Writing RPGs is Hard, Part 1

2019-09-11

Modified 2019-12-31

In two previous posts (1) (2) I talked about trying to write custom RPGs. To summarize, here’s my working list of projects:

It’s worth noting that Astral and Paranormality, my oldest ideas, provide a system-light or systemless adjunct to other RPGs. Troika! Redo and Untitled #2 rewrite existing systems to my liking. only The Elf Game, Zeta World, and “Untitled #1” consist of wholly independent RPGs with their own settings. Even then, The Elf Game is a riff on D&D tropes, and Zeta World is both a reference to another post-apocalyptic RPG and a potentially multi-genre set of rules.

Which is all just to say I’ve yet to turn one of these ideas into a usable product. So take what I say with a huge grain of salt.

Below I’ll discuss some of my design principles and preferences. I’m not saying these are the only design choices designers can make, nor even the “best”. These are just the kind of games I prefer to play.

Use Common Components

Especially when starting from scratch, keep material requirements simple. Players should truly need only pencils or pens, paper, a handful of dice, and a rules reference. Other visual aids can help, but if a design demands specialized dice, reference cards, counters, miniatures, maps, phone apps, and the like, printing and sourcing those components isn’t the problem; the problem is expecting players and GMs to juggle so much state at once.

Any design should require no more than what most players might already own. Maybe Fantasy Flight Games might be able to manufacture proprietary dice and require people to play their RPGs with them. I, as an amateur with more time than money, cannot.

That means someone should be able to play one of my games only with:

  1. A copy of the rules, digital or dead tree

  2. Paper, index cards, or something else to write on

  3. Pens, pencils, or something else to write with

  4. One or more common randomizers, namely:

    • one or more six-sided dice (d6s)
    • a single polyhedral die (probably a d10, d12, or d20; nobody likes d4s)
    • a d100 (i.e. two d10s, one marked as “tens” somehow)
    • playing cards
  5. Poker chips, glass beads, a countdown die, or the like to track one or a few rapidly changing resources without wearing a hole in a piece of paper.

Stuff to avoid includes:

Prefer Simple Mechanics and A Single Pass/Fail Continuum

Randomizers and dice mechanics that sound “brilliant” and “innovative” don’t always play well at the table. Boring old polyhedrals and six-siders with numbers on them, read in one of a few common ways, have stood the test of time for a reason.

For example, Fantasy Flight Games’ Genesys system, descended from their systems for Warhammer Fantasy RPG 3rd Edition and Star Wars, uses custom color-coded dice. To resolve a situation, a player builds up a pool of Ability dice (green d8s) equal to a base characteristic, Proficiency dice (yellow d12s) equal to a skill rank, Boost dice (blue d6s) for added advantages, Setback dice (black d6s) for added threats, Difficulty dice (purple d8s) for the level of resistance, and Challenge dice (red d12s) for any extraordinary risks. Then the player rolls the dice, add up the number of symbols of each type – Success, Advantage, Triumph – subtracts the number of opposing symbols – Failure, Threat, Despair, respectively – and (so the theory goes) use the tally on each scale for a clear description of what happened. I on the other hand kept getting results like Two Failures with an Advantage and a Triumph, but no net Successes. Not exactly clear.

Far better, I think, to have a single linear scale of success or failure, with the “nuanced” bits derived from a position on the scale.

HeroQuest (the new RPG, not the old board game) had both player and GM roll a d20 and from their relative values derive which side succeede Divd or failed – assuming no tie – and whether the success/failure was marginal, minor, or major.

Even that seems too complicated, though. I’d prefer a simple scale of Succeed, Tie, or Fail (“Win, Lose, or Draw”?), with improbable results denoting automatic or “Critical” successes/failures.

Use a Bell Curve For Vaguely Realistic Probabilities

Most of the mechanics I’m looking at use d6s; they’re common and with enough of them I can make bell curves.

Each is essentially the same mechanic with a different probability curve. 2d6 or (d6 vs. d6) is a pyramid, 3d6 closer to a bell curve, more total d6s even closer. As the number of random variables increases, all distributions approximate a bell curve. Thus it seems more suited to tasks with a lot of random variables and an assumption that a character has an “average” or “median” level of performance.

Untitled RPG #1 uses a distinct mechanic I’m calling P(1/3):

Technically pool of successes isn’t a bell curve, but a binomial distribution, and any probability besides 50% creates a skewed curve … but it’s close enough.

Use Flat Distributions For Simple Probabilities

Some people prefer flat distributions, notably:

Especially when the extremes indicate extraordinary success or failure, e.g. D&D’s Critical Failure on 1 and Critical Success on 20, the “extraordinary” can happen 1% or even 5% of the time. Compare to 3d6, where the extremes of 3 or 18 come up 0.46% of the time.

Flat distributions do have the advantage that every +/-1 in in a modifier or target number alters the probabilities by the same amount: +/-5% for a d20, and (obviously) +/-1% for a d100.

In any game emulating or parodying D&D a d20 is practically mandatory. Likewise, if I want to follow the traditions of d100-based RuneQuest I’ll have to stick with a d100. (King Arthur Pendragon and its d20 Roll Low mechanic notwithstanding.)

FWIW, any single die – d6, d10, d12, d30 – and a deck of cards each have a flat probability distribution. Unless there’s a mechanical or thematic reason that overrides the “keep it simple” principle above, I’ll stick with the traditional d20 or d100.

Minimize the Amount of Math at the Table

Also worth noting: humans and computers find some mathematical operations easier than others. These operations, in order of speed and cognitive resources, are:

  1. Comparing two numbers and determining which is higher.
  2. Adding two numbers together.
  3. Subtracting one number from another.
  4. Multiplying two numbers.
  5. Dividing one number from another.

Beyond that are exponentiation (repeated addition and subtraction), conversion to logarithms, and other esoterica. Reducing the number and complexity of operations can speed up play enormously.

For that reason, I shy away from any “roll under” mechanic if I can avoid it. All have a mathematically equivalent roll over version, and most of the time the roll over version is cleaner and quicker.

For example, Troika! defines success on unopposed skill rolls as

2d6 ≤ Skill + Special Skill + modifiers

defined thus:

Skill
a basic attribute all PCs and NPCs have.2 PCs start with a Skill of 3 + 1d3 that never improves.3 NPC Skill depends on species or role and varies from 3 (Gremlins) through 5 (Goblins) to 14 (Dragons).
Special Skill
a specific ability typically ranked between 1 and 6, e.g. Sword Fighting 2. If PCs lack a Special Skill, it defaults to 0. NPCs have no Special Skills.
modifiers
Anything else the GM wants to throw in, based on the situation.

This probability distribution is effectively equivalent to

2d6 + ( Skill + Special Skill + modifiers ) ≥ 14

It would be simpler and fairer to replace Skill with a constant 5, yielding:

2d6 + Special Skill + modifiers ≥ 9

Now Special Skills, or simply Skills, won’t add to targets for roll-under for unopposed tests but add to die rolls for roll-over for opposed tests and combat. NPC Skill ratings for Enemies in the book will need to drop by 5 points, to between -2 (Gremlin) and +9 (Dragon).

Troika! also has a roll-under Luck mechanic, though; and I’m debating whether to keep it as is or do something radical like gradually eroding “Luck Points” that add to a Luck Save, also defined as 2d6 + Luck ≥ (some number).

d100 is inherently roll-under, and gets complicated when introducing “difficulty modifiers”. (I’ll have more to say about that later.) Unfortunately rolling d100 dice gets less comprehensible in a roll-over system, precisely because if a player’s target is 52 on a simple d100, they know their chance of making it is exactly 52%.

Resolve As Many Situations As Possible with a Central Mechanic

The first game masters (or “dungeon masters”) invented brand new rules with each unexpected situation. As a result, early D&D used a hodgepodge of mechanics.

And so on.

Tunnels & Trolls, the less “complicated” D&D “knockoff”, reduced this to three circumstances:

  1. Want to hit something (in melee combat)? Pick sides, roll a bunch of d6s plus “adds” for damage; higher total inflicts the difference on the losing side. One reviewer called the combat system “the worst I’ve ever read”. But if you really don’t care about skirmish tactics and just want to see who wins rolling dice simultaneously is definitely faster.

  2. Want to cast a spell? Pay the STR (later WIZ) point cost and it happens.

  3. Want to try or avoid something else? That’s a “saving throw”: roll 2d6, Doubles Add and Roll Over4, add a relevant characteristic (one of six to eight, depending on version), and compare to the target number for that saving throw level. The Target Number is 15 + Level x 5; Level 1 or L1 is 20, L2 is 25, etc. If the total is ≥ the Target Number it succeeds.

Subsequent rules chose a single resolution mechanic for most actions.

Want to do anything in …

Even with a single mechanic combat can get a little fiddly, thanks to RPGs’ wargaming roots, but ideally not by much. In WEG D6 games the same die conventions determine whether the attack succeeds and the severity or amount of damage done. Notably, many World of Darkness games determine success and damage in the same roll.

The huge advantage of this approach is that a GM can make an on-the-spot ruling quickly by using a pre-established method. Chaosium rulebooks would have an entire chapter devoted to “Spot Rules”, all of which boiled down to making a percentile roll against a skill (or a characteristic X some number, usually 5). Having a very small number of go-to mechanics makes the GM’s job easier, and the resulting rulings more consistent and easier to remember.

D&D 3.0 finally caught up by settling on a d20 based mechanic for (nearly) everything. Not that everyone approves. A few “old-school” holdouts keep the d20 roll high mechanic for combat and saving throws, then use some other mechanic – d6, 2d6, d20 roll low vs. a characteristic value, whatever – for everything else. Backwards compatibility? Nostalgia? Contrariness? Who knows?

After more than forty years of game design, GMs don’t have to invent new mechanics from scratch. Designers have already done it. The most successful designs pick one simple procedure to resolve the vast majority of expected situations, and provide guidelines and examples to extend it to the unexpected. This tactic makes new rulings consistent and less stressful.

NOTE: On Dec 31, 2019, I fixed some markup errors, some related to the new Markdown engine `hugo` uses, and changed some URLs to refer to the Chaosium site rather than DriveThruRPG.

  1. After my fake coin collecting mania of a few years ago I’d like to write a game which requires massive fluctuations in silver pennies / credits / quatloos / (mumble) points, but I’ve yet to discover such a theme. Plus most people would use cheap plastic or clay poker chips instead. ↩︎

  2. In the Fighting Fantasy game books, from which Troika! “borrows” its system, Skill is the only measure of a character’s ability in fighting, sneaking, etc. ↩︎

  3. In Advanced Fighting Fantasy, Arion Games’s licensed rewrite, players allocate points among SKILL, STAMINA (reduced by damage), LUCK (rolled for saving throws), and optionally MAGIC (Skill for spell-casting), although Arion feels the need to SHOUT their names. ↩︎

  4. That is, if the dice show doubles, add the total to the result of another 2d6; if that’s doubles, add that and roll again, etc. ↩︎

  5. Many of which are out of print either because their licenses expired (e.g. Stormbringer/Elric!, ElfQuest, Ringworld) or poor sales; most of the latter are still available as PDFs. ↩︎