How dice probability works: average is only the start

Choose the event you care about before comparing dice averages. A higher average does not describe every target's success chance. Selection rules also change the distribution's shape, even when the dice themselves remain fair.

This guide assumes independent rolls and equally likely faces on each die. It treats notation as a complete rule for producing a total. A game can add critical hits, automatic failures, or other rules that this total does not capture.

Average describes repeated rolls, not the next roll

Expected value weights each possible total by its probability. In symbols, E[X] = Σ xP(X=x). A total with more probability contributes more to the average. OpenStax's chapter on expected value and standard deviation explains this calculation.

For a fair die with faces from 1 through s, the mean is (s+1)/2. Independent dice added together contribute their means to the total. Adding a fixed bonus shifts the mean without changing the spread.

Enter 1d6 in the dice average calculator. It returns the exact mean 7/2, shown as 3.5000. That average is not a possible face on the die. It describes the average across repeated rolls under the model.

For 2d6, the calculator returns 7/1, shown as 7.0000. Here, the mean also matches the most likely total, 7. That coincidence does not make the next roll predictable. The chance of exactly that total is 16.6667% (1/6).

Spread explains how far totals move around the mean

Variance weights each squared distance from the mean by its probability. Standard deviation takes the square root, returning to the total's units. The calculator reports the distribution's standard deviation, not an estimate from simulated rolls.

For 1d6, standard deviation is approximately 1.7078. For 2d6, it is approximately 2.4152. The summed dice have a larger absolute spread, even though their middle totals receive more probability than their extremes.

Do not confuse a bell-like shape with a guarantee of small deviations. Read the actual target probability when a particular total matters. Use standard deviation to describe dispersion, then inspect the distribution to understand where that dispersion occurs.

Advantage selects a maximum instead of adding dice

In this calculator, 1d20adv rolls independent copies of a d20 and keeps the higher result. The equivalent expression is 2d20kh1. Keeping the higher result differs from adding both results.

Let F(x) mean P(X≤x) for the ordinary roll. Both independent rolls finish at or below x with probability F(x)². Therefore, the maximum has cumulative probability F(x)². LibreTexts' order-statistics chapter gives this maximum rule.

For an integer target t, success with advantage has probability 1−F(t−1)². This expression counts the complement of both rolls failing. It avoids double-counting outcomes where both rolls succeed.

Set the target to 15. The ordinary 1d20 succeeds with probability 30.0000% (3/10). The advantage expression succeeds with probability 51.0000% (51/100). Its exact mean is 553/40, shown as 13.8250.

Do not replace advantage with a fixed bonus for every question. A fixed bonus moves the distribution without selecting outcomes. Advantage changes which totals receive probability, so its benefit depends on the target.

The calculator applies adv to the preceding dice group. On a group containing several dice, it compares independent group totals. Check your game's wording before applying that convention to a compound expression.

Keep-highest changes which dice contribute

Enter 4d6kh3 to roll four six-sided dice and add the highest three. The equivalent expression 4d6dl1 drops the lowest die. Ties do not change the total because tied faces have equal values.

The calculator returns 15869/1296 as the exact mean, or 12.2446 after rounding. Its possible totals run from 3 through 18. The most likely total is 13. The probability of reaching at least 15 is 23.1481% (25/108).

Sorting the rolls changes the distributions of the retained dice. You cannot treat each retained die as an untouched fair die. You also cannot subtract an ordinary die's mean to represent dropping the minimum.

Use the exact distribution when comparing character-generation or damage rules. First confirm whether the rule keeps individual dice or selects a whole group total. Those operations answer different probability questions.

Reroll once means the replacement can still be low

Enter 1d6r1 to reroll an initial face of 1 once. Keep the replacement even when it also shows 1. This differs from rerolling repeatedly until the face changes.

The final face can arise through separate paths. A non-triggering initial face stays. A triggering face leads to an independent replacement roll. Add these mutually exclusive paths to obtain the final face's probability.

OpenStax's multiplication and addition rules justify multiplying independent roll probabilities and adding disjoint paths. A branch diagram must retain the initial trigger event; it cannot assign equal probability to every final face.

The calculator gives 2.7778% (1/36) for a final face of 1. For a final face of 2, it gives 19.4444% (7/36). The other non-triggering faces share that latter probability. The exact mean becomes 47/12, shown as 3.9167.

The modifier r<=V rerolls an initial face at or below V once. Write the threshold and reroll count explicitly when sharing a calculation. A rule that permits repeated rerolls needs a different model.

Common mistakes when reading exact results

• Reading “at least” as “greater than” changes whether the target itself succeeds. The calculator includes the target.
• Averaging possible totals equally ignores their different probabilities. Summed dice usually have more paths to middle totals.
• Adding standard deviations gives the wrong spread for independent sums. Add their variances, then take the square root.
• Confusing rounded decimals with exact fractions loses precision. Keep the fraction when comparing close results.
• Treating a game label as a probability specification hides rule differences. Record the expression and the target beside the result.

These examples use exact finite distributions rather than simulation. Their displayed standard deviations remain rounded decimals. If you add exploding dice, this calculator caps explosions after four extra rolls per die. Its probabilities then describe that capped rule, not unlimited explosions.