How to Write a Tabletop RPG

This chronological rundown of how I approach RPG design will go over the most universal principles I've learned or copied from others.


All rules are bad, but some rules are worth it.

The rest of the ideas here will talk about technical implementation, but the core thought, riding alongside every notion, is that every rule slows down the game, makes demands on people's time, and creates a barrier to crafting a narrative. Therefore, every rule must justify its existence, and the more difficult the rule, the more it has to improve the game in order to be allowed in.

Genre: Ins & Outs

The first question is genre. I don't want to give any precise definitions here - call it 'category', or 'setting', or whatever. The point is to look at what kind of literature or TV you want to reproduce, then make those things have inputs and outputs. We're not considering rules just yet - only making a list of what things we care about.

Star Trek

If you're making a Star Trek game, it's more important to know that Worf is a warrior, and Data knows everything, than knowing exactly how good Worf is with a bat'leth vs a spear.

The outputs for these things are the kinds of resolutions we see, scene by scene. They have no money - characters don't have wealth except for Quark, and he's really only 'poor' or 'rich', which changes nothing for him. Everyone get injured however - in fact everyone gets regularly stunned. 'Red-shirts' may die, but characters do not, so no need to include it as a possibility. If the players need stakes, then the stakes should be rank, discovery, reputation, or ability (such as science).

So far we have a few properties we care about:

  • warrior
  • rich
  • stunned
  • rank
  • science ability

These might end up pure Booleans ('rich' or 'not-rich').

Buffy the Vampire Slayer

We care about weapons, we care about relationships between characters, and we care about how strong someone is. We care about research ability, which needs to result in useful information when it's used. We also need to ensure side-characters have a part to play, without having any strength. One Buffy RPG included some ability for side-characters called 'guys, I think I'm okay', which allowed them to shrug off massive damage by retconning it as dumb luck.

What we don't care about is the difference between an Uzi and an AK-47, or the ability to speak modern languages - all vampires speak English.

Mapping Inputs and Outputs

Now we have our ins and outs, we need a system which gets us from one to the other as fast as possible. When someone fires a phaser at a klingon, what might the results be?

Bad Example

  • roll to see if you hit with Dexterity + Phasers
  • roll the enemy's dodge to see if they can avoid the stun
  • roll to see how much damage you've done (assuming the enemy did not dodge)
  • roll to see how the enemy reacts

This process clearly demands too many steps from the player. The 'inputs' are just:

  • 'how good you are at phasers'
  • enemy's ability to dodging

The 'output' is

  • type of damage

Better Example

Star Trek officers don't interact with 'Dexterity' a lot, so we don't need to make sure that Dextrous characters do certain things well. We can reduce the skill-roll to just 'phasers'. 'Dodging' phasers seems to be mostly a matter of putting oneself behind cover, if any, so we'll leave that for now, and just say whoever fires first, wins. 'Damage' is just whatever you set your phaser to. In a lot of Star Trek, phasers automatically kill, but the best Star Trek is obviously DS9, so I'm going to go with how they do things and say that phasers can give somewhere between light wounds and death. So the possible results might be:

  • Miss
  • Stunned (fall over, lose next action)
  • Light wound (another light wound will go to the next Damage phase)
  • Serious wound (penalty to movement)
  • Death

No we can add some numbers. If we want a really 'swingy' random system, the D20 works well:

Roll Result
<8 Miss
9 Stunned (fall over, lose next action)
16 Light wound (another light wound will go to the next Damage phase)
17 Serious wound (penalty to movement)
18 Awful wound (penalties all round, will die without medical treatment)
>19 Death

Or for more reliable numbers, we roll 2D8, because the D8 seems like a sci-fi-die.

Roll Result
<6 Miss
8 Stunned (fall over, lose next action)
10 Light wound (another light wound will go to the next Damage phase)
12 Serious wound (penalty to movement)
14 Awful wound (penalties all round, will die without medical treatment)
>16 Death

We can add a bonus for people trained in using phasers, and adjust all the numbers to accommodate how lethal we want the game to feel.

Of course we need plenty more rules:

  • something for initiative (Star Trek deals with quick-draw situations a lot),
  • how cover works,
  • another for plot-points (main characters never just die),
  • something to resolve social situations ("cool it, Jake - let me talk to the Klingons!"),
  • et c.

So we've gone from our 'input', to the 'output' in a single roll.

More Star Trek Thoughts


Smaller systems are always better. We've already added inputs and outputs when constructing the genre - that part has finished. Everything which comes during mapping should draw a line between those two as quickly as possible.

Dice do not need to be assumed - if a system becomes unpredictable without dice, best to remove them (or remove them anyway if a situation should be predictable).

Some kind of 'base-system', is generally required as well. Old D&D games can feel clunky because it's not clear how to do 'stuff' in general; one roll is for hitting enemies, another for dodging magic wands, and another for searching rooms - all totally unrelated! By and large, this base system will determine how swingy the game is.

Modern D&D's reliance on the D20 mean pretty much every roll swings around wildly. Fate Core's fudge-dice keep rolls consistent, so the bonus-points from tokens play a big part. White Wolf's dice can be consistent with a low difficulty, or wildly swingy, with a high difficulty. These rules let snake-vampires look tough when they have their snake-skins on, and let the artsy vampires consistently spot details. They let fights on a moving train give crazy results simply by changing the dice-difficulties.

Feedback Loops

When systems make outputs that become inputs, it entices players. Consider D&D's 'Pick Pockets' Skill; you use the skill and gain gold pieces, then use those gold pieces to buy equipment, and use the equipment to kill monsters, who also have gold pieces. Our options are opening up already!

On the other hand, Vampire: The Dark Ages has a dynamic system which promises to do anything, but when you pick someone's pockets, the system just says 'okay, I guess you have money or whatever'. The DM can fill in the details - 'three coins, two from the New Hungarian King, and another which looks like it may be East Roman', but the system remains silent. When the system wants to represent money, it does so with the 'Resources' Trait, which is simply rated 1-5, and picking someone's pocket will not influence it, so from the system's point-of-view, picking pockets does nothing.

We've come full-circle back to genre once again. A loop helps focus player attention, without being 'grabby'. If killing Klingons in Star Trek gives you XP, that's bad, because players will see the output 'dead Klingon' results in the input 'XP', and killing Klingons. On the other hand, if the output 'resolve a conflict', results in 'more Star Fleet buttons', then they will most definitely want to resolve conflicts. And of course anyone playing a Buffy RPG should be interested in uncovering lore, so every enemy should have some kind of lore which helps defeat it, and every enemy slain can help build skills.


Hopefully the idea of mapping inputs to outputs feels fairly clear by now, because I want to start again. The best way to understand any RPG is removing the dice . Take the bare inputs and outputs, and look at how many steps a player has to take. Every step is bad, because it creates more work, so every step has to be justified because:

  • it lets us represent the genre well, or
  • it allows players to make an interesting decision.

Notice that with A,D&D fighters make no decisions except whom to hit, while mages make new decision every round. Note also that a lot of decisions are pointless when you stop to think about them. If a fighter has a special 'fighting surge' which they can use once per combat, then the best time to use it is immediately, assuming all enemies are the same. In White Wolf's World of Darkness rules, the ability to 'take aim' should never be used when rolling with 4 or more dice, but should always be used with 3 or fewer, due to the average damage expected. These illusory decisions are cheap shots, and even when players can't do the Maths by hand, they will eventually gain a feel for the system, and uncover the system's foetid banality.

Ability + Skill Pairs

This is something best done right, or not at all. If some Skill uses an Ability (e.g. Academics always gets a bonus from Intelligence), then we may as well just have 'Academics' on its own. The only thin 'Intelligence' does here is make busy-work for the players, and set a minimum to their Academic ability (since someone with Intelligence +2 cannot start their learning with an Academics roll at +1: it must be +3 to begin with). Such a system may as well just use the Skills, and nothing else.

On the other hand, Ability + Skill pairs can save everyone a lot of time, if use correctly.

(Skill/ Abilities) Strength (2) Intelligence (3) Charisma (-1)
Academics (2) Shouting a tale to a crowd Recalling facts Telling an entrancing story
Deceit (1) Threatening someone Conceiving of a devious lie Looking remorseful
Survival (3) Sleeping in the cold Navigation Convincing someone they will live

This chart shows only 3 Attributes, with 3 Abilities, but with only 6 stats written down, we have distinguished 9 different abilities, each of which could have their own rating, and a clear reason why. Now 'shouting a tale to a crowd' may not come up very often, but this misses the point - any time a character does something they've never done before, and which nobody could have guessed they would do, we still get a good measure of their overall ability to do that thing in a way that makes sense. It makes sense that this person can spin a good tale and navigate, but cannot make fake apologies or tell very good stories.

Consider the DM

Many games have no systems for the DM. This seems like freedom, but it creates a lot of responsibility.

This is a good time to return to the genre and for more inputs and outputs. What do the enemies want? The DM will never gain XP for killing players, but an enemy's 'win' might turn into another encounter, a problematic situation for the players, or might be a good time to reveal an enemy's hidden base.

Having some rules (or at last procedures) for the DM can really help, especially if the system seems like it could answer the question 'what the hell do I do now?'.

Reduce, Reuse, Recycle

Reusing rules reduces mechanics, and all mechanics are bad, therefore recycling is good.

If your cowboys game needs rules for chases on horseback, do the quick-fire rules cover the same thing? Could you count 'jammed gun', as 'bucked off', or something, then pull them all into a single system with different applications?

If the dead-space pirates need a system for asking questions on the streets, can you re-use the system for scavenging wreckages, and just plug in different stats?

Recycled rules don't have to call upon the same rule explicitly - they may simply use the same patterns repeatedly. Suppose you've made specialist rules for crafting poisons, chatting to courtiers, and finding traps in dungeons, then you have three irritating little thorns in your system. But if commonalities between them can help suggest more unified systems. Each one of these activities may come in steps, and reveal clues - perhaps traps will yield information to be given to the players slowly, such as 'this room has some suspicious disturbances around the floor in the centre'. Poison may not seem to fit this mould, but perhaps poison-rolls can involve less poison-brewing, and more steps for the PC to discover how to slip the brew into the target's food. And of course, a failed roll to find traps results in springing a trap, but what about the poisons and courtiers? Well, they could add their own dangers, so a failed roll means the PC gets some bad outcome. Or they could become threatening, so the GM informs the player that some unknown danger looms, and gives them the chance to give up...or to make one last roll.

Crafting these little patterns in the rules presents a serious hazard - the 'uncanny valley'. If players read that rules A, and B start with steps '1, 2, and 3', then everything which looks similar must follow the same pattern. If hunting for clues during a murder investigation suddenly presents rules '1, 3, Gamma', this won't just shock people, but irritate them.

These little similarities mean players never have to feel surprised by the rules. They can skim the book, find specialized rules for spreading disgusting rumours about nobles among the peasants, and instantly think 'yea, okay - that makes sense', because they remember the core of the idea. And similarly, you're going to stop reading this, and forget these last few paragraphs. But hopefully you'll still feel the core of the idea, that 'all rules are bad'.