DragnaCarta
patreon
Dragna's Blog: Method Dungeons
🕑 Added 2024-09-08 16:55:24 +0000 UTCIn my home game, I’m currently running my players through (a heavily remixed version of) A Deep & Creeping Darkness from Candlekeep Mysteries. For those who haven’t read it, the adventure is about investigating an abandoned mine and eliminating the nest of meenlocks that still haunts it.
As I prepared for an upcoming session, I took note that my players had reached the mine and were about to enter it. The question became: What should this dungeon look like?
I didn’t mean this question literally, though - more so narratively. Those who have followed my work on the Patreon Discord server know that I’ve recently (over the past month or so) made great use of what I’ve been calling “goal diagrams”—flowcharts (usually created using an Obsidian canvas) that model each dramatic question’s inciting incident and goal, as well as any subsequent obstacles. This allows an easy visualization of the players’ current dramatic question(s), as well as whether the current dramatic question provides gameplay or (as is common) requires an additional obstacle to create that gameplay.
In this particular adventure, my then-current goal diagram for the mine looked something like this:
“Investigate the mine” would provide gameplay via a method choice—an open-ended scenario that would be resolved only when the players achieved the goal (similar to a combat encounter). So how would this look in practice? To answer this question, we must first ask: Why is it called Dungeons & Dragons?
Choice & Challenge
“Agency” or “gameplay” has two components: apparent player control of outcomes (“choice”) and actual player control of outcomes (“challenge”). There are three kinds of choices: action (whether or not to do X), option (whether to do X or Y), and method (how to do X). Of these, method choices provide the deepest gameplay, because they allow the players to make creative use of a broad portfolio of gameplay verbs (e.g., cast fireball, climb a cliff, recall information about something) without requiring the DM to create a preordained set or sequence of verbs that the players must use to progress.
The core aspect of a method choice is emergent complexity: a simple scenario that can produce countless sequences and outcomes depending on the verbs chosen by the players. (A single combat encounter illustrates the concept quite well—can you ever really say that two playgroups’ approach to the same fight are exactly the same, especially when each player has full power to customize their character, movement, and tactics?)
Meanwhile, there are three kinds of challenges: solution (the players achieve X or don’t), path (the players achieve X by optimal path Y instead of the default path Z), and outcome (the players achieve X with varying non-discrete levels of success). Of these, outcome challenges provide the deepest gameplay outputs, because they reward or punish players for every discrete choice made (e.g., which square to move to, which enemy to attack, or which spell to cast).
Significantly, while you can have a method choice without an outcome challenge (e.g., “How do you want to get past this locked and barred door?”), you can’t have an outcome challenge without a method choice; outcome challenges represent the emergent complexity of the possible outcomes of a method choice. In an outcome challenge, neither the DM or players can predict exactly how many hit points or spell slots, which conditions (if any), or which knowledge or relationships the players might emerge with after the gameplay has concluded.
The deepest gameplay, then, emerges from a method choice paired with an outcome challenge: an open-ended “How do you want to do this?”, where the players’ specific level of success is unknowable upfront and near-completely dependent on the choices they make during gameplay.
But how does this relate to Dungeons & Dragons?
Method Dungeons
DMing blogger The Alexandrian has previously written about the importance of nonlinear dungeons. Here’s how he describes a particular dungeon created by designer Jennell Jacquays:
[I]n Caverns of Thracia Jaquays includes three separate entrances to the first level of the dungeon. And from Level 1 of the dungeon you will find two conventional paths and no less than eight unconventional or secret paths leading down to the lower levels. (And Level 2 is where things start getting really interesting.)
The result is a fantastically complex and dynamic environment: You can literally run dozens of groups through this module and every one of them will have a fresh and unique experience.
But there’s more value here than just recycling an old module: That same dynamic flexibility which allows multiple groups to have unique experiences also allows each individual group to chart their own course. In other words, it’s not just random chance that’s resulting in different groups having different experiences: Each group is actively making the dungeon their own. They can retreat, circle around, rush ahead, go back over old ground, poke around, sneak through, interrogate the locals for secret routes… The possibilities are endless because the environment isn’t forcing them along a pre-designed path. And throughout it all, the players are experiencing the thrill of truly exploring the dungeon complex.
Sound familiar? The Alexandrian is describing a dungeon whose nonlinearity itself creates a method choice and outcome challenge. This is (one) secret to why dungeons have remained a cornerstone of D&D gameplay: like combat, nonlinear dungeons are an easy, drag-and-drop route to meaningful gameplay that require no game-designer knowledge from their creator.
Not convinced? Let’s take a look at the introductory dungeon to Lost Mines of Phandelver: Cragmaw Hideout.
Despite being the literal first dungeon the players encounter, this cavern is heavily nonlinear. Here’s what it looks like if we break it down into a basic flowchart:
Just from this basic structure alone, we can see how many simple option choices combine to product a single, complex method choice:
Do the players investigate the Goblin Blind before entering the Cave Mouth?
Do the players head left from the Steep Passage (into the Goblin Den), straight (beneath the Overpass to Twin Pools Cave), or right (to the Kennel)?
If they enter the Kennel, do they return to the Steep Passage or try to ascend to Klarg’s Cave?
Once they reach Twin Pools Cave, do they head west (into the Goblin Den) or southeast (into Klarg’s Cave)?
Significantly, all of these choices have meaningful tradeoffs! Checking the Goblin Blind before ascending to the Steep Passage is important because, if you fail to do so, the goblins will ambush and attack you anyway, and you’ll wind up on the back foot. Ascending the chimney from the Kennel into Klarg’s Cave is actually a bad move, because the prisoner you’re looking for isn’t there, and you lose all leverage with the goblins in the Goblin Den (who want Klarg dead in exchange for the prisoner’s freedom).
And heck—players who head from the Steep Passage straight into the Goblin Den could conceivably turn the (goblin) combat encounters in Twin Pools Cave and the Overpass into social encounters because the goblin boss in the Den wants the bugbear in Klarg’s Cave dead, and players with the boss’s blessing could conceivably convince goblins elsewhere in the cave to stand down! (The boss might even tell the players, if the right questions are asked, that the chimney in the Kennel leads to Klarg’s Cave, which lets clever parties ambush Klarg by climbing up from an unlikely angle.)
This is incredibly deep gameplay. By contrast, imagine how this might look if we removed some of these loops:
Sure—we’ve still got some action choices (investigate the Blind/Kennel?) and an option choice (investigate the Goblin Den or Klarg’s Cave?). But by eliminating just two nonlinear paths—one connecting the Kennel to Klarg’s Cave and one connecting the Steep Passage to the Goblin Den—we’ve eliminated all possible emergent complexity from the adventure. The DM can easily see every possible outcome from every possible choice. At most, we’ve got the following permutations:
Do the players investigate the Goblin Blind or ignore it?
Do the players investigate the Kennel or ignore it?
Do the players investigate the Goblin Den first or Klarg’s Cave first?
None of these choices really matter except for the last one—and even then, the only possible outcomes of the last one are, “The players talk to the goblin boss first,” or “The players talk to Klarg first.” There’s no deeper complexity that informs how the players traverse and negotiate the rest of the dungeon; it all comes down to a coin flip.
Now, Wizards of the Coast is doing something interesting here—by having the goblin boss demand Klarg’s death before handing over the prisoner, the designers are forcing/encouraging the players to 100% complete the dungeon (i.e., explore basically every room) before they can exit, just like a more linear dungeon would.
However, unlike a linear dungeon, the loops still create emergent complexity, such that every possible run has at least the potential to be unique by virtue of the players’ gameplay choices. To quote The Alexandrian again:
In a linear dungeon, the pseudo-choices the PCs make will lead them along a pre-designed, railroad-like route. In a [nonlinear] dungeon, on the other hand, the choices the PCs make will have a meaningful impact on how the adventure plays out, but the actual running of the adventure isn’t more complex as a result.
Worldbuilding is Hard
So how can we create a nonlinear dungeon? We begin with Newnist design.
As I’ve previously written, “Newnism” is a term I coined that basically means “inspiration-driven design.” We’re not worried about immersion or verisimilitude, and we’re definitely not concerned about narrative design or gameplay. Instead, “Newnist” design just asks: “What do I, as someone with specific aesthetic and/or dramatic preferences, want to see in this narrative?”
One of the interesting things about working with a module (or setting sourcebook) is that it’s already done the Newnist work for you; if you’re someone like me, who lacks a muse and hates the act of pure creation, then modules are great! In this case, since I was working from an existing module, I already knew what the Newnist design would look like: an old, abandoned platinum mine with Feywild influences infested by a nest of fear-loving meenlocks. I had already jotted down a few other ideas during previous encounters (e.g., a kind of black mold that can make you hallucinate or cause fear), and the book also supplies some additional information (e.g., a group of miners was trapped down here after a cave-in, then subsequently rescued), so I noted them here as well.
The next step of the creation process: Watsonian design: “What makes sense?” We know there’s an old, abandoned mine with meenlocks lairing in it; what does that look like? I could have used the map in the book, but it felt unsatisfactory to me—too small for a proper mining operation (though, to its credit, some parts of it were nonlinear, providing a confined method choice). But I don’t know anything about mining, let alone D&D mines. What was I to do?
One of the secrets about using LLMs/AI like ChatGPT or Claude is that they’re absolutely worthless when you need them to convey novel information (e.g., “Write this blog post for me”), but that they’re godsends when you need to obtain or synthesize words that convey no new information. While nobody should accuse them of being proper search engines (or even enyclopedias), LLMs are exceptionally good when you need someone with approximate knowledge of many things. And when it comes to figuring out the layout of a hypothetical D&D platinum mine, that’s exactly the kind of task where LLMs shine.
So I asked Claude to play the role of Watson for me: to give me a list of rooms that might be found in a platinum mine in the Middle Ages. Then, I asked it to present those rooms as a flowchart/diagram depicting how they might connect, then asked it to make that diagram nonlinear, with lots more loops and hidden routes. To finish it off, I asked it to give me brief descriptions for each area, just so I could get a sense of what I was working with. (This overall process was more of a conversation than a series of instructions - I’d often tell it to tweak its output, add/remove stuff, or redirect it down another path.)
After about ten minutes of chatting, I wound up with something like this:
This was much better. I did, however, still need an actual, navigable map.
I started by asking Claude for the length and width of each room in the dungeon. From there, I placed each room down as boxes in Canva, sizing each “room” appropriately proportionately to its size by specifying the dimensions in pixels. As I went, I frequently sent Claude screenshots of my work and asked if I should rearrange what I had—for example, it helped me figure out how to arrange the Mine Entrance, Guard Post, Foreman’s Office, Equipment Storage, and Main Tunnel in a realistic way, while still preserving the connections and loops that I wanted.
After some additional brainstorming and a good amount of back-and-forth, I wound up with this map (blue boxes are at ground level, red boxes are below ground level; purple boxes are the meenlock nest; the green box is a house in the village nearby):
It was heavily nonlinear, with a substantial number of rooms (and, eventually, assorted encounters and secrets to find). Throughout this process, of course, was woven the Doylist process of game design: What do I, as the game designer, need for the players to encounter, and in which order? (For example, even though I wound up with two possible entrances to the meenlock nest, I still knew that it needed to be the last thing that the players entered.
Finally, once I had the full dungeon structure, I continued the conversation with Claude to figure out what each room might contain (e.g., what kinds of tools might be found in the Tool Repository?). I also added in a few additional encounters of my own to add gameplay (e.g., a meenlock ambush halfway down the right branch; some violet fungi in the large cavern; etc.), plus some discoveries (e.g., a miner’s journal documenting the meenlock attacks following the cave-in).
Total time to design the entire dungeon from scratch (including encounters, discoveries, etc.): approximately four hours. The time required to create the basic architecture/layout was even less—perhaps an hour (and most of that time was spent drawing it out in Canva, rather than actually doing “design work”). And that was, again, working from scratch, in a situation where I knew nothing about the dungeon’s potential contents.
By sheer coincidence, approximately a week after I ran the first half of this dungeon, I began outlining my plans for Berez in Reloaded. After going through an annoyingly long process of write/revise/throw out/repeat (which I’ll detail in a future blog post), I settled on a simple concept: The players needed to navigate to Baba’s hut, where they had to fight her. To prevent them from doing so, Baba would use mirage arcane to protect her hut with a giant thicket maze, which turned the open-plan Berez into a closed-plan dungeon with discrete “rooms.”
Here's what I wound up with, after spending around two minutes connecting locations on a printed version of the Berez map:
Is my work done? No. Is the annoying part done? Yes. Do I now have the framework for a satisfying, method choice-driven, nonlinear dungeon?
Absolutely.
