Dice Roll Probability Explained: The Odds for One Die, Two Dice and Beyond

Dice are one of the oldest tools for generating a random outcome, and also one of the clearest ways to see basic probability in action, because every possible result can be counted by hand rather than estimated. This piece works through dice roll probability from a single die up to the full distribution of rolling two dice together, with every count verified rather than just asserted.

Dice roll probability: why a single die is the cleanest example

A standard six-sided die, a d6, has six faces, each showing a different one of the numbers 1 through 6. If the die is fair, meaning no face is weighted or shaped to land more often than another, every one of those six outcomes is exactly equally likely. That gives each face a probability of 1 divided by 6, which works out to roughly 16.67%, or a little better than a 1-in-6 chance.

This is what statisticians call a uniform distribution: every possible outcome shares the same probability, with none favored over the others. It is the simplest possible probability distribution to reason about, which is exactly why a single die roll shows up in almost every introductory probability lesson. Rolling the die does not change this probability from one roll to the next either. Each roll is an independent event, meaning the outcome of the current roll has no memory of, and no influence from, any roll that came before it. A die that has just landed on 6 three times in a row is still exactly as likely to land on 6 a fourth time as it was on the very first roll, a fact that runs against a common but mistaken intuition called the gambler's fallacy, the belief that a result becomes "due" after not appearing for a while.

Rolling two dice: why the sums are not equally likely

The moment a second die enters the picture, the outcome that matters most often is not either individual face but the two faces added together, and this is where dice roll probability becomes genuinely more interesting than a single die's flat, uniform spread.

Each of the two dice still has 6 equally likely faces on its own, and since the two dice are independent of each other, the number of distinct ordered combinations across both dice is 6 multiplied by 6, or 36 total combinations, each exactly as likely as any other individual combination. A roll of (1, 4) and a roll of (4, 1), for example, count as two separate combinations even though they add up to the same sum, because the two dice are still distinguishable outcomes even when rolled at the same time.

Sums, though, are not spread evenly across those 36 combinations, because different sums can be reached by different numbers of combinations. A sum of 2 can only happen one way, both dice showing a 1. A sum of 7, by contrast, can happen six different ways. More paths to the same total means a higher probability for that total, which is the entire reason the sums on two dice are not uniformly distributed even though each individual die still is.

The full 2d6 distribution, counted by hand

Counting every one of the 36 combinations and grouping them by sum produces the following table, which can be verified directly by listing out the ordered pairs for each total:

Adding up every count in that table, 1 + 2 + 3 + 4 + 5 + 6 + 5 + 4 + 3 + 2 + 1, comes to exactly 36, matching the total number of possible combinations, which confirms every combination has been accounted for exactly once.

Why 7 is the single most common roll

The table above makes the reason plain: 7 sits at the exact middle of the possible range from 2 to 12, and every sum near the middle of that range has more combinations that can produce it than a sum near either extreme. There are six distinct ways to make 7, more than any other total, giving it a 6-in-36 probability, which simplifies to exactly 1 in 6, or about 16.67%. That is the same probability as any single face on one die, an easy-to-miss coincidence rather than a deeper mathematical connection, since the two situations are counting entirely different kinds of events.

Compare that to the two least likely sums, 2 and 12, each reachable only one way and each carrying a probability of 1 in 36, about 2.78%, roughly six times less likely than rolling a 7. This lopsided shape, rising from the extremes toward a peak at the middle, is the beginning of what statisticians call a discrete triangular distribution, an early cousin of the smoother bell-curve shape that shows up when even more dice are added together.

This exact fact, that 7 is the single most likely sum on two standard dice, is the mathematical foundation behind the dice game Craps, where a roll of 7 on the first throw is an automatic win and a roll of 7 after a point number has been set is an automatic loss, precisely because 7 comes up more often than any other total.

The expected value of a two-dice roll

Beyond just knowing which sum is most likely, probability also gives a precise answer for what a two-dice roll averages out to over many throws, a figure called the expected value. It is calculated by multiplying each possible sum by its own probability and adding all eleven products together:

(2×1 + 3×2 + 4×3 + 5×4 + 6×5 + 7×6 + 8×5 + 9×4 + 10×3 + 11×2 + 12×1) ÷ 36 = (2 + 6 + 12 + 20 + 30 + 42 + 40 + 36 + 30 + 22 + 12) ÷ 36 = 252 ÷ 36 = 7

The expected value works out to exactly 7, the same number that is also the single most likely individual outcome. That is not a coincidence baked into two dice specifically, it follows directly from the distribution's symmetry: every sum has a mirror-image sum on the opposite side of 7 with an identical probability (3 and 11 each carry a 2/36 chance, 4 and 10 each carry 3/36, and so on), so the probability-weighted average has to land exactly on the point those pairs are balanced around. Roll two dice a handful of times and the total will bounce around unpredictably; roll them a few thousand times and record the average of every roll, and that average will drift closer and closer to 7.

A brief history: dice and the birth of probability theory

Dice are ancient, with bone and ivory examples recovered from archaeological sites thousands of years old, but the mathematics of dice probability is comparatively recent. The Italian physician and mathematician Gerolamo Cardano wrote one of the earliest known systematic treatments of dice odds in the 16th century, working out combination counts similar to the 2d6 table above well before probability existed as a formal branch of mathematics. A century later, a French nobleman's gambling question about dice, passed to the mathematicians Blaise Pascal and Pierre de Fermat, led to a written correspondence between the two that many historians credit as the founding exchange of modern probability theory. The specific question that started it, essentially how to fairly split a prize between two players when a dice game is interrupted early, only makes sense to answer once the exact odds behind each possible dice outcome are known, which is exactly the kind of counting problem worked through earlier in this piece.

That history is worth knowing for a practical reason as much as a historical one: dice probability is not an abstract or modern invention layered on top of an old game. Every combination-counting method used to build the distribution table above is the same basic technique mathematicians were already using centuries ago to settle real disputes over dice games, just organized more systematically today.

Fair dice versus loaded or biased dice

Every probability figure in this article assumes a fair die, one where manufacturing has not given any face an advantage over any other. Real physical dice can fall short of that ideal in small ways: a die drilled with pips loses a tiny amount of material from the faces with more pips (typically 6, and to a lesser extent 5), which can shift its center of mass just slightly away from perfect and bias results, on average, a very small amount over a very large number of rolls. Casino-grade dice are manufactured to tight tolerances specifically to minimize this, with pips filled flush using a material matched in density to the die itself, precisely to keep every face's probability as close to the ideal 1-in-6 as manufacturing allows.

A "loaded" die is a different matter entirely: a die deliberately weighted or shaved to favor a particular face, used historically to cheat at dice games. Detecting a loaded die is itself a probability problem, essentially rolling it enough times to notice that the actual frequency of each face has drifted further from the expected 1-in-6 than random chance alone would plausibly explain. A well-built digital dice roller sidesteps physical manufacturing concerns entirely, since each face's probability is set directly by the code generating the roll rather than by the physical shape and weight distribution of an object (see the dice roller's own page for exactly how that random draw is generated and scaled).

Beyond two d6: polyhedral dice and flatter distributions

Tabletop role-playing games routinely reach for dice with more than six sides, a d8, d10, d12, or the signature d20, and the probability behavior of a single one of these dice works exactly like a single d6: every face on a fair die of any number of sides is equally likely, giving a flat, uniform distribution with no favored outcome. A d20 gives each number from 1 to 20 an even 1-in-20 chance, roughly 5%, which is one reason game designers favor a single d20 roll for resolving actions: the flat distribution is simple to design difficulty thresholds around and easy for a player to reason about mid-game, unlike the bell-curve-shaped spread that summing multiple dice produces.

Combining multiple dice, on the other hand, always pulls the distribution back toward that bell-curve shape, the same way two d6s pulled toward a peak at 7. A percentile roll, generated either from two d10s read as tens and ones digits or, in a digital tool, directly from a single roll across 1 to 100, is a special case worth knowing: because it is built from independent digit positions rather than a true sum, a percentile roll actually stays close to a flat, uniform distribution across the full 1-to-100 range, unlike a genuine sum of multiple dice.

Dice pools in other games: Yahtzee, backgammon, and Craps

The same combination-counting approach that built the 2d6 table above extends to more complex dice games, even when the full calculation gets considerably more involved. Yahtzee rolls five dice at once and scores specific combinations, like five of a kind or a full house, and the probability of landing any one of those combinations on a single roll of five dice can be worked out the same way, by counting how many of the possible outcomes (6 to the 5th power, or 7,776 total combinations for five distinct six-sided dice) satisfy that particular pattern. Backgammon uses two dice exactly like the table above, and doubles (rolling the same number on both dice) carry the same combined significance the game assigns them, four moves instead of two, precisely because a specific double is one of the 36 equally likely combinations, no more and no less special than any other single combination despite feeling notable when it lands.

Craps, mentioned earlier for why a 7 carries such weight in the game's structure, is built almost entirely around the 2d6 distribution table above. The house edge on nearly every bet at a Craps table can be calculated directly from that table: a bet that pays out only when a specific, low-probability sum like 2 or 12 comes up before a 7 is priced by the casino using exactly the same 36-combination counting method used earlier in this piece, just applied to a slightly more complex sequence of rolls rather than a single one.

The gambler's fallacy and independent trials

It bears repeating because it trips up intuition so reliably: each dice roll is an independent trial. A die, virtual or physical, retains no memory of its previous results. If a single d6 has landed on 6 five times in a row, an entirely genuine, if statistically uncommon, run of results, the probability of a sixth 6 is still exactly 1 in 6, identical to every roll before it. The belief that a different number is somehow "due" after a streak is called the gambler's fallacy, and it applies just as directly to two dice added together: a string of low sums does not make a 7 more or less likely on the next roll, because the two dice being thrown have no awareness of what came before.

Dice probability inside real games, not just textbooks

The counting methods above are not only a classroom exercise. Board and tabletop games build their rules directly on top of exactly this kind of probability, whether or not the designer ever writes the math down.

Monopoly moves every player using the sum of two standard d6, so the same 2d6 table built earlier in this piece governs how far a typical turn travels: a roll of 7 is the single most likely move, and a roll of 2 or 12 is the rarest. Rolling doubles grants an extra turn, but three doubles in a row sends a player directly to jail instead, a rule that only makes sense once you know how often doubles come up in the first place. That answer is a genuinely different question from the odds of any one specific double: while a specific pair like (4, 4) is only 1 of the 36 combinations, doubles of any kind, (1,1) through (6,6), cover six of those 36 combinations, for a 6/36, or exactly 1-in-6, chance on a given roll.

Yahtzee scales the same counting idea up to five dice at once, 6 to the 5th power, or 7,776 total combinations, and scores specific patterns among them. A "full house," three of one number plus two of another, and "five of a kind," the game's namesake Yahtzee, are both counted by working out how many of those 7,776 combinations match the pattern, exactly the same technique used to build the 2d6 table, just applied to a larger set of dice.

Dungeons & Dragons' well-known "advantage" mechanic, rolling two d20s and keeping the higher result, is a clean example of a probability tree: two independent branches, each with 20 equally likely outcomes, combine into 400 total paths. The odds of rolling at least one 20 with advantage work out to 1 minus the odds of missing on both dice, 1 − (19/20 × 19/20) = 1 − 361/400 = 39/400, about 9.75%, nearly double a single d20's flat 5% chance of the same result.

Try it yourself

Every dice roll probability worked through in this piece is a mathematical prediction about what happens over a very large number of rolls, not a guarantee about what any single roll or short run of rolls will show. The dice roller on this site includes a quick two-die tool built for exactly this kind of hands-on check: roll it repeatedly, tally how often each sum from 2 to 12 comes up, and watch the results gradually settle toward the 36-combination distribution worked through above the more rolls you record. A short run will look streaky and uneven almost by definition; a long enough run will not.