Roll D10
One ten-sided die. Each face is a clean 10%, which makes the d10 the easiest die to reason about, and two of them together give a flat percentage from 1 to 100.
1d10 1
Press roll.
Lowest possible 1, highest 10, average 5.5.
10% a face, average 5.5
Ten faces, one tenth each. No number is favoured, and the long-run average is 5.5, a value the die cannot actually show.
Two d10 make a percentage
Read one die as tens and the other as units and you get 100 outcomes, each worth exactly 1%. That is the whole basis of percentile systems, and it is identical in distribution to a single d100. If you only need a percentage, roll two of these rather than hunting for a hundred-sided die.
Ten rolls will not give you ten different numbers
This is the one people get wrong. The chance that ten rolls produce each face exactly once is 0.036%, about one time in 2,756. What actually happens is repeats:
- In four rolls, the chance of a repeat is 49.6%, near enough a coin flip.
- After ten rolls, any particular face is still missing 34.9% of the time.
- It takes about 29 rolls on average before all ten faces have shown up.
So a d10 that has not produced a 7 all evening is behaving normally. There is no memory in it and no correction coming.
Where the d10 gets used
Dice pools in Storyteller-style systems, where you roll a handful and count successes. Weapon damage. Percentile tables. Any ten-item list where each entry deserves the same chance.
For the same odds without dice, take a random number from 1 to 10, or pick a different die in the dice roller.