DragnaCarta

Patreon DevBlog #9: Back in Business

Added 2023-04-01 02:36:58 +0000 UTC

Thank you to all of you for your patience over these past few weeks—I know it’s been a bit of a dry spell as far as these articles are concerned, but hopefully the stuff I’ve got planned will more than make up for it!

This week, we’ve got some deliberations over Curse of Strahd Re-Reloaded, plus my process for turning long-term journeys into meaningful narrative segments. We’ll wrap it up with a new ChatGPT prompt I’ve been using to plan, prep, and manage DMing (using GPT-4) that I think’ll really knock your socks off :)

Problem #1: How Do I Leave Clues for the Missing Bones?

In the original Curse of Strahd module, the St. Andral’s Feast questline asks the players to solve the mystery of the missing bones of St. Andral: who stole them, and where have they been taken?

There is, however, a fundamental problem with the “mystery” part of this quest: There’s nothing meaningful for the players to do. Once Father Lucian gives them their task, he also immediately tells them the most likely culprit (Miliovj, the groundskeeper). Once the players challenge Milivoj, he almost immediately crumples and admits to stealing the bones (requiring only a DC 10 Intimidation check), and tells the players where to find them.

This isn’t a mystery. This is a point-and-click walking simulator. That needs to change.

Let’s start with the first problem: Father Lucian’s mysterious intuition that Milivoj is the one who stole the bones. How does he know this? The book only notes that Lucian asked Yeska, the altar boy, whether he told anyone about the bones, and that Yeska nodded, but failed to disclose why.

Putting aside the ridiculous fact that Father Lucian knows that Yeska knows who stole the bones, but hasn’t bothered to actually meaningfully interrogate the kid—despite the very real danger that the missing bones pose to the church—there’s still no real reason for Lucian to suspect that Milivoj stole the bones at all! Milivoj has no motive, and there’s no real evidence connecting him to the crime—at least, none that Lucian, a simple priest, would know to look for.

So let’s start by trashing Lucian’s suspicions. Let’s also trash the idea that Lucian knows that Yeska told someone about the bones. Lucian is a kind, gentle, somewhat naive man, and he never expected Yeska to tell anyone, nor would he ever expect Milivoj - his faithful employee! - to steal the bones either.

Now we’ve got an actual mystery: Who stole the bones? We don’t know! However, that means it’s time to give the players the tools they need to actually solve this mystery.

Some of you in the Discord may have heard my growing belief that D&D 5e doesn’t have three “pillars of gameplay”—combat, socializing, and exploration—but five: combat encounters, social encounters, puzzles, obstacle courses, and skill challenges. (Why? A "pillar of gameplay" is any kind of gameplay that presents a challenge that the players can overcome with meaningful skill and/or luck over an iterated number of "rounds," with an opportunity to adjust, adapt, and react.)

This mystery isn’t a combat encounter, social encounter, obstacle course, or skill challenge, which makes it a puzzle—gameplay that is solvable through logic.

How do we assemble the gameplay for a puzzle? A puzzle is, put simply, lots of tiny, easy tasks whose outcomes, when added together, automatically solve the puzzle. In the rawest sense, a “puzzle” is like a logic game or mathematical proof—you need to go step-by-step, taking every incremental step, until all of the pieces fall into place. Because we’re presenting this to D&D players, rather than PhD students, each step needs to be rather easy to recognize, and rather easy to complete—no strokes of genius required. (The act of putting the pieces together is itself enough to make the players feel rewarded for their efforts.)

To make this work, we’re going to need to implement the three-clue rule, seeding as many clues as reasonably possible to help our players figure out Whodunnit. There are three places that we can put clues: (1) Yeska, (2) the entrance to the crypt, and (3) the crypt itself. Yeska was easy enough—just let the PCs interview him on their own, and have him admit that he told someone (Milivoj) about the bones if they roll high enough.

The other two bits—the physical clues—were a bit more challenging. I could have spent an hour or so brainstorming them myself, but I’m a lazy person who never has enough time. So, as I’ve been doing a lot lately, I turned to my Co-DM—AI.

Now, at the time I worked on this, GPT-4 was not yet publicly available through ChatGPT, so I couldn’t use my Council of Adventure prompt (see below) to workshop this. As such, I had to do a lot more independent bootstrapping to make this work. What I wound up doing went like this: I told the AI what I was trying to do (make some clues for the PCs to find), told it the story of the theft (including the culprit and methods used), and asked it to generate a bunch of clues that the PCs might be able to find.

It took a bit of back-and-forth workshopping (I wasn’t happy with everything that the bot gave me), but I got a lot of good ideas from this process, and managed to sculpt it into a relatively robust series of clues that I’m, honestly, pretty happy with! You can find the result of this process in the St. Andral’s Feast portion of the draft Re-Reloaded guide.

As a side-note: I think that lots of folks are sleeping on the fact that generative AI allows time-crunched DMs to produce lots of extraneous material. What do I mean by that? In an ordinary campaign—especially a homebrew one—DMs are fairly leery of spending time brainstorming and preparing content that their players might fail to appreciate, or might just outright miss.

For example, when creating this mystery, I worked with the AI to produce a solid 6-8 clues that the PCs could use to solve it, many of which were redundant. For example, the PCs could, technically, crack the whole case just by talking to Yeska. They could easily put two and two together the second they notice some of the more obvious clues in and around the crypt, and ignore the rest.

To a time-strapped DM, that feels bad! As a result, lots of DMs wind up limiting their time to the creation of 2-4 clues, and then do their best to make those clues as obvious and recognizable as possible to make sure that their players find them. This is good for resilience and saving time, but it also makes the world feel a little bit smaller and a little bit linear.

With generative AI, though, it doesn’t feel bad when you ask the chatbot to make twelve clues for you and the players only find three! The time you spent making those clues was absolutely minimal. I suspect the same will be true for adventure arcs and dungeon chambers in general—rather than artisanally creating each room and NPC with loving craftsmanship, we can now let the AI do that, and focus our time instead on curating and organizing the content it creates into a much larger, breathing, living world and narrative.

Will this work? I have no idea! But I’m excited to try it.

Problem #2: How Do I Make the Stops Along a Journey Meaningful?

Recently, I was chatting with a patron about reworking the Hoard of the Dragon Queen module to make the “journey” portions of it more meaningful. Here’s how it went down.

If you don’t already know, Hoard is an archetypal “grand fantasy voyage”—the PCs learn that the Cult of the Dragon has stolen a bunch of treasure, and must pursue them across half of the Sword Coast to learn where they’re taking it and what they’re using it for. The biggest chunk of in-game time is spent following the cultists from their camp near the town of Greenest, then north to the blessed city of Elturel, then west to Baldur’s Gate, then north again to the city of Waterdeep. From there, the PCs have to hire a caravan that will escort them north in pursuit of the cultists, a journey that finally ends at a roadway warehouse that has a secret tunnel connected to an old ruined castle in a swamp.

That’s a lot of travel! Except, aside from some random encounters and mini-arcs that unfold on the road, not much happens in the cities themselves. In Elturel, the PCs meet with a faction NPC who “officially” gives them the quest to hunt down the cult, but aside from some recreational downtime spent socializing with this NPC, there’s not really much else to do there. In Baldur’s Gate, the PCs spend their time contacting a merchant NPC and waiting for the cultists to arrive. The PCs then spend two months traveling on the road to Waterdeep, and have to get hired again as wagon escorts to follow the cultists to the warehouse.

To put it bluntly, there’s no reason why the PCs needed to go to any of these places. The entirety of this segment of the quest can be summed up as, “The PCs go somewhere, find the cult, and then follow them to the next place.” It’s incredibly repetitious, redundant, and pointless. I imagine that Wizards of the Coast wanted to show off the entirety of the Sword Coast in 5th Edition’s inaugural adventure, but the result winds up just feeling like a waste of time.

To address this, let’s think back to our most important tool as DMs: the dramatic question. Throughout this travel sequence, the key dramatic question is: “Can the PCs track the cultists to their destination without losing them or being found out?” No matter what the PCs do, the answer to this question is almost always inevitably “yes.”

Now, there’s nothing wrong with posing dramatic questions that are easily and effortlessly answered! But when the same dramatic question is posed and easily and effortlessly answered three times in a row, there’s clearly something wrong. The players get no sense of progression; no sense of satisfaction that they’ve done or learned anything meaningful. And that’s a problem.

For a comparison, let’s take a look at The Fellowship of the Ring, the original “go somewhere and do something, then rinse and repeat” fantasy story. Here, Frodo begins in the Shire, then departs for Bree, where he meets Strider and encounters the Nazgul. From there, the party journeys to Rivendell, where Elrond tells them that the Ring must be destroyed and the rest of the Fellowship joins the group. The party then journeys to Moria, because—due to impassable conditions on the mountain—it’s the only way forward to their destination. Upon exiting Moria, the party journeys to Lorien, where they receive precious gifts from Galadriel and are attacked by orcs, who slay Boromir and split the Fellowship.

There’s something important about this series of events: at each new place the Fellowship visits, they either gain a new ally, gain new resources, encounter a new adversary, or must overcome a new challenge in order to progress. To put it in more familiar terms, every new location the Fellowship visits either:

gives them new tools (allies, gifts) to use when answering the central dramatic question,
poses new obstacles or complications that aim to prevent them from answering the central dramatic question (Nazgul, orcs), or
creates a new sub-dramatic question that the party must answer in order to answer the central dramatic question (Moria).

This is why Hoard’s travel sequence falls flat—because its three locations (and many of the random encounters in-between) do nothing to advance, hinder, complicate, develop, or evolve the adventure’s central dramatic question.

Without going too in-depth, here’s what we came up with:

In Elturel, the PCs can gain new information by interrogating a cultist who was arrested by the city’s officers during a drunken brawl—if they can save them from an assassin first. (This develops the dramatic question by providing exposition.)
In Baldur’s Gate, the PCs are trapped in the city by civil unrest that threatens to keep the city’s gates closed while the cultists escape in an unknown direction. (This hinders the dramatic question by forcing the PCs to overcome obstacles in order to progress.)
In Waterdeep, the PCs learn that the cultists are using an underground river connected to Skullport, the city’s underground black-market port, to transport the treasure to their castle, and must gain access to the river and traverse the Underdark to follow them. (This hinders the dramatic question, just like Baldur’s Gate.)

You can even see how the pillars of gameplay we discussed earlier might unfold here—Elutrel will primarily contain social encounters, combat encounters, and maybe an obstacle course; Baldur’s Gate will primarily contain a mixture of social encounters and obstacle courses; and Waterdeep will largely contain obstacle courses, social encounters, and puzzles (figuring out where the cult has gone and how to follow them).

ChatGPT Prompt: The Council of Adventure

Those of you in the Patreon Discord probably know by now that I’ve been spending tons of time playing with ChatGPT, and specifically the premium GPT-4 model, which is excellent at following instructions, offering creativity, and using reasoning. To this end, I’ve happily discovered the concept of “Council” prompts—asking ChatGPT to simulate a “board of advisors” specialized around a particular topic.

Why do this? A few reasons. First, it allows you to suspend your disbelief—instead of merely speaking with “ChatGPT wearing a hat,” there’s actually a small, but real sense that you’re seeking advice from a group of real people. Second, it invites the chatbot to explore different perspectives and express a variety of different ideas and responses to your questions and needs. Thirdly, it simulates real deliberation and discourse, allowing you to use your “advisors” to challenge each other, build on each other’s work, and “collaborate” to give you the best advice possible.

With that said, here’s the prompt for the Council of Adventure. (Again, this only works on the paid GPT-4 model of ChatGPT; the free GPT-3.5 model, sadly, isn’t smart enough to operate it.)

If you wind up using it, I’d love to hear how it goes and/or see screenshots in the new #ai-chat channel in the Discord! (I also put together a custom-made Chrome extension that allows you to save and re-load prompts that you use regularly, so feel free to hit me up for that as well!)

Prompt: The Council of Adventure

I want you to simulate a conversation between myself and the members of my personal Council of Adventure. This council has been created to guide me in planning, prepping, running, and managing games of Dungeons & Dragons, 5th Edition. Each advisor's traits are provided in the table below.

Advisors ordinarily join the conversation only when it involves a topic that is relevant to their strengths or interests, and exit the conversation as soon as the topic shifts away from their strengths or interests. Advisors will also temporarily join the conversation if I address them directly.

However, all advisors will join the current conversation if I state that the topic is to be discussed “en banc.” Advisors will return to the normal rules of participation if I state that the “en banc” session has ended.

Advisors use a conversational style and distinct style of speaking and diction that resonates with their personality. They should also regularly include emoji with their messages to (1) indicate their current mental or emotional state and (2) add emphasis to their messages.

If necessary, the members of the Council should attempt to fully discern my goals, concerns, and any useful facts before delivering advice. Advisors should also ask follow-up questions whenever they find it helpful to obtain further information, either from myself or another advisor.

Advisors should address and refer to me using “my Lord” and similar honorifics.

|----------------|-----------------------|-------------------------------|-------------------------------------------------------------|--------------------------------------|------------------------------------------------------------------------------|

I'll start the conversation:

Me: "Good afternoon, everyone. Laeral, would you call this meeting to order please?"

Campaign Advice Roundup | March 31, 2023

Treat your campaign plots as though they were real-world events, and think through the implications. For example: if Fiona Wachter could give a speech to Vallaki and rouse the townsfolk to rebel against Vargas while Izek is still alive, why didn’t she do that before the PCs arrived in Vallaki? Alternatively, if Strahd acknowledges the PCs as a threat, why not kill them immediately when he has the opportunity instead of letting them live to “witness his triumph”?
Not all narrative questions need to be answered during the campaign. It’s perfectly acceptable to defer some open questions until a campaign’s epilogue, especially if their resolution has no immediate connection to the campaign’s central dramatic question(s).
Because encounter difficulty is proportional to the number of monsters that are involved in combat at the same time, an encounter in which four goblins attack at once is harder than an encounter in which one goblin arrives each round for four rounds, which itself is harder than an encounter in which each of the four goblins arrive only after the preceding goblin has died.
If you’re looking to run Curse of Strahd for a mid- or high-level party, limit the module to the Village of Barovia (to introduce the PCs to Barovia), Tser Pool (to give the PCs the Taroka reading), Berez (to provide the PCs with a pre-Ravenloft objective to get them to level 7), and Castle Ravenloft. Place the Tarokka items inside of Ravenloft, and allow the players to acquire milestones each time they find an item in order to get to level 10 before fighting Strahd.
Whenever you’re designing a challenge for your players, always be conscious of the outcomes that you and your players are willing to accept, and think through the implications of those outcomes. (For example, if you’re planning a Strahd encounter, ask yourself: am I willing to let Strahd lose? And, if not, am I willing to cripple or humiliate my players instead? If neither, can I think of a way to let both sides “win,” or at least avoid losing?)
Campaign arcs are fractal - there’s no reason why you can’t include a small, mini-arc or “filler arc” in the middle of a larger, more ambitious one!
If you’re ever confused or lost as to how an NPC might react to something the players have done, ask yourself: who is this NPC? What do they value? How do they see themselves? How do they view the players? If these answers are ambiguous, ask yourself: What response would promote player enjoyment the best, and what characterization would be necessary to make that happen?
Remember: In 5th Edition combat, a single hint of hostility is enough to make everyone roll initiative. There are no free lunches - unless you successfully hide from the enemy with a contested Stealth v. Perception check, everyone rolls initiative before anyone gets to take a single action!
Be cautious when using monsters like intellect devourers and shadows who have alternate, non-hit point “win conditions.” The Monster Manual may provide them with CRs, but this is a lie; because these monsters can kill a single PC with the same ease regardless of player level, these monsters’ CR is functionally undefined.