Dice Average Calculator

Every dice pool has a centre of gravity, and the game turns on knowing it. Enter how many dice you roll and how many sides they have, and this calculator returns the average total, the spread around it, and the chance of rolling one exact total — computed from the full distribution, not simulated or guessed.

Why Averages of Dice Are Exact (and Delightfully Small)

A single fair die's average is the mean of its faces — (1 + 2 + … + S) ÷ S = (S + 1) ÷ 2: 3.5 for the household d6. Pools obey the statistician's friendliest theorem, the linearity of expectation: the average of a sum is the sum of the averages, no matter how the dice tumble together, so N dice average N × (S + 1) ÷ 2. Spread is nearly as cooperative: a die's variance is (S² − 1) ÷ 12, independent dice add their variances, and the pool's standard deviation is the square root of N × (S² − 1) ÷ 12. The third actor — the chance of an exact total — refuses a formula and demands the count the widget performs: convolve the distributions (build every total's number of ways, die by die) and divide by S^N. That enumeration is also why the calculator caps its inputs: 40 dice of 100 sides is eleven quintillion outcomes counted exactly by a dynamic program that tallies ways, not outcomes — arithmetic on integers, exact to the last roll.

Worked Example: The Immortal 2d6

Two six-siders — the atoms of board-game probability. Average: 2 × 3.5 = 7. Variance: 2 × 35/12 ≈ 5.83, so the standard deviation is 2.42 and typical rolls crowd between about 4.6 and 9.4. The distribution's shape is the famous triangle: 36 equally likely outcomes, and the totals' ways run 1, 2, 3, 4, 5, 6, 5, 4, 3, 2, 1 for 2 through 12 — the calculator's exact-total line prices the peak: exactly 7 in 6 of 36 outcomes = 16.67%, while the glamorous 12 (and the snake-eyes 2) manage 1 of 36, 2.78% each. Every craps bet, every Monopoly expectation and every backgammon doubling decision is this triangle wearing different clothes, and the Settlers of Catan board prints it in dots under each number — the game's designer conceding that the distribution is the strategy. Players who internalise the triangle stop being surprised by the game and start being surprised only by the dice, which is the correct division of astonishment.

The Bell Arrives by the Third Die

Watch the distribution morph as dice accumulate: one die is flat (every face equally likely — an average with no favourite total), two dice form the triangle, three already round into the bell that the central limit theorem promises, and by five or six dice the pool behaves like a small normal distribution centred on its mean with the calculator's standard deviation. The practical consequence for gamers is counterintuitive and total: more dice means more predictable totals — relative spread shrinks as the pool grows (the standard deviation grows like √N while the mean grows like N), which is why game designers reach for dice pools when they want skill to matter and for single dice when they want drama. The same theorem is why this site's probability pages converge on the normal curve from every direction: dice are where most people first meet it, in a casino or at a kitchen table, long before a classroom names it. There is a quiet delight in that ordering — the bell curve, statistics' most consequential shape, arrives in most lives not as a formula but as the felt tendency of two dice to make sevens and of three dice to make tens and elevens, experience preceding notation by years.

Dice in the Wild: Where the Triangle Earns Its Living

The 2d6 triangle quietly runs a surprising share of the world's leisure: craps is built entirely from its probabilities (the pass line's 49.3% win rate is the triangle summed with the game's point rules), Monopoly's movement expectations — and therefore its property valuations — are doubles-inclusive variations on it, and backgammon's doubling cube prices positions whose equity is dice-distribution arithmetic all the way down. Beyond the table, the same convolution machinery prices insurance layers, inventory buffers and any sum of independent small risks: actuaries call it the aggregate distribution method and charge accordingly for what this page does to dice for free. The pedagogical moral is worth the section: learn one enumeration honestly, on objects you can hold and shake, and a startling fraction of applied probability turns out to be that enumeration with the dice painted over and the stakes denominated in money.

Loaded Dice and Honest Audits

Every formula on this page assumes fair dice, and the assumption is itself a testable hypothesis — casinos audit for it industriously, because a die biased by a single pip's weight shifts its average measurably over ten thousand rolls. The audit runs this page's arithmetic in reverse: roll the suspect die a few thousand times, compare its observed average and face frequencies against (S+1)/2 and uniformity (a chi-square test, whose page lives elsewhere on this hub), and the die confesses or clears itself. The same machinery audits random number generators, shuffled playlists and draft lotteries — anywhere fairness is a claim with data available. A calculator of averages turns out to be an instrument of suspicion as readily as of expectation, which is true of most honest statistics and nearly all interesting ones.

Related Tools

More from the Statistics Calculators hub: the Dice Probability Calculator for at-least and at-most questions on the same pools, and the Roulette Payout Calculator — the casino's other distribution, priced the same honest way.

Frequently Asked Questions

What is the average roll of a d20, exactly?

(20 + 1) ÷ 2 = 10.5 — and the half-point matters: a d20's average sits between faces, which is why advantage/disadvantage mechanics (roll two, take the better or worse) move the effective average so dramatically (to about 13.8 for advantage: the maximum of two dice averages far above the single die, a computation this page's enumeration performs if you ask it about pools of a different kind). The d20's flat distribution also makes it the drama die of the earlier section — every face 5%, no centre of gravity worth the name, one roll deciding everything.

Why is 7 the most common 2d6 roll, then?

Because six of the 36 outcomes compose it — (1,6), (2,5), (3,4), (4,3), (5,2), (6,1) — more decompositions than any other total: the middle total pairs every face with its complement to 7, and the count of pairings peaks where the two dice's ranges fully overlap. The calculator's convolution literally counts these decompositions; the triangle's symmetry around 7 is the arithmetic of two identical ranges sliding across each other. Craps built its pass line on the number's frequency and its mystique on the same fact — the seven that makes you also breaks you, by the identical 6-of-36.

Does the average change if the dice are different sizes entirely?

The means still add — a d6 plus a d8 averages 3.5 + 4.5 = 8.0 — because expectation's linearity never asked the dice to match. The distribution's enumeration grows more interesting (mixed pools convolve different ranges, producing trapezoid middles), and this calculator's same-size design covers the standard pools of actual play; for a mixed pool, adding the individual averages gives the centre exactly, while the exact-total probabilities want the mixed convolution a dedicated tool should perform. Knowing which questions survive a simplification is most of applied probability's craft.

How do exploding dice and rerolls change the average?

They bend the linearity politely: an exploding die (reroll and add on a maximum face) averages (S+1)/2 × S/(S−1) — a d6 that explodes averages 4.2 instead of 3.5 — because each face's contribution gains the geometric tail of possible explosions; reroll-ones mechanics add their own computable increments. These variants are exactly enumerable by the same convolution with modified face weights, and their lesson generalises: any dice rule you can state precisely, this machinery can price precisely — the game's flavour text is a distribution in costume.

What does "3d6" average, and why do role-players care?

3 × 3.5 = 10.5, with a standard deviation of about 2.96 — the distribution that generated generations of ability scores sits between 3 and 18, bulging at 10 and 11 (the most likely totals), with 18 arriving in just 1 of 216 rolls (0.46%). The bell's kindness is the point: characters cluster near average with heroes and the hopeless at the tails, a population model in three dice. The calculator prices the whole pool as written; the culture that grew around 10.5 is a reminder that distributions are worldbuilding.

Average versus most likely — why do they differ on some pools?

On symmetric pools they coincide or bracket (2d6's mean 7 is its mode; 3d6's mean 10.5 splits its twin modes 10 and 11), and on skewed constructions — mixed dice, exploding rules, drop-the-lowest mechanics — they part company meaningfully: drop-lowest pools (4d6, discard the minimum) average about 12.24 while their most likely total sits at 13, the rule having sawn off the distribution's left tail. The calculator reports both for its pools precisely so the difference stays visible: the average is the long-run settlement price, the mode is the single most probable guest, and confusing them is the original sin of gambling intuition.