Game Design Is Input/Output


“It is a truth universally acknowledged, that a game designer in possession of even a modest audience must be in want of a clever new way to talk about fundamental design.” — Jane Austen, on The Forge, circa 2003, probably.

This is one of those abstract-theory posts that might be a total rehash of something someone else wrote like fifteen years ago, but for me it is independent invention and the beginning of further thought, so maybe you’ll come along with me on this. I want to talk about roleplaying games and their rules as a content input/output machine.

We talk about roleplaying games as a collaborative storytelling experience – that it isn’t just the GM telling the players a story, which they passively absorb. The players get to contribute content also. (For GM-less/ful games, it isn’t one of the people at the table telling the whole story.) The most basic rule is some variation of passing the mic:

  • control passes when the speaker reaches a cliffhanger
  • initiative
  • passing control clockwise around the table
  • conditional permission to interject or seize control
  • a party “caller” (as in ultra-old-school D&D) who leads the players and aggregates their actions for the GM’s convenience
  • the GM’s sense of pacing and cinematography – here I’m thinking of PBTA games
  • (this list is not exhaustive)

In terms of the content machine, this is who gets to decide what is fed into the machine and when they get to decide it. It’s necessary because humans aren’t good at resolving everyone in a conversation speaking at once. It’s also a control to make sure shyer players can still take part while more assertive players are in the group.

The next step of any game’s flow is what the speaker gets to decide. This is the shape of the input. First, the speaker has to be willing to work within some of the established context – if everyone has agreed to a Roman historical epic, you’re rejecting the game outright if you attempt a story input about your Sherman tank wiping out Pompey’s forces. If you’re playing a time-traveling romp, on the other hand, this may be perfectly normal.

Within a game’s context, one of the major things that separates roleplaying games from, say, board games or video games is the breadth of variety in possible inputs. This point is part of how roleplaying games came to be distinct from adjudicated wargames, as Ben Robbins explains here. The majority of player and GM inputs are narration and dialogue that change the shared context (in a way that fits the shared expectation) but don’t trigger any rules. The kinds of actions that do or do not trigger rules vary from game to game. Imagine a player describing their sword flourishes that end in a thrust against their hated foe. In D&D, this might be an attack roll, or an attack roll with a maneuver on top, while over in Fate that might be a way to capitalize on a whole slew of Aspects, and in Amber Diceless/Lords of Gossamer and Shadow… well, you get the picture. One of the things that makes Fate tick is that just vocalizing your input is a drumbeat of character or scene establishment or reminders.

It’s possible to have a bad experience at the table as part of the input, though as it’s the initiation of player agency within the rules, you can usually push past any rough patches in the input to get to seeing the results of the actions. For example, if other people at the table have abilities that trigger during your vocalization of input (and for some reason can’t or don’t wait until you’re done speaking), that can be very frustrating, though most tabletop games are cooperative enough only the GM gets interrupted like this. The frequency of interrupt features (not, I want to note, any rudeness on the part of my players) was part of what soured me on 4e.

It’s also possible to have an excessively complicated input process, as in 3.x’s attack roll incantation: “Did you remember bless prayer haste bull’s strength?” Then there are situations dense with modifiers, such as a Tracking roll in most editions of D&D (and certainly OD&D through 3.5). There’s also building a die pool in some games, and the step of parsing the die roll, and so on, which… well, some players are good at mental math, others aren’t. I am, and I still prefer to keep math simple so that I can still perform it at the close of an all-day session. On the other hand, it’s possible to make a micro-game out of parsing dice (or other randomizer) results, as with games that reward sets or runs in the results.

Continuing with issues at the point of input, situational constraints in the game may narrow the acting player’s options to the point that the situation becomes more frustrating than merited. For example, in 3.x, a character who is bleeding to death gets a stabilization roll each round, and typically has no other options – a frustrating situation perhaps, but at least each roll is tense and you presumably got to do something fun while losing those hit points. (1st-level characters with d4 Hit Dice notwithstanding.) What about a character who gets grappled, and isn’t a grappling specialist? There’s probably not a lot you can do each round but roll to escape, fail the roll, and hope your allies save you. (This amounts to a bounded accuracy issue, and also an issue with just how screwed you are if an enemy wins a few consecutive grappling contested rolls.) Then there are 3.x’s versions of other really bad conditions, like stunned, paralyzed, or Tasha’s hideous laughter. Depending on the exact source of the effect, you might get a new save each round, but probably not. What I’m trying to get at with all of these examples is that constrained input has a lot of potential for frustration, even ruining a player’s whole experience of that encounter. Mitigating that frustration mostly requires the player to feel like their own choices put them there, one way or another. (I’m also not the least bit sorry to see just about all negative conditions become “save ends” in 4e and grant a fresh save each round in 5e, even though that erodes crowd-control considerably.)

Okay, so I think you get what I’m saying about “content input.” The game (often in the form of GM narration) returns a content output, such as a description of the monster taking damage, possibly even suffering a messy demise. In D&D, it’s very much on the DM to spice up the content output beyond “yes, that attack hits. The monster takes how much damage? Cool, got it.” I’ve been guilty of leaving it there… more often than not, but keeping attack descriptions fresh over many, many rounds of combat is not my forte. Content outputs in other situations range from the dreaded “roll fails, story stops” situation that the philosophy of failing forward is intended to address, or provoking a content input from someone else – like dialogue from an NPC.

D&D largely favors and teaches straightforward, binary content outputs, with the barest of exceptions for crits and the variation of damage rolls. It relies mainly on a large number of content outputs and characters having a large-ish resistance to catastrophic failure (conventionally called “hit points”) to create nuance in success and failure. This has the knock-on effect of increasing the importance of every kind of modifier, as a large number of rolls diminishes the effect of statistical outliers. Obviously, D&D is hardly alone in this approach.

Other games, though, encode more yes-and, yes-but, no-but, or no-and outcomes. You can see all of these cases at work in any PBTA game, though I don’t mean to suggest that they invented it – after all, a critical hit is just a yes-and, and a fumble (though a bad idea in D&D gamerunning) is a no-and. (The pushback when D&D tried to introduce no-but outcomes in 4e and D&D Next was ferocious – here I’m talking about “some damage on a miss” mechanics.) PBTA’s move resolution explicitly eschews a flat no as an outcome; 7-9 is your yes-but or no-but outcome, 10+ is yes or yes-and… but 6- is not just a no, it also immediately gives the MC a chance at a content input, which will nearly always be bad for one or more PCs.

PBTA is a big part of why I’m even writing this post, so let me dig into that a little more. Its move structure is about ruthlessly making sure that action stays interesting. It does this by funneling content inputs toward decision trees that it has already constructed to be varied and interesting. There’s a distant kinship here to the structure of powers in 4e D&D, which don’t grant a lot of micromanaging internal decisions, but once you’ve chosen one, you’re getting a few camera shots of a highly cinematic experience.

The potential content of, say, Apocalypse World is just as infinite as any D&D game; I’m not interested in a qualitative judgment here. AW’s moves, along with the incentives worked into playbook features and the structure of advances, emphasize content outputs that lead back to new conflict and tension. I’m sure it’s possible for tension to get fully resolved in AW, but it is not here for that – the best it really offers is shifting the camera to someone else so that the player is out of the spotlight and catching a breather. There’s nothing going on here that a skilled D&D DM wouldn’t already be doing – it’s all the kind of dirty tricks (er, legitimate dirty tricks) that I used to see called Rat Bastard DMing, in tones of glowing approval.

The dynamics of content input and output can vary enormously even within a single system. I first started thinking about game systems as input/output thanks to my domain rulership series over in Tribality. Many of those systems highlight a classic form of content output – the random encounter table that has been part of D&D since forever, or tables of random content outcomes (such as the 5e DMG’s Carousing rules). Roll some dice, and get a text block. It’s the most straightforward form possible, so the challenge is in writing compelling text that changes the status quo enough to drive new actions, without becoming a non sequitur. This kind of content is most ideal for GMs that are strong on embracing unexpected system outcomes and justifying them within the story. Games like ACKS use tables to conceal huge volumes of mechanical (in ACKS’s case, economic) and creative work that the designer has done for you, in an attempt to turn a fairly dry input into a compelling output.

In a similar vein, many games deliberately present the GM with content that they “must” (for whatever value of “must”) weave in during the course of the session or adventure, such as working out Icon Relationship rolls for each session of 13th Age. The Fate GM that doesn’t aggressively turn back to the players’ Aspects and figure out ways to phrase them as compels is doing the game’s whole point economy a serious disservice. They’re sort of a case of a desirable content output leading the GM to hunt around for a suitable input to get them there.

Okay, so you’ve come with me this far. What is the point of talking about games as input/output machines? If there’s anything to this as a conversational approach to games, it is to talk about the exact step of an exchange in which a gameplay loop becomes unsatisfying – the input, the process, or the output. Of course, that can also be situational, based on factors like player investment in the outcome and the GM’s sense of the significance of an encounter. If the party’s rogue finalizes a boss in a devastating ambush, that might be incredibly disappointing to the rest of the party (why were we here?), or the result of the whole party working together to set up the hit and provide backup if things went south. Both cases could happen in most games, but the first is a lot more likely in the typical D&D contexts, while the second is a lot more likely in, say, Leverage or Blades in the Dark (ignoring for a moment that no one is “the party’s rogue” in either of those games – they’re all rogues). D&D builds its bosses to avoid the dissatisfaction of the first outcome, while Leverage and BitD build their whole scenarios to provide the satisfaction of the latter outcome.

The content input/output machine separates roleplaying games from storytelling by making it collaborative in a controlled way, by introducing the designer’s ideas for the group to incorporate or reject, by introducing randomized outcomes to build tension, and other things I’m forgetting at the moment. There are as many things that this model doesn’t fit as things it does, but still. I hope this can be a jumping-off point for an interesting conversation or two. Ultimately, I think we want content output to generate the therefore of the next content input – the fundamental propulsion of the narrative, as long as you incorporate character motivations.

Leave a comment

Your email address will not be published. Required fields are marked *