one room has a Mirror of Opposition . At least one boss fight for every stage of the dungeon(I want it to carry them from level 3 to level 6 at least).


They've been investigating this figure called The Ram that's been trying to build a goblinoid(as in hobgoblins, goblins, and bugbears) army for some unknown reason. My idea is that they'll first encounter him towards the end of this dungeon. I don't want to reveal his intentions or anything, I just want them to meet him for the first time, and maybe get his attention. He is a level 10 Cleric.

The dungeon is a cave system near a coastal city they recently arrived at.

So what all should I consider for this dungeon if I follow the method that I described above? What in the DMG would be most helpful to me?

[Zirconia]

I would actually say your first time, where you pre-designed everything, is likely to give a more satisfactory experience. If you have to generate most of the encounters on the fly, unless you have all the monsters memorized, they will have to sit around while you look up abilities, figure out where to place the critters, etc. Also, how creative are you likely to be able to be with architecture on the fly? Either you or they will have to record it as they go, so if they suddenly decide to flee from something you can figure out if the corridor connects or whatever.

I would say just address the specific complaints they had. Narrow corridors you are certainly allowed as a DM, next time they might be natural caves, or gigantic, or something else depending on who lives there. If you have everything designed, pre-generate a few random encounters you can throw in as complications, maybe most of the way through a fight something is attracted by the noise if the fight was too easy. Or, conversely, something can jump in and attack the baddies if you make TOO hard of a boss fight.

In general, having more prepped means you can keep the flow of action going faster. Something I have seen people suggest, though, is run with ideas the players have. If you have a useless but decorative column in a room, and someone speculates that it conceals a trap door if they can move it, let them move it and find a crypt with ghouls underneath. Or come to life, or turn out to be a Mimic. That gives less of a "dead space" feel.

[MonkeySage]

What if... I draw out a map on grid paper, plan out my traps, only have randomly generated encounters in the corridors? Each designated "room" could have planned encounters... And out of the random encounters, have smaller selection and thus less need to look up monster entries.

At least half of these monsters were rolled by myself(for example, level 2 hobgoblin fighters were used in our last encounter. I rolled them up on Dingle Games.)

[Popertop]

I'd definitely say planning has its merits, I've been trying to build on the fly and its hard to come up with stuff right away.

As long as you have a decent outline of the dungeon, and a few enemies rolled up already, as well as environmental stuff (traps, lava pits, etc) to throw in whatever room you need to, that should be a solid pool of resources to have flexibility with, without taking too much time to generate spontaneous content on the fly.


tl;dr about half preparation is what I call this, have some stuff rolled up, and an outline of the rooms/furnishings.

[Knaight]

I'm an extremely improvisational GM. I barely plan at all for most things, and find it works out well. Dungeons? Those are pretty much the one big exception where planning suddenly becomes necessary, unless you're going with something really small. It's part of the reason I tend not to use them anymore. This applies much more to physical layout than anything else though, so I wouldn't necessarily bother with planned encounters.

[NichG]

Planning suffers from wastage - if you want to make something nonlinear, you have to accept that parts will not be used. I would suggest that planning is best used to create 'constraints' on the dungeon rather than specific layouts.

For example 'this dungeon has a clan of kobolds and a clan of duergar that have an unsteady peace; there are sites of old battles, places where there are heavy traps on dead ends to discourage the other faction from random searching, and shibboleths that can be used to identify allies vs enemies, such as something that shows up to duergars' 120ft darkvision but can't be detected with kobolds' 60ft darkvision, or something that only a kobold could stand the smell of, or whatever. The duergar have a complex religious structure with a secluded 'inner circle', whereas the kobolds are animistic and tribal'. On top of that, I might design a handful of set pieces that are hard to do on the fly (e.g. things with puzzle-like elements or particular scenarios you think might be interesting, like a fight on a narrow rock ledge).

Then, as the party explores, I would make up a story about the current area to myself and fill things in that are self-consistent within that story and what the players have discovered so far, possibly incorporating one of the set pieces at a choke-point that is likely to get explored. Basically I'd mentally map ahead maybe 2-3 rooms (e.g. 'this is a kobold living space, so theres a communal room where food is prepared and eaten, a room where the older kobolds make clothing, weapons, and traps, a barracks area, and a hidden 'safe room' that only kobold-sized creatures can fit into). That way player decisions and scouting can still be meaningful (because the layout fits a self-consistent picture of the functions of a kobold lair) without me having to map out places that may never be explored.

Another way to do it (or rather, a complementary technique) is to design rooms ahead in accordance to the hints you drop beforehand. In other words, when the players are listening around/etc, decide then and there 'okay, this is the next room' and then give a hint as to the contents of that room. This again gives the players the ability to plan ahead for future threats or decide to go around, while not requiring you to intricately plan out every detail.

Overall the trick with improvisation is to come up with stuff not when the players first set foot into it, but rather when the players should first be able to become aware of it. If you can do that well, then the overall experience from the player point of view should be indistinguishable from a pre-planned dungeon.

[KorbeltheReader]

I would strongly recommend that, if you randomly generate stuff, you randomly generate it before the session begins. Nothing slows a game down like the GM thumbing through tables and rolling dice on his own.

Also, while randomly generated traps are easy for the GM, they tend to be pretty boring for everyone except the skillmonkey. I know a guy who refers to them as "wandering damage," which is pretty apt. A little bit of brainpower poured in to making traps that are most than just "[flavor text] 1d6 damage, save for half, DC 20 disable device" can improve trap encounters greatly.

Also, don't fault yourself for miscalibrated encounters for ECL 1 characters. At level 1 the happy medium between "easy" and "TPK" is super narrow.