Dice Probability Basics for Board-Game Night

A single die: every face is equally likely

For a single fair die of any size -- d4, d6, d8, d10, d12, or d20 -- every face has an identical probability: 1 divided by the number of faces. A d6 gives each number 1 through 6 a 1-in-6 (about 16.7%) chance; a d20 gives each number 1 through 20 an exact 1-in-20 (5%) chance. This is true regardless of what you rolled last, since each roll is fully independent -- rolling a 6 five times in a row doesn't make a 6 any less likely on the sixth roll, a common misconception known as the gambler's fallacy.

For this to hold in practice, the die itself has to be genuinely fair, which is harder to guarantee physically than it sounds -- manufacturing imperfections, rounded edges, and uneven weight distribution can all introduce small real biases in physical dice. A digital dice roller built on rejection sampling avoids that specific category of problem, guaranteeing each face a genuinely identical probability regardless of the die size.

Two dice together: why 7 is the most common roll

Roll two six-sided dice and add them together, and the result isn't flat across 2 through 12 -- it clusters around 7, the most statistically likely total. That's because there are more ways to make certain sums than others: only one combination makes 2 (1+1) and only one makes 12 (6+6), but six different combinations make 7 (1+6, 2+5, 3+4, 4+3, 5+2, 6+1). Each individual die remains perfectly fair on its own; it's the number of combinations that add up to each possible total that creates the bell-curve-shaped distribution of the sum.

This matters practically in games like Settlers of Catan or Monopoly, where knowing that 7 comes up roughly 1 in 6 rolls (versus 2 or 12, each roughly 1 in 36) is genuinely useful strategic information, not just trivia.

Advantage, disadvantage, and rolling multiple dice

Games like Dungeons & Dragons use mechanics like 'advantage' (roll two d20s, keep the higher) and 'disadvantage' (roll two, keep the lower), which meaningfully shift the odds compared to a single roll. Rolling with advantage roughly doubles your chance of hitting a high number like 18+ compared to a single d20 roll, while disadvantage does the reverse. Understanding this is part of why exact per-face fairness on each individual die matters so much -- if the underlying single-die rolls carried even a small statistical skew, mechanics built on top of them (advantage, disadvantage, damage rolls with modifiers) would compound that skew rather than average it out.

Common misconceptions worth clearing up

A streak of the same number doesn't make a different number 'due' -- each roll is independent, full stop. A d20 rolling a natural 20 twice in a session isn't suspicious on its own; with enough rolls across a long campaign, that kind of repeat is expected to happen occasionally by pure chance. And 'hot' or 'cold' dice -- the belief that a specific physical die is currently rolling well or poorly -- isn't statistically meaningful for a genuinely fair die; any apparent streak is just the normal variance you'd expect from independent random events, the same variance a coin flip or a digital dice roller shows over a small sample.

Why exact d20 fairness matters more than it might seem

In systems like Dungeons & Dragons, a natural 20 and a natural 1 on a d20 roll trigger special rules -- critical success and critical failure -- that are specifically designed around an exact 1-in-20 probability for each. If the underlying roll carried even a small statistical bias favoring, say, numbers 1 through 16 over 17 through 20 (exactly the kind of bias a naive byte-modulo-20 approach would introduce), the practical rate of critical hits and misses would drift over a long campaign in a way players would eventually start to notice, even if no single roll looked obviously wrong. This is the concrete, game-relevant reason rejection sampling isn't just an abstract mathematical nicety for a dice tool -- it's the difference between a die that's actually fair over hundreds of rolls and one that only looks fair on a small sample.

Probability versus 'feel': why variance is normal

A genuinely fair die will still occasionally produce results that feel wrong -- three natural 1s in a row, or a long cold streak on important rolls. That's not a sign of a broken die; it's the expected behavior of independent random events over a small sample. The gambler's fallacy -- expecting a 'correction' after a streak -- is one of the most persistent misconceptions in casual probability, and it's worth remembering specifically at the table, where a bad roll can feel personal even though it carries zero information about what the next roll will do.

Frequently asked questions