Adding Dice Narrows It
Sum two dice and you get a triangle. Sum three and you get a bell. Every die added makes the middle more probable and the edges rarer than most players ever realize.
01The Shape That Grows in the Middle
A single die is flat: every face appears with identical frequency. Roll a d6 and each number from one to six comes up roughly one time in six. Now roll two d6s and add them. The result can only be two or twelve one way — snake eyes, boxcars — but it can be seven in six different ways. A triangle emerges, peaked at the midpoint, sloping symmetrically down toward each extreme.
Key numbers
- 011d6: six outcomes, each at ~16.7% probability — perfectly flat
- 022d6: eleven outcomes; seven is the peak at ~16.7%, two and twelve each at ~2.8%
- 033d6: range 3–18; ten and eleven together account for roughly 25% of all results; three and eighteen each appear about 0.46% of the time
- 04Standard deviation scales as the square root of the number of dice — doubling dice shrinks relative spread, not absolute spread
Add a third die and that triangle rounds into something close to a bell curve. Four dice, and the bell narrows further and rises higher at the center. The technical name for what is happening is the central limit theorem at work: independent, identically distributed random variables, when summed, converge toward a normal distribution as the count grows. In practice, this means that 3d6 — the classic attribute-generation roll in many games — does not merely produce a different average than a d20; it produces a fundamentally different shape. Results near eighteen or near three are genuinely unusual. Results near ten and eleven are crowded.
This is the mechanism that makes dice pools feel reliable at scale. A player rolling fifteen dice to count successes is effectively sampling from a much narrower band of outcomes than a player rolling one die, even if the expected value per die is identical. More dice does not mean more chaos — it means less of it.
Systems that want wild outcomes cling to the flat line.
What narrows is the spread. Statisticians measure this as standard deviation, which grows with added dice but grows slower than the range does. When you move from 1d6 to 3d6, the possible range triples — three to eighteen versus one to six — but the standard deviation only grows by the square root of three, roughly 1.73 times, not three. The middle becomes proportionally dominant.
What it means at the table
- 01Ability-score generation on 3d6 makes scores below 6 or above 15 uncommon events, each side turning up under 5% of the time
- 02A single-die system lets a dragon and a farmer share an equal chance of a freak outcome; a summed-dice system makes that outcome structurally inaccessible for most rolls
- 03Switching from "roll one die" to "roll several and sum" is not a modifier — it is a change to the probability architecture
Designers exploit this deliberately. Systems that want consistent heroes use summed pools rather than single dice. Systems that want wild outcomes cling to the flat line. The choice is not aesthetic preference; it is a direct claim about how often exceptional things happen. The extreme faces of a single d20, a 1 or a 20, come up one roll in ten. The extreme totals on 3d6, a 3 or an 18, come up roughly one roll in a hundred, which is functionally unreachable in normal play.
Add dice to stabilize. Drop to one die to open the range. The shape is the decision.