Sunday, March 21, 2021

On Dungeon Size

In the most recent Questing Beast Q&A he and his guests gave their thoughts of "ideal dungeon size" and it got me thinking. Here's a link to the part of the video where they discuss it. After some consideration, I want to propose 4 basic size classes of dungeon, divided partially by number of rooms but, more importantly, by the effect they have on the core gameplay loop of your campaign.

Micro Dungeons

These are your 5-Room Dungeons, your Delian Tombs, your DCC funnels, etc. Pretty much anything between 1 and 6 or 7 rooms. The 5E starter kit has one as its introductory dungeon immediately after the combat encounter that kicks off the campaign, and it's pretty damn solid. By definition, you can pretty much always finish these dungeons within a single session. That makes them ideal for one-shots, and indeed, they're popular at conventions and instances where you're introducing new players into the hobby.

But if you're playing a whole campaign, I think they can serve a special role that's underappreciated. For one thing, collecting tons and tons of tiny dungeons is a great thing to populate an open world sandbox map. You just sprinkle them in everywhere and boom, you have a whole "land of adventure!" There's something really cool about how Breath of the Wild broke from the Zelda tradition and switched to having dozens and dozens of tiny shrines dotting the map instead of 6-10 medium-sized dungeons. This is also (sorta) the idea behind Barrowmaze, which is a really cool adventure that does the same concept.

My sessions are averaging around 5 hours at this point. A medium sized dungeon usually takes us 2 or 3 sessions to play through. But a micro dungeon takes us 1 or less. In fact, we've had sessions where we complete 2-3 micro dungeons all before the end of the night. And I mean the full cycle. Get the quest in town, do some research and prep, travel to the dungeon, clear it out, haul the treasure back to town, rinse and repeat, multiple times in one session. It feels unbelievably cool to do that. Obviously you can point out that micro dungeons have less creative potential than medium dungeons and are going to be a bit shallow (they especially can't do anything fancy with layout and Jaquaysing techniques and whatnot) but the tradeoff is an incredible sense of accomplishment after the session is over.

"We just did 3 quests in one night! Holy shit!" Even if it's the same total amount of content overall, it feels more satisfying in a way.

So yeah, I actually think micro dungeons are really good for when you want your campaign to be focused on dungeoncrawling as the main activity. Either your players are operating out of one hub town and they're slowly clearing out the dozens of shrines in the surrounding wilderness, gradually cleaning out the whole local region, or you have your players on the move throughout the kingdom and just doing tons of shrines along the way.

You can also easily make a procedural generation system for this kind of campaign. I use 24-mile hexes for my large scale and 3-mile hexes for my small scale. So you could have a procedure that says "every 24 mile hex has 1d4+1 micro dungeons in it." When the party enters, you roll 1d4+1 and then place each dungeon into its own 3 mile hex within the 24 mile area, then put one rumor for each dungeon in the main settlement of that 24 mile area.

Medium Dungeons

These are the majority of all dungeons that get designed. Most of the ones in The Lost Mines of Phandelver and other 5E books, the Sunless Citadel, Tomb of the Serpent Kings, Tomb of Black Sand, the Waking of Willowby Hall, most Dyson Logos dungeons, Hole in the Oak, I could go on and on. Unintuitively, most 1-page dungeons would actually be in this category, just because people are getting really good at packing lots of content into a small space.

How do you make good medium dungeons? What can you do with a medium dungeon? What are they good for? Well, to be honest, most advice people give on designing dungeons just assumes this size by default, so go ahead and apply most of what you know about dungeons to these ones. Try to make each room cool, have some variety, use interesting layout, maybe some competing monsters, a few secrets here and there, etc.

What I tend to find that medium dungeons are really good for, as weird as it may seem, is campaigns that don't focus on dungeoncrawling. Or rather, ones that only occasionally have dungeons. If you want a campaign that's close to 100% dungeoncrawling, then I think the other options on this list are better, but if you want about 1/4 or 1/5 of your sessions to be spent in dungeons then you should make them this size.

Like, the 5E campaign I've been a player in for the last few years I would say has had about 1/5th of its sessions be dungeoncrawls. Another 1/5 might be wilderness travel, another 1/5 is spent in settlements (talking to NPCs, investigating, shopping, etc.), and another 1/5 is just boss fights. Since high-level 5E has really long combats and it's very much an action game, we rarely have combats that aren't bosses. If the fight is gunna take us hours to do, then we might as well have something fucking awesome and complicated to deal with, and luckily my DM is really good at designing 5E combat scenarios that can keep you interested and thinking about your choices for 5 hours straight. It's just that you definitely wouldn't want that kind of thing more than 1/5th of the time. The last 1/5 I would say is just "miscellaneous" stuff, especially goofy roleplay sessions.

So if your dungeons are medium sized then you get about 2 or 3 sessions out of each one, and that feels like a good amount of dungeon if you know that it'll only be comprising a fraction of the total time. A megadungeon would be wasted if you only spend 1/5th of your time engaging in it and a micro dungeon would feel like a really insubstantial session when you had other adventures that were, like, 2-3 session-long arcs about mystery or travel.

Megadungeons, in two types

Megadungeons are just defined as "unusually large dungeons," but over time there's arisen some traditions and best practices in megadungeon design. If you've ever played a Metroidvania (Metroid, Castlevania, Hollow Knight, the first Dark Souls, etc.) then you can picture a megadungeon. It's so large that it's usually split up into regions that could each be their own medium-sized dungeon. This is your Rappan Athuk, Maze of the Blue Medusa, Caves of Chaos, Undermountain, Dead in Thay, etc. And megadungeons are absolutely optimized for dungeon-focused campaigns, because they offer so much to keep discovering. It would be a waste of a good megadungeon to only visit once or twice and then move on to other stuff.

Thing is, megadungeons were actually the default back when the hobby was created. In the Original D&D booklets, in the section where they define dungeons and how to design one, this is what they say [emphasis mine]:

In beginning a dungeon it is advisable to construct at least three levels at once, noting where stairs, trap doors (and chimneys) and slanting passages come out on lower levels, as well as the mouths of chutes and teleportation terminals. In doing the lowest level of such a set it is also necessary to leave space for the various methods of egress to still lower levels. A good dungeon will have no less than a dozen levels down, with offshoot levels in addition, and new levels under construction so that players will never grow tired of it. There is no real limit to the number of levels, nor is there any restriction on their size (other than the size of graph paper available). “Greyhawk Castle,” for example, has over a dozen levels in succession downwards, more than that number branching from these, and not less than two new levels under construction at any given time.

A dozen levels?? And this is the example drawing they offer:

So it's basically a given that a megadungeon is so large that either 1) it'll take well over 5 sessions to complete, more in the ballpark of a dozen or two or 2) it'll never be completed, since the incremental progress made with each delve can't be sustained every time the party is forced to retreat. The megadungeon will repopulate its vacant rooms with new monsters. There will definitely be competing factions in the dungeon, and they'll be pushing and pulling for control over the different regions. And the DM likely only has to design small chunks at a time, since they know "there's no way the party can get down to the 6th level in this session, so I can leave it blank for now and work on it later." The dungeon can also have multiple entrances so that the PCs can think about their approach and have a totally different experience coming from each direction.

Normally you'd want to go in the front entrance, but the drawbridge is controlled by the orcs, right? So you could go around the other side and head up the mountain face, scouring it for caves that might lead into the dungeon. But you can see that the trees of the mountainside are patrolled by goblin scouts. So maybe you decide to go to the base of the mountain, find the river, follow it back up to the mouth, and enter the lowest area of the dungeon through that cave.

You definitely want to Jaquay this fucker up, and probably adhere to some classic design traditions like having the depth correspond to difficulty of monster/amount of treasure, along with some "mythic underworld" traits like hostile architecture, self-shutting doors, changing layout, and monsters generation from thin air as a mere quality of being in such a hateful environment. It's like the dungeon itself is alive, it sees the PCs as an infectious disease, and the monsters and traps and mimics are its immune system.

Why do I say there are two types? Because dungeons at this scale take on a secondary quality that becomes important: density. As in, the density of rooms to each other can make a pretty big difference on the players' relationship to the dungeon. I don't know if I've ever seen anyone else comment on this before so I'm gunna lay out my thinking.

Dense Megadungeons

This is the more traditional kind and the one I bet most people reading this are thinking of. These megadungeons are still constructed as though they're one building. There may or may not be hallways between rooms, just as with all dungeons, but even if there are I bet they aren't super long. It may be a truly stupid number of rooms, like dozens or even hundreds of rooms, but overall they span an area that's still no bigger than, like, an airport or a school or something. For example, here's the layout of the Maze of the Blue Medusa:

Huge for a dungeon, but still just a single building.

The idea for these kinds of megadungeons is that they're probably the focus of the entire campaign. The basic cycle of gameplay is 1) assemble today's party of adventurers, 2) venture out to the local megadungeon and choose an entrance, 3) go in for a delve and plunder as much as you can, 4) return home with your haul. Every session you discover a little bit more and maybe even change the dungeon's situation a bit, but you never "complete" the dungeon. Plus, this cycle is basically fueling the local economy. The megadungeon is the biggest source of capital for the community and adventurers going in for treasure is an entire industry unto itself.

One thing that helps this is the "stuck doors" and "auto-close doors" mechanics, which go hand in hand. In old school D&D, part of the mythic underworld traits was 1) an assumption that all doors are stuck by default, and only have a 1-in-6 chance of being opened, and 2) doors automatically swing shut after you've passed through, unless you have a way of holding them open. Do these assumptions make sense? Not really, but that's why the explanation is "the dungeon itself hates you and wants to screw you over." The combination of these two traits actually creates a randomization effect, so that no two delves are exactly the same. It requires that there are lots of doors immediately available to the PCs and that no part of the dungeon ever becomes too linear, because all it takes is 1 stuck door to then create a dead-end. But if there are 6 doors right off the bat and only 1 of them is going to open today, then that alone immediately defines what parts of the megadungeon the PCs get to experience today.

I'm gunna show an example below with a megadungeon map I found on Google. This will be the same map shown 4 different times, each one representing a different delve into the dungeon. The red line shows the party's primary path from the entrance they picked to the exit they left through. Every time there's a blue line, it represents an attempt to take a path that failed because of the stuck doors mechanic. Every time there's a pink line, that represents a path they took that worked out for them but ultimately resulted in a dead-end, thus forcing them to backtrack to an earlier decision point and choosing another way. 

Click to enlarge

In the top left (our first delve), you can see that the party entered from the south and was immediately put through a series of randomization chambers. Each of those diamond-shaped rooms has 4 doors, and since every attempt to open one has a 5/6 chance of failing, it means that the party's path through the very first stretch is heavily randomized by which doors opened for them versus which ones stayed stuck today. But as you can see, they ended up exploring eastward and went through quite a few rooms before finding an exit to the north, which had another series of randomization chambers (each hexagon shaped, this time).

In the top right (our second delve), the party used the same entrance but ended up having to explore the totally opposite side of the dungeon because of where the randomization chambers led them. Their whole delve ended up being pretty short but they discovered a secret exit to the west, which is neat.

The bottom left (third delve) show them trying the same entrance, doing the usual running of the doors, and then exiting out into the same area of the dungeon as their first delve. But instead of exploring the same direction eastward, they went more north to see what the middle of the dungeon is like. They had a very frustrating experiencing going through that one hallway and trying every door on their right, only to keep finding them stuck. But eventually they explored through some northern halls that link up with the area they explored in their first delve, and they decided to head south and found another exit.

The bottom right (fourth delve) shows them trying out the secret entrance they discovered 2 delves ago. This time they got to explore more of the northwest and a bit of the same center-dungeon they explored last time, once again exiting through the north.

So what's neat about this is that you'll never even have the chance to exhaustively survey the dungeon. You'll gradually learn more and more about the layout each time but you can't rely on returning to the same rooms as before unless you know multiple routes to reach it. If you're playing at an open table campaign or a West Marches game or something, then you can tell other players about what you experienced and maybe give them a partial map of what you explored, but they have to understand that their own delve might look completely different.

Sparse Megadungeons

This is the rare one, and does not have nearly as much history of design advice backing it up. So imagine a dungeon that spans miles of area. It still may have the same total number of rooms as the examples from before, but the paths connecting them are, on average, much longer. If you've ever read or seen The Lord of the Rings, you may recall the part in The Fellowship of the Ring where the party is traveling through the Mines of Moria. Well it may not have been obvious in the movie, but they actually spend several days in there. It's literally a dwarven kingdom they're journeying through, even if it has the vibe of a dungeon the whole time.

So rather than having one big dangerous castle or tower or something, you have pockets of dungeon living beneath the surface spread throughout a region. Here's an example diagram of a sparse megadungeon (imagine it as a cross section):

So you have a city on the right, a mountain on the left, and some farmland in between. But underneath all of that is a series of dungeons that are all interconnected. In this way, it feels a bit more accurate to refer to the dungeon as "the Underworld" since it's literally just a layer directly beneath our world.

You can also see that I've marked which areas are under control of which factions to start out. Three large ones and one small one is pretty good. So wherever the players enter the dungeon from, they aren't just going to clear out that space and haul in all the riches. They'll be facing one piece of a larger network of allied forces. And while they could totally wipe their enemies out, they'll just as likely force a retreat to some other area. Plus, the factions are scheming against each other to expand their own territory in key areas, so the makeup of the megadungeon will change even without the players' involvement. Let's take a look at how this dungeon might transform after you throw some PCs at it.

Looks to me like the PCs entered the gold mines, drove the kobolds further up the mountain, and then descended into the magic vault to kill off the frost giants and grab some magic items. But the dwarves expanded into the castle ruins and the kobolds took the sewers away from the undead. Now let's look again after another bout.

The undead fought back and they massively expanded. It was inevitable that the vacuum in the magic vault would be filled if the players just left it behind. So the kobolds and the dwarves both tried claiming the gold mine and the dwarves won with the PCs' help. The kobolds were doing well for themselves before, but are now almost gone. Fun, right?

So you can see how the different "regions" of the megadungeon are emphasized a bit more? Claiming ownership over each one grants a clear advantage. Power over the burrows gives you a food supply directly above. The gold mines provides capital. The magic vault is full of artifacts. The castle ruins have lore and are fortified. The sewers gives you the best access to the city. And so on.

The scale also means that the concept of "dungeon factions" has a bit more room to breath. When you're talking about the factions in the Caves of Chaos and you say there's a "tribe of hobgoblins" or whatever, we're talking, like, 40 hobgoblins in total. They have the manpower to hold down less than 10 rooms. When different factions are fighting for control over a dense megadungeon, it feels like small squads of people holding down a few rooms at once, maybe being able to claim a wing of the building. Competing factions cannot, in my imagination, survive in such close proximity for any amount of time longer than, like, a few hours.

But with a sparse megadungeon, you can very easily imagine something like 100-200 members in some of these factions, spread across multiple regions and holding them down for days or weeks. For one faction to conquer a new region of the megadungeon would require mounting a pretty substantial invasion.

And as adventurers, discovering the true size of the dungeon can be kind of astonishing. You get paid to go down into the sewers in town and clean up some kobolds, right? But as you explored further and further, you found yourself suddenly in a network of caves, until finally you see some light up above, and...

...You pop out of the well in Farmer Jenkins's backyard? Way outside of town?

So if we return to the idea of the procedurally-generated dungeons as previously discussed, then imagine this: every 24 mile hex has 1d4+1 dungeon areas, but they're all connected across the region. When you zoom in to the 3 mile hex scale, you start sprinkling their entrances in the Overworld throughout the 24 mile area. The hub town, the forest, the swamp, the lake, and the ravine all have their own dungeons underneath them that are all part of a greater network.

I also like this scale because you can have a "room" of the dungeon be, like, a whole cliff face that needs to be climbed. Or an entire chasm in the earth. Or, I dunno, like a lake. And you can put things like Underworld settlements in as part of the megadungeon. A whole-ass goblin town is just part of the network as one of the regions. Or something nice and chunky like an arena.

Also, as I've ragged on about on this blog before, the standard assumption of "120 feet per 10 minute turn" is beyond absurd. You don't need to read this whole post if you don't want to (it's actually quite out of date) but if you skip to "Big Difference Number 1: Movement" then you can see me put this one to death. There is no combination of excuses or reasonings that can ever possibly justify such an insanely slow pace, even if you go with the interpretation that "it assumes players are being thorough and careful." It's actually difficult to travel at such a slow pace, like, physically. Try it. Try stretching a 120' walk out to 10 whole fucking minutes.

So instead, if you go with a reasonable pace (the most realistic being 3000 feet per 10 minutes but I'll accept something lower), then having a really wide dungeon helps you get more out of that. If you're walking from the sewers to the burrows to the gold mine then maybe it'll take you a few hours at a pace of 300ft/minute. If you did 12ft/minute it could take you days.


I don't really have any way to wrap this up. It was just something I was thinking about and I thought maybe others would find it interesting. Obviously the line between each broad size category is pretty blurry but I do find it interesting how important it can be in defining the entire dungeon experience.



  1. Fantastic stuff - so many thoughts to chew on. How many rooms do you think would make up a typical Medium dungeon? What's that middle range between micro-mega?

    1. Boy, that's a good question. Especially since, despite what OD&D says, I'm pretty sure you hit megadungeon status well before you've reached 12 floors. And I've played plenty of dungeons that I thought of as "medium sized" but still were divided into multiple "regions" like a megadungeon (Tomb of the Serpent Kings, for example).

      If a micro dungeon is 1-6 rooms then maybe we can say a medium dungeon is 7-50ish. But of course, it also depends on how one labels the "rooms" their map. Hole in the Oak definitely isn't a megadungeon but it's got more than 50 labeled rooms. But then of course you also have dungeons that can get a lot of content out of very few rooms, especially if their puzzles require lots of backtracking through the same rooms again and again.

      I guess that's kind of why I think it makes more sense to measure in "number of sessions it takes to play," but that's really tricky to estimate without actually playing it. Not to mention differences in session length and how "productive" your group is.

  2. Regarding the stuck doors -- can't the players just break them down or force them open? Or is a door stuck just magically unpassable no matter what?

    1. Actually, you WOULD be forcing them open using an "Open Doors" roll, which used to be a thing in early editions and was one of the main uses for your Strength. But if you failed on that roll, then presumably that indicates that your character wasn't strong enough to open that door, and so you wouldn't be able to re-roll.

      But also, yes, most of the justifications here are due to magic bullshit. That's the notion of the "Mythic Underworld" that I alluded to. The all-encompassing explanation given for the quirks of early D&D dungeon rules is that, "the dungeon is a living thing that hates the players and will constantly shift in order to screw them over."

      Not saying I LIKE this style of doing dungeons, but yes. That is the logic behind the rules, which is why they seem a bit nonsensical.

  3. Also, one more Q: does gold get spread out in such dungeons? I understand you have a certain amount per floor -- but it's quite a different story if the floor is 5 rooms VS 50 rooms. Shouldn't the treasure be less in the former and more in the latter?

    1. Actually, the amount of treasure per floor would be randomly rolled if you're following the guidelines in the book. For example, in B/X D&D, the third floor of a dungeon would have a base 1d12 × 100sp. After that, there's a 50% chance of there being 1d6 × 100gp, then a 10% chance of there being 1d6 gems, then a 5% chance of there being 1d6 pieces of jewellery, and lastly an 8% chance of there being 1 magic item.

      So no, treasure isn't proportional to room count. It varies dramatically based on the dice. Again, not saying I LIKE this style, but those were the rules.

  4. I'd like to hear more about the sparse mega dungeons. It's how I've always imagined the underdark feels. What games/articles best encapsulate this? Any play experience with a dungeon like the one you described?