Mechanics, Clear the Way!

Why D&D needs procedures, not storytellers.

Narrative DMs don’t exist in the d20 space; referees run open games or just do whatever they want.

Don’t you wish the story vs mechanics discussion would die? WhickTV invited me for a panel discussion on Sandbox vs Narrative D&D, where Harmony Ginger and I put one more nail in the coffin:

I prepped quite a bit for this debate (thanks again to Harmony for being excellent and Finneas for coaching me) and there’s still plenty to discuss. For the sake of collecting my thoughts and not wanting to retread this discussion, I’ll delineate open styles of play¹ from Authored Branching Narrative and further distinguish ABNs from railroads. This essay isn’t a comprehensive positions page, it just follows from my priors. If you disagree with one of those priors (or just want me to expound on them in the future), let me know in the comments and I will consider it for a future post.

Authored Branching Narrative

Authored branching narrative is a style of play which prioritizes high player freedom, but not high player agency. All encounters are written by the author. Whether the encounter shows up in game, player options for interacting with the encounter, and even when the encounter appears are all handled by preconditions². My go-to explanation was taken from this talk by Jonathan Ingold below:

Because all content is authored, the designers are responsible for a ton of logic to ensure each encounter and accompanying player responses always make sense. Every pre-written encounter inflicts debt on any reasonably connected encounter because it shows the author is paying attention to your choices.

Writing content is a piece of cake in ABNs because content is all atomized anyways. A computer will easily save every line of description you can imagine. Think of a rude thing to say to an orc; the designer now has this line. They don’t need to know where exactly this goes into the game, it just goes into a bucket of content the designer might use later on. Why stop myself if I know the player might interact with orcs at some point? Invent a problem the player can solve within the same dialogue tree. Into a bucket it goes! NPCs, weapons, you name it. When Jon imagines a fun description or joke, he writes it down. He doesn’t need to plan out every appearance or reference in advance, this is a task for later. Write as much content as you want, you will need it later on.

Connecting dialogue/encounters is the heavy lifting of ABN development. I’ve written an NPC named Farmer Bob, a puzzle called the Lantern Puzzle, and a village where my players will encounter both. Is the Farmer Bob encounter any different before or after solving the Lantern Puzzle nearby? Maybe Farmer Bob gives you a solution to the Lantern Puzzle! Better yet, Farmer Bob might be looking for something trapped or hidden behind the Lantern Puzzle! If my players solve the Lantern Puzzle before they meet Farmer Bob, they should have different lines to account for this new context. Everyone’s eyes are peeled for these little love notes from the author. Unique dialogue interactions are moments of recognition from the author without breaking the fourth wall, a fairly novel experience. The conceit of authored branching narrative is a game that doesn’t just respond to player choices; player choices create narrative because the game is designed to cope.

Fragile decision trees are an unfortunate fail-state for ABNs and the reason they require so much logic to cope with player decisions. Say a player walks into a room with three doors and a symbol in the center. What are their options? Investigate the symbol. Walk through door 1, 2, or 3. Easy enough, but what if the player already knows what the symbol is? Could seeing the symbol cause the player trouble? Does the symbol invoke madness? We’ll give the player a choice to avoid this symbol based on a precondition. Wait, the player turned the lights off earlier! There are no torches, the room is dark and this symbol must be obscured. We take away any options related to the symbol using another precondition. Maybe the player caused a cave-in earlier, blocking off some of the doors leading out. Now they’re back to investigating the room, only it’s dark and the button to light a torch sits begging to be pressed in full view.

Anyone can see the problem here, we’re funneling the player towards a dead end. Remember, each question implies a state or precondition which then modifies our player’s available responses. They also pose more opportunities for failure! I’ve created a room where a swarthy, trenchcoat-wearing GAME OVER is more likely to flash our player the more we respond to their decisions! Our trees need to let the player navigate a highly contextual environment without a real chance to strategize³. That’s a fancy way of saying it takes a ton of work to make this room playable.

Grand Campaigns

I’ll now argue that D&D has the opposite problem. Game content in tabletop rpgs is actually quite hard! Even if you don’t decide to author every single monster ad-hoc, you still have thousands of monsters available for D&D alone. What’s the perfect monster for your table? Congratulations, you have a new problem! There are literally hundreds of variables you can track in a single adventuring day; will you independently validate each and every monster, player, and game statistic to distill the perfect encounter?

AD&D declares this a waste of time and elects to use “random” generators instead. Monsters in the dungeon are grouped by their associated dungeon level. Each level can hold monsters from a range of other levels. Every table has a wide range of monster types, which themselves range in difficulty but rarely mean instant death. AD&D generates high-variance encounters out of the box because doing so by hand is prohibitively complex and time consuming. It would take hours to produce this same variance by browsing monster books and selecting what you liked. You’re more likely to ditch the variance and produce a more boring set of encounters to save time.

Rooms (corridors, chambers, etc) themselves follow a similar pattern; they mostly adhere to standard sizes and shapes but can also generate odd environments. This pattern repeats in room contents. A slight majority of rooms are first generated empty (random encounters can populate these later). Monsters make up the remaining majority of results, some of which are guarding treasure. Treasure can also come with traps or be hidden. Every seemingly empty room could be holding hidden treasure! Besides tricks and traps, we have one last interesting result having rolled an 18:

Special* or contains stairway…

*Determine by balance of level or put in what you desire; otherwise put in stairs as indicated.

The implication of this d20 table result is subtle; your hand-written rooms are sufficient to fill maybe 5% of a full dungeon. Referees will run out of ideas well before they run out of rooms, so AD&D anticipates and copes by creating them for you. A ref who can invent 5 rooms by their lonesome can actually run a 100 room dungeon (so long as they don’t quit at 5 rooms)⁴.

A sufficiently busy adventuring day (given a variety of monster types, number appearing, difficulty, and so on) will satisfy players just fine. Healthy games don’t need encounters specially tuned to each individual statistic and player circumstance; a day of lots of varied encounters will more or less trip over these statistics in a fraction of the time. So what is the referee for, exactly? Running monsters, sure. Rolling on tables, got it! Is there anything for a referee to do besides read the rules to me? Is the referee necessary? Constant tirades insisting the referee would create more exciting quests and NPCs if they only felt more comfortable breaking the rules cast a dim light on the matter⁵. An answer came to me shortly after I found the perfect twitter argument to prod my thoughts:

From The Basic Expert:

Sometimes, I want to publish more modules and dungeon adventures, but the thing that stops me is the idea of restocking the dungeon. That's a big deal to me now as a referee. Sure, I can put gnolls in a room in the dungeon, but what will be in that room when the players come back? I hope Refs are versed enough to know to restock but I don't know if many are. Do you restock your dungeons?

What I mean by this is that using Appendix A in 1e and zero prep, I don't see the point in modules. That generator has restocking built into it with its room content table.

From Evheil:

Oh really now? What Appendix is the In world lore reason to roll for why your randomly rolled assortment of monster is there? It's like saying you like movies and books where a bunch of random stuff happens over a well writen story.

The Basic Expert’s Reply:

Why do you act like making sense of the rolls is impossible? I'm mediocre at best as a Ref and GM and do this all the time just fine. It's fun.

Evheil’s Retorr:

Sorry for not being convinced that if you need random rolls for monsters, you're going to make compelling reasons for them to be there on the fly... As apposed to someone that planned everything out before hand, in a module. Hell ask A.I. to write it then rather then tables.

Portions of my response are repeated elsewhere in this essay but here’s the full quote:

Tables include the full range of the game's design, browsing through various monster manuals for something you feel like running does not. You're not engaged in some brilliant level design, just window shopping. Independently validating each and every encounter statistic takes a ton of time because these games are mechanically dense. There are hundreds of variables you *could* track in a single adventuring day. To explain why an ooze is in the dungeon, I just need one lie. Literally one context clue is enough, it's the difference between writing up a character backstory and reacting to an encounter in-session. Players justify their reactions/ presence every single encounter despite not knowing everything behind the screen.

There’s the issue. Random tables accidentally made encounter design seem trivial. It’s so easy, a table can do it! Referees might aspire to be even better at encounter design as time goes on, surely this requires a hands-on approach. Nonsense; the tables are calibrated to meet a great number of needs at once.

What the game can’t do is connect encounters together; lucky us, we have a game master. Context builds over the course of a session as players and refs alike invent new details about the game. NPCs gain allegiances, monsters wield and guard magic items, favors are traded between factions. Each small improvisation adds to a library of references; flesh out characters by giving them the same livery or last name, attach random encounters to local bad guys by making them mercenaries.

The ref rolls dice and the game spits out encounters. All the ref has to do is justify the rolls and the party’s actions post-hoc.

Game’s Up

Some referees swear by hand-authored encounters in their games. Monsters are meaningful in combat. Characters have fleshed out personalities ready to dazzle players with their words. Recall the development process of authored branching narrative. Now, think back to a more railroaded campaign of your own; did it seem like your referee was developing an ABN? Did the game expand the more you tested boundaries? Did unorthodox play give you more unique options to progress?

Of course not! Railroad campaigns only account for a few variables, a few methods of approaching encounters at most. Every monster or stuck door is a bottleneck designed to funnel you to the next encounter in sequence.

Why Not Both?

You’ve just read some thousand words or so of claims; so-called authored D&D campaigns don’t really exist, authored branching narrative requires logic to connect encounters whereas a referee only needs to reference their contextual environment, etc. Nobody would blame you for wondering, “What’s so bad about pre-authored encounters if referees are so naturally good at connecting random encounters?”

If you’ve been paying attention you know our encounters check for states to determine how the players may interact. Encounters don’t just connect with one another; they edit future encounters. Authored content is fragile. Maybe one pre-written encounter won’t fall apart thanks to the previous encounter’s outcome, sure. How about 3? How about 6, the number of encounters D&D 5E recommends for a single adventuring day?

Let’s cap this essay off with an example. You decide Farmer Bob lost the key to his shed while walking near the Lantern Puzzle. The Lantern Puzzle is built into a hill side, he was on top of the hill, wouldn’t you know it Bob fumbles his keys and they slip through some grate or vent running into the puzzle dungeon. Farmer Bob will let the players take supplies from his shed if they get the key…but unbeknownst to the party, this is a trap! The Cult of the Reptile God is actually waiting below the shed’s floorboards for new bodies. Your plan is in motion; your players will travel to the Lantern Puzzle, fight some monsters to soften them up a bit, collect some cool treasure they’ll need to reacquire later, retrieve the key, go back to Farmer Bob’s shed, and right when they would normally turn in for the night…BAM! They’ll get clubbed over the head, PCs won’t put up much of a fight after a full day’s adventure.

All’s well until the Thief pipes up. “Can I just pick the lock?” The DM panics and says Farmer Bob doesn’t just want his shed unlocked, he needs the key back to keep his shed safe. In my view the DM has yet to overstep⁶. One player asks to perform an insight check to see if Farmer Bob is lying. The DM shuts it down and tells the players to get on with it, the guy wants his key back. Pod people attack the party on their way to the Lantern Puzzle and things go poorly. Everyone heads back having not engaged with the puzzle to interrogate Farmer Bob for new leads. Now the DM fumbles for excuses in their head. The party might rest at Farmer Bob’s, but the key is missing. Will the cult capture them anyways? Bust out of the shed? That doesn’t seem right, the ambush is below the shed’s floorboards but if the key is retrieved tomorrow, the players won’t be battered enough to rest at Farmer Bob’s!

Our DM’s bold gamble is to let the party rest and hope other random encounters bruise them enough to stay put tomorrow⁷. Here’s where everything starts to go sideways:

1. Farmer Bob is interrogated by the Bard on these pod people. There’s no shutting down an insight check this time, the party knows something is up.

2. The shed’s lock is simple and everyone’s on edge, so the Fighter asks the Artificer to cut a new key. Farmer Bob gets a new key, the party collects supplies and leaves at first light.

3. Speaking of simple locks, the Thief plies their trade and moves to open the shed.

Everyone might as well be sabotaging the DM on purpose! Farmer Bob doesn’t know everything so the DM can stall the interrogation. What of the Fighter, Artificer, and Thief? Now the players are planning to leave, they’ll escape the trap! Even worse, one player has defected and opened the shed! The DM has to fudge again; the thief either gets captured alone, doesn’t find the cult’s secret entrance, or finds it but with no cultists. A sprung trap is better than no trap, even though important loot will be left behind in the Lantern Puzzle. Combat begins and the Thief miraculously escapes, bringing the party back! Hours of prepped encounters are seconds away from returning to their proper sequence! Just kidding, combat ends with the players victorious. They’re barely victorious; this breaks the DM. Another party of cultists now arrives from Farmer Bob’s house, rushing to knock the players unconscious on the orders of a foul-tempered creature playing god. No need to expand the example, you know it’s a farce.

Any of the party’s actions above could have snowballed into an unforgettable night, if they were allowed. Put yourself in the DM’s shoes. What are you not doing, DM? You’re not connecting encounters anymore. You’re not adhering to the game’s mechanical structure. You’re not testing states and preconditions to see where the players will go.

Instead, you’re chiseling away at a pre-written encounter and your party’s choices alike to make them fit together⁸. There we have it; I’ve wondered where to draw this line for a few years myself and at the risk of sounding vain, I’ve just clearly delineated free-form gaming from 50 years of railroads.

1. Braunsteins, Grand Campaigns, and West Marches.

2. (or lack thereof)

3. See Jon Ingold’s definition of strategy.

4. Puts 5 room dungeons in a new light, in my opinion.

5. I could never completely abandon the idea of referees.

6. Antagonists should be free to trick or trap the players through conversation and I would never regard this as cheating; at the very least I wouldn’t compare it to fudging dice.

7. Remember, the example is a DM trying to adhere to the rules and mechanical structure while preserving the integrity of their authored encounters.

8. I could add they’re aiming for a specific sequence of events, no sense in getting too granular.