One of the most divisive topics in the RPG community revolves around prefered dice systems. The hobby has dozens of different approaches, each with hundreds of implementations that are uniquely tailored to each game. Despite this, live-and-let-live seems impossible over nearly-sectarian divides. Divisions have been drawn and insults hurled over notions of realism, genre emulation, bell curves, and other such trivialities. Yet amid this chaos, no one seems to have taken stock of what an RPG dice system is actually supposed to do. The system of the game is not a goal in itself. It is a tool designed to facilitate play, resolve challenges, and enhance narrative suspense. Putting aside all posturing, there are three paramount concerns for a dice system: (1) that it be fast and easy to learn, (2) fast and easy to use, and (3) that it yields a satisfactory resolution to challenges. When we keep this in mind, along with the other trappings of an RPG (especially the presence and role of the GM), we find out that most of the other concerns about things like realism or probability distribution simply melt away.

The most important aspect of a dice system is, by far, that it be fast and easy to learn. The learning curve for an RPG matters significantly to players that are new to the hobby, have limited free time, or limited patience. The amount of time and frustration before the game becomes fun acts as a significant barrier to entry. This barrier is even higher for GMs, who must understand the systems exceedingly well to properly run the game.

Dice systems form a part of that learning curve. Rolling a single die and comparing it to a target number is a very simple and intuitive probability structure to grasp. Any added element increases the complexity of the system, requiring more time to learn. For example, rolling three dice and adding them means that all numbers are not equally likely, and so increases the time taken to grasp the distribution. Rolling many dice and revaluing them, such as in White Wolf and other dice pool games (where a die can mean 0, 1, or 2) complicates the calculation further by making the distribution very complex. Some systems like exploding dice (where a maximized roll gets rerolled indefinitely) can make a large high-end tail that is extremely difficult to anticipate. Such complications make it harder for players to fully understand their capabilities, and harder for the GM to properly set challenges.

Ease and speed of use in play tends to be highly related to the learning curve. While play speeds up as players become familiar with the dice system, complex dice systems usually continue to take longer than simple ones. Rolling and reading a single die is an act that takes a few seconds at most. Rolling multiple dice takes a little longer to find and count the dice, but then requires reading, retaining, and conducting mathematical operations on multiple numbers. This may be only an extra ten or more seconds per roll. However, in game sessions where rolls are frequent, this this can add up to a substantial loss. This loss only increases as more modifications need to be conducted on each die. This also complicates and slows the GM’s tasks, as a firmly grasp of probabilities is required to design appropriate challenges. Added up, the cost of a complicated system in time alone can be considerable, equating to the loss of entire scenes in a single session or entire sessions over the course of a campaign. For this reason alone, the simple should be preferred to the complex when all else is equal.

The final major concern is that a dice system yields satisfactory outcomes. Generally, this means that for any task that a GM deems a roll appropriate, the system needs to give an appropriate chance at success based on the character’s capability and the challenge. Most dice systems are capable of generating odds of success covering most of the range between 0% and 100%. This is true whether one is using a single 20-sided die (d20), three 6-sided dice (3d6) and adding them, or a large number of 10-sided dice and counting dice of certain values. The key criteria is not what the dice system is, but how the target values for success are set. The main limitation imposed by the dice system is what specific values can be reached. For instance, a system based on single d20 can only set values in increments of 5%. A 3d6 system is much more coarse-grained near the middle (the difference in probability of exceeding a 10 and 11 is 12.5%) but finer near the tails (the difference in probability of exceeding a 3 and 4 is about 0.05%). Most dice pool systems lead to more complicated binomial distributions.

A key takeaway is that the dice system flows together with the rest of the game system, and does not operate in a vacuum. A game system should leverage its dice system’s distribution to best yield the results desired. For example: GURPS places a heavy emphasis on differences of skill, even a single point. Relying on the gaussian distribution (commonly called a bell curve) provided by three 6-sided dice emphasizes the value of a single point of difference near the center of the range. Weapons of the Gods wants characters with greater skill to have a chance for truly massive outcomes. Its unique blend of skill-based dice pools and yahtzee results does just that, with skill gains causing moderate rises in the average but tremendous top-end growth.

All of this brings us back to that much maligned and abused grandfather of the hobby: Dungeons and Dragons. Whatever else, the simple d20-based resolution system remains a mark of true genius. Yet that genius goes underappreciated in this day and age due to its “unrealism.”

Putting aside the fair question of whether realism is even desirable in RPGs, this view marks one of the greatest delusions among modern RPG hobbyists. Flatter distributions produced by single dice tend to be called unrealistic, while binomial or gaussian distributions produced by multiple dice tend to be called realistic. But this is based on a faulty assumption: that all dice systems should use the same approach to setting target numbers. It is true that the d20 system fails if you assume each integer increase of skill or attack bonus is supposed to be a massive difference. It is also true that the d20 system fails if the GM calls for rolls on things where there is a 1% chance of failure. That’s how GURPS was designed to work.

But that’s not how the d20 system was designed to work. Dungeons and Dragons always had certain “benchmark” levels when capabilities radically changed. These were typically spaced about 4 levels apart. Thus, we should not be comparing the incremental differences of +1, but the benchmark differences of +4. A +4 on a d20 is quite substantial, and represents a meaningful increase in the reliability of existing capabilities and the development of new capabilities.

More importantly, Dungeons and Dragons assumed the GM would consider what checks to call for and which to ignore. Is the matter insignificant? Is the character’s skill so far above or below it that success or failure aren’t guaranteed? Then don’t roll. This was the most important guarantor of “reliability” in the game. Human judgement provided a sense of realism that no universal dice model could provide. The d20 system thrived in the middle-ground that was most important for play, while effectively relying on a GM with common sense to control edge-failure cases. The d20 system delivered where it mattered most. Even after the active GM fell out of favor, tools like “taking 10” were added to provide the reliability that many craved.

On the other hand, the d20 system provides one of the simplest systems ever designed to learn: roll a d20, add some modifiers, and compare to the difficulty. It’s simple in play: roll a d20, add some modifiers, and compare to the difficulty. It’s straightforward for players to use information to guess their own odds: take the difficulty, subtract the player’s bonuses plus one, then multiply by 5% to get the odds of failure. It’s intuitive for the GM to set challenges: for 50% success, set a difficulty of the player’s bonuses plus 11; each integer change adjusts the odds by 5%.

Yet this simplicity costs almost nothing in usability. It scales smoothly from 5% to 95% odds in 5% increments. This is in contrast to other games, which sacrifice granularity in the middle of the range (where most rolls are concerned) to gain it at the edge (when rolls often needn’t be made at all). This allows the GM to set almost any needed probability, but is not too fine-grained as to be a distraction. The edge 5% cases occur just frequently enough that players never quite count them out. When the GM decides even 5% is too often, it provides a benchmark to obviate a roll. It is also amenable to many of the classic modifiers: bonuses, subtractions, and rerolls.

Few systems can boast such an easy learning curve, fast use in play, and generate as broadly useful outcomes as the classic d20 system. Its elegance and simplicity represent a genius yet unmatched by any other systems that exist today. Those that do not respect it merely have not taken the opportunity to learn its proper use. After all, dice systems are not the game. They are simply tools to allow the game to play. The simpler, faster, and more usable a dice system is, the better any game that uses it. And I can think of no systems generally better suited for the needs of RPGs than the d20.