DICE PROBABILITY

Which total comes up most with two or three dice, how often you roll a 6, doubles and the d20 with advantage: the exact math and what happened in a million rolls.

By the person behind Decide for me · Updated

With one die, every face is equally likely: 1 in 6, or 16.67%. With two or more dice that changes: some totals come up far more than others, which is what makes games like Catan, tabletop RPGs and craps work. This guide has the exact tables and, next to them, what came up when we rolled a million times with this site's dice roller.

One die: 1 in 6

A fair six-sided die gives 16.67% to each number. In our million rolls, each face came up between 16.62% and 16.70% of the time: differences of thousandths, exactly what randomness looks like. The probability of rolling a 6 on one roll is 1/6; of not rolling it, 5/6.

Two dice: the 2d6 probability table

Two dice give 36 equally likely combinations (6 × 6). The probability of each total is the number of combinations that make it, divided by 36:

SumWaysProbabilityMeasured
211/36 · 2.78%2.81%
321/18 · 5.56%5.56%
431/12 · 8.33%8.29%
541/9 · 11.11%11.17%
655/36 · 13.89%13.86%
761/6 · 16.67%16.70%
855/36 · 13.89%13.92%
941/9 · 11.11%11.05%
1031/12 · 8.33%8.34%
1121/18 · 5.56%5.52%
1211/36 · 2.78%2.78%

The “Measured” column is what came up in a million rolls of two dice. The curve is a triangle: the extremes (2 and 12) can only be made one way, and the middle six ways.

Why does 7 come up so often?

Because it has the most combinations: 1+6, 2+5, 3+4, 4+3, 5+2 and 6+1. That's 6 out of 36, 1 roll in 6 (16.67%); in the experiment, 16.70%. That's why the robber moves on a 7 in Catan, and why the 6 and 8 tokens have the most dots (5/36 each, 13.89%).

Common two-dice questions

Three dice: the 3d6 bell curve

Three dice give 216 combinations (6 × 6 × 6) and the shape already looks like a bell: the middle totals, 10 and 11, come up 12.50% of the time each, and the extremes (3 and 18) only 0.46%.

SumWaysProbabilityMeasured
310.46%0.46%
431.39%1.39%
562.78%2.78%
6104.63%4.63%
7156.94%6.92%
8219.72%9.74%
92511.57%11.58%
102712.50%12.52%
112712.50%12.53%
122511.57%11.51%
13219.72%9.74%
14156.94%6.95%
15104.63%4.62%
1662.78%2.77%
1731.39%1.40%
1810.46%0.46%

Rolling three of a kind is 6/216 = 1/36, 2.78% (measured: 2.77%).

At least one 6 in several rolls

For “at least one”, work out the opposite: the chance of rolling no 6 in n rolls is (5/6)n, so the chance of at least one is 1 − (5/6)n. Rolling one die n times is the same as rolling n dice at once.

RollsAt least one 6
116.7%
230.6%
342.1%
451.8%
559.8%
666.5%

The d20 with advantage and disadvantage

In Dungeons & Dragons, with advantage you roll two d20s and keep the higher; with disadvantage, the lower. The average goes from 10.5 to 13.82 with advantage and 7.17 with disadvantage (measured over a million rolls). This is how the chance of beating each target changes (measured in brackets):

Roll at leastNormalAdvantageDisadvantage
580.0% (80.0%)96.0% (96.0%)64.0% (64.0%)
1055.0% (55.0%)79.8% (79.7%)30.3% (30.2%)
1530.0% (29.9%)51.0% (51.0%)9.0% (9.0%)
205.0% (5.0%)9.8% (9.8%)0.3% (0.2%)

Advantage is worth the most for middling targets: rolling 11 or more goes from 50% to 75%. A natural 20 goes from 5% to 9.75%.

The problem that invented probability

In 1654 the Chevalier de Méré, a French gambler, brought Blaise Pascal a puzzle: he won money betting on at least one 6 in 4 rolls of a die, but lost betting on at least one double six in 24 rolls of two dice, even though by his reckoning the two bets were the same. Pascal and Pierre de Fermat solved it by letter and, along the way, laid the foundations of probability theory. The first bet wins 51.77% of the time and the second 49.14%: one in his favor, one against. In our million games of each, they came up 51.74% and 49.16%.

How to calculate dice probability

  1. Count the possible combinations: 6 with one die, 36 with two, 216 with three (6 to the power of the number of dice).
  2. Count the ones that give what you want (for example, the six that make 7).
  3. Divide favorable by possible. For “at least one”, use 1 minus the chance of none.

Other dice work the same way: a d8 gives 1/8 to each face, a d12 1/12, a d20 1/20.

The experiment: a million rolls

Every “measured” figure comes from rolls made in the browser with the same generator this site's dice roller uses: the browser's cryptographic generator, with each die decided by 1 + Math.floor(Math.random() × faces) (on this site, Math.random draws its numbers from that generator). A million rolls for each case, with nothing filtered or repeated. The chi-square test gives p-values of 0.634, 0.066, 0.902, 0.401 (one die, two, three and the d20): none below 0.05, meaning the differences from theory are ordinary randomness. The full counts are in this data file, free to cite; and the test of every tool is in is the wheel really random?

Frequently asked questions

What is the probability of rolling a 7 with two dice?

It is 6/36 = 1/6, or 16.67%. It is the most likely total with two dice because it can be made six ways: 1+6, 2+5, 3+4, 4+3, 5+2 and 6+1.

What is the probability of rolling a 6 on one die?

It is 1/6, or 16.67%. Over two rolls, the chance of at least one 6 is 11/36 (30.56%); over four rolls, 51.77%.

What are the odds of rolling doubles with two dice?

It is 6/36 = 1/6, or 16.67%. A specific double, such as double six or snake eyes, is 1/36: 2.78%.

Which total is most likely with 3d6?

10 and 11, each with 27 of the 216 combinations: 12.50%. The least likely are 3 and 18, 0.46% each.

How much does advantage help on a d20?

The average goes up from 10.5 to about 13.8. Rolling 11 or more goes from 50% to 75%, and a natural 20 from 5% to 9.75%.

Are online dice fair?

This site's dice use the browser's cryptographic generator. Over a million rolls, each face came up between 16.62% and 16.70% of the time, within what randomness gives.

Want to check? Roll the dice: 4 to 100 sides, with a modifier and history. And for coins, the history and probability of the coin flip.

Keep reading