Standard dice notation with keep, drop, advantage, exploding dice and re-rolls. Shows the exact probability before you roll, and a seed makes any roll reproducible from a shared link.
Initializing in your browser…
Slide numbered tiles to combine and reach 2048! Features touch/swipe support, undo functionality, and score tracking
Classic tic-tac-toe game with AI opponent using minimax algorithm. Three difficulty levels: Easy, Medium, Hard
Classic snake game with three speed levels, pause/resume, mobile D-pad controls, and high score tracking
You are attacking with advantage and want to know your real chance against armour class 16.
Input
2d20kh1+7 · target 16
Output
Both d20s shown, lower struck through, total 24; chance of 16+ reported before you roll
Advantage is just 2d20kh1: roll two, keep the higher, which lifts the average from 10.5 to 13.825. Because keeping the highest has no simple closed form, the odds are simulated over twenty thousand throws and the panel says so rather than presenting an estimate as an exact figure. Both dice stay on screen so the arithmetic is checkable.
Roll dice using standard notation: 2d6+3, 4d6kh3, 2d20kh1 for advantage, exploding dice and re-rolls. Kept and dropped dice are both shown, the exact probability of the roll is worked out before you throw, and a seed makes any roll reproducible from a shared link.
Roll attacks, saves, damage and ability scores in the notation your rulebook already uses.
A full dice set on any device, including the ones you never own enough of.
See the real chance of hitting a number before you commit to a plan, rather than guessing from the average.
Use a seed so a roll made in a play-by-post game or over chat can be checked by everyone.
Replace lost dice, or show a class what a probability distribution actually looks like.
Enter 4d6kh3 and roll six times. All four dice show each time with the lowest struck through, so the arithmetic is visible rather than assumed. The odds panel puts the average at about 12.24, which is why this method produces better characters than 3d6.
Enter 2d20kh1+7 and set the target to your opponent's armour class. Both d20s appear with the lower one struck through, and the chance readout tells you how much advantage is actually worth against that number.
Set a seed such as the current date plus your character name, roll, and share the link. Anyone opening it sees the same rolls in the same order, so nobody has to take your word for it.
This reads standard dice notation, which is what anyone rolling for a game actually types. 2d6+3, 4d6kh3 for ability scores, 2d20kh1 for advantage and 2d20kl1 for disadvantage, 1d8+2d6-1 mixing several dice in one throw, d% for percentile, 3d6! for exploding dice, and 2d6r2+4 to re-roll ones and twos once. The previous version could only roll N dice of a single size with a flat modifier, so none of that was possible. What you type is echoed back in normalised form, and anything that cannot work is explained before you roll rather than after: asking to keep five of two dice says so, and the Roll button stays disabled.
Keep and drop are shown, not just applied. Roll 4d6kh3 and all four dice appear, with the discarded one struck through, so you can see the tool did what you asked. Exploding dice carry a marker and their whole chain in a tooltip, and re-rolled dice show what they were before. Maximum and minimum rolls are detected on the underlying faces, so an exploding die that happens to total twenty is not mistaken for a natural twenty.
Alongside the roll sits the probability of the roll you are about to make. For plain dice sums the distribution is computed exactly by convolving the dice, so 2d6 reports a mean of exactly 7 and a 1-in-6 chance of rolling one; nothing is estimated. Keep, drop, exploding and re-roll rules have no simple closed form, so those are simulated over twenty thousand throws, and the panel says which method it used rather than presenting an estimate as fact. A target field turns that into the number people actually want: the chance of hitting 15 with 1d20+5 reads 55 percent, because the die has to show 10 or more, which is 11 faces out of 20.
Randomness comes from crypto.getRandomValues with rejection sampling, so every face is equally likely with no modulo bias at any die size, verified by chi-square over sixty thousand rolls through the real code path. A seed field switches to a deterministic stream instead: enter a seed, share the link, and anyone opening it gets the identical sequence of rolls, which is how you settle a disagreement about a roll made online. History keeps the last forty rolls with the individual dice and a running average, and can be copied out as text.
NdS with an optional count, plus and minus modifiers, and several groups in one expression, so 1d8+2d6-1 works. Then kh and kl to keep the highest or lowest N, dh and dl to drop them, ! for exploding dice, rN to re-roll anything at or below N once, and d% for percentile. Whatever you type is echoed back in normalised form so you can confirm it was read correctly.
Enter 2d20kh1: roll two twenty-sided dice and keep the higher. Disadvantage is 2d20kl1. Both dice are shown with the discarded one struck through. Advantage raises the average from 10.5 to 13.825, which the odds panel shows.
Both, depending on what you asked for, and the panel says which. Plain dice sums are convolved exactly, so 2d6 gives a mean of exactly 7 and a probability of exactly 6/36 for a total of seven. Keep, drop, exploding and re-roll rules have no simple closed form, so those are simulated over twenty thousand rolls and labelled as estimated.
Yes. Values come from crypto.getRandomValues, and are converted to a die face with rejection sampling, which discards the small unusable tail of the range rather than taking a remainder. That removes modulo bias exactly rather than approximately. Chi-square testing over sixty thousand rolls through the actual code path shows uniform face frequencies.
It switches from unpredictable randomness to a deterministic sequence. With a seed set, the link carries it, so anyone who opens that link and rolls gets exactly the same results in the same order. That makes a roll verifiable by other people, which matters for play-by-post games and any roll someone might dispute.
So you can see the tool did what you asked. The discarded die is struck through rather than hidden, which makes the arithmetic checkable and matches what you would see rolling physical dice.
It makes dice explode: when a die shows its highest face, it is rolled again and added, repeatedly. An exploding d6 averages 4.2 rather than 3.5. The chain of values is kept, so hovering a die shows every roll it passed through.
The game runs entirely in your browser. No account is needed and no gameplay data is collected.