D&D Dice Roller

It is 7:58 p.m. The session starts at 8:00. Your dice bag is in the jacket you wore Tuesday, the jacket is at the office, and the office is closed. Someone mumbles about a dice-roller app with three ads and a subscription screen, someone else starts shaking their phone like a maraca, and the barbarian is already rolling damage on a random number generator that only does 1 to 100 with no way to add the +4.

Tabletop has a dice problem and it is almost never the dice — it is the situation. The remote game where physical dice are unverifiable. The one-shot at the shop where you forgot everything. The late-night solo journaling session where you want a d12 answer without building a dice ritual out of coins. What you need in all of these is a roller that speaks your game's language, shows its work, and never asks for an account.

That is this D&D dice roller. Full RPG set from d4 to d100, real dice notation with advantage, disadvantage, keep/drop, and modifiers, a visible breakdown of every individual die, crit highlighting, and a session history you can scroll when someone accuses the universe of conspiracy. It runs in the browser, free, with the fairness story told openly — which matters, because a dice roller is a trust tool before it is a math tool.

Serious rollers read dice expressions rather than buttons, because notation scales to everything the game asks for. The grammar is compact: N dice of size X, plus or minus a modifier, plus optional keep/drop instructions. Once you can read it, you can type any roll the rules can describe — including the gnarly ones that make button-based rollers collapse, like rolling stats and applying a modifier in one expression.

1d20+5        standard attack roll with proficiency and Strength
2d20kh1       advantage (roll two, keep highest)
2d20kl1       disadvantage (roll two, keep lowest)
4d6kh3        stat generation: four dice, keep the best three
4d6+2dl1      four dice, +2, drop the lowest single die
8d6           fireball, cast at third level, no survivors
1d8+2d6+5     longsword plus sneak attack plus Dexterity
2d10%         percentile pair: tens die and units die, 00+0 = 100
1d100         single true hundred-sider for the daring

The two stat-generation spellings — kh3 (keep highest three) and dl1 (drop lowest one) — are synonyms, and the tool treats them that way. The breakdown view shows every die with kept dice bolded and dropped or discarded dice struck through, so when you roll 4d6kh3 and see ~~3~~ 6 4 6 = 16, the arithmetic is right there on the screen. No trust required; it is arithmetic in public.

1. 1

   Type or tap your expression

   The quick panel covers the common rolls — d20, advantage, disadvantage, damage by die size. The expression field covers everything else, including the ones nobody has a button for.

2. 2

   Add modifiers in the expression

   Type them as part of the roll (1d20+7) rather than doing the addition in your head mid-combat. The history records the total you actually used, which future-you will appreciate during rules arguments.

3. 3

   Read the breakdown

   Every die, every keep and drop, every modifier, and the total. On a d20 attack roll, a natural 20 lights up and a natural 1 gets its own funereal shade, because celebrating crits is half of what dice are for.

4. 4

   Chain rolls when the turn needs them

   Separate expressions with commas to roll attack and damage together — the order of operations the turn actually has, in one click.

5. 5

   Scroll the history when someone doubts

   Every roll of the session is recorded with its full breakdown. Screenshot the streak of 18s, or confirm the streak never happened. Both outcomes are popular.

Advantage — roll two d20s, keep the higher — is 5e's signature mechanic, and its value is delightfully non-uniform. On raw average, advantage moves you from 10.5 to about 13.82, a gain of roughly 3.3 points that players often compare to a +3 or +4 bonus. But the average hides the interesting part: the bonus is worth the most exactly where most checks actually live.

Consider a check you succeed on an 11 or better — the dead center of the d20, and the most common situation at a real table. Flat, that is a 50 percent chance. With advantage, you succeed unless both dice are 10 or lower: 1 minus 0.5 squared, a clean 75 percent. To match that with a flat bonus you would need +5, which is why optimizers rate advantage as better than most items that grant +1 swords. With disadvantage, the same check collapses to 25 percent, the mirror image — you fail unless both dice come up 11 or higher.

The natural-20 row is the one players quote at each other: advantage nearly doubles your crit chance from 5 to 9.75 percent, while disadvantage drops it to a quarter of a percent — the reason a frightened, prone rogue is a sad spectacle. And the natural-1 row explains why advantage is, statistically, a small insurance policy: you can still roll a 1 with advantage, but both dice have to agree to betray you, and they almost never do.

The greataxe (1d12) and the greatsword (2d6) sit half a point apart on average damage: 6.5 for the axe, 7 for the sword. That sliver is close enough to ignore on any single swing, but the shape around that average is what decides how the weapon feels across a campaign. Two dice summed produce a bell curve — middling totals dominate, extremes are rare — while a single big die is flat: every face from 1 to 12 lands with the same 8.33 percent.

Consistency favors the greatsword: 58.3 percent of its rolls clear 7 or better versus 50 percent for the axe, which is why grinders and defenders like it. Spike damage favors the greataxe, and the gap widens on the one turn everyone remembers — the crit. A critical hit rolls the weapon's dice twice, so the axe's crit is 2d12 with its full spread, while the sword's crit is 4d6, which concentrates near 14. Crit-fishing builds — half-orcs with their extra damage die, barbarians who reckless-attack for permanent advantage — get more from each doubled d12 than from two doubled d6s.

The same curve logic explains why ability scores roll 4d6kh3 instead of a flat draw, why fireball's 8d6 lands between 24 and 32 more often than players expect, and why a single d20's chaos makes sessions swingy by design. Whenever the rules want drama, they use one big die; whenever they want reliability, they stack several small ones. You can watch the difference live here — roll 2d6, 1d12, and 8d6 a dozen times each and read the histories side by side.

Spread has a number, and it is the d20's whole personality. A uniform d20 runs a standard deviation of 5.77 — computed as the square root of (n² − 1) ÷ 12 — while 2d6 sits at 2.42 and 3d6 at 2.96. The single big die is more than twice as volatile as the pair summing to the same neighborhood, which is why a d20 check can feel like chaos while 2d6 damage feels like a metronome. Game designers exploit this deliberately: drama rolls on one die, economy rolls on several.

Character generation is the most-rolled expression in the game, and its distribution is stranger than intuition expects. Roll 4d6, keep the highest three, and enumerate all 1,296 ordered outcomes: the average lands at 12.24 — well above the 10.5 midpoint — and the reason is a built-in asymmetry. Scoring an 18 needs only three sixes (the fourth die can be anything, since a 6 is always kept): 1.62 percent. Scoring a 3 requires all four dice to be 1s, because any higher die would displace a kept one: 0.077 percent, one stat in roughly 1,300. The mechanic cuts the bad tail far harder than the good one.

The array-level view completes it: six stats average 73.47 total points against the standard array's 72, so the two systems are nearly identical in expectation — the difference is that rolled arrays carry variance, and 9.34 percent of them contain at least one 18. That is the honest trade the table makes when it votes down point buy: same mean, fatter tails, and stories about the time someone rolled three 18s and a 6.

Let us follow one combat turn through the roller, the way a real turn actually flows. Our fighter, +7 to hit with her longsword and 16 Strength, swings at an armored knight while her ally has knocked him prone — melee attacks against a prone target have advantage, so the attack expression is 2d20kh1+7.

The first die was a 9, which would have missed (9 + 7 = 16 against AC 18), but the advantage die rescued the turn — exactly the mechanic doing its job at the middle of the range where it is strongest. Now damage: the longsword is 1d8+3, and the fighter's subclass lets her add a d4 on a hit, so the damage expression is 1d8+1d4+3.

Two expressions, two clicks, every intermediate number preserved in the history for the rules-lawyer in all of us. Contrast the stat-generation use: clicking 4d6kh3 six times during character creation gives six ability scores with each die shown — and if you are building the character rather than just rolling it, our D&D Character Generator automates the whole sheet, not just the dice.

A natural 20 on an attack triggers the game's most commonly miscomputed line of math: critical hits roll the weapon's damage dice twice and add every modifier once. Dice double; flat numbers do not. The errors this produces are consistent enough to catalog.

• Doubling the modifier. A longsword wielder with +5 Strength rolls 1d8+5, so the crit is 2d8+5 — average 14 — not 2d8+10. The Strength, the proficiency-driven attack bonus, and flat damage like Great Weapon Master's +10 are all added once.
• Declaring maximum damage. The crit does not award the weapon's top total; it awards another roll of the dice. A 1d8+5 crit can still total 7 on two sad eights-rolled-low, which is statistically honest and emotionally unfair.
• Forgetting which extra dice double. Sneak attack, Hunter's Mark, Divine Smite, and every other rider expressed in dice does double on a crit — the rogue's 1d8+3d6+4 hit averages 19 normally and 34 as 2d8+6d6+4. Spell saves never crit: the fireball that rolls 8d6 does not double for anything, no matter how cinematic the moment.
• Treating 2d6 and 1d12 as interchangeable on crits. They sit within half a point on normal hits — 6.5 versus 7 — but doubled they diverge hard: 4d6 concentrates near 14 while 2d12 spreads from 2 to 24. The crit-fishing builds go big-die for exactly this reason.

The related decision math optimizers argue about is the Great Weapon Master trade: take a −5 penalty on the attack for +10 damage. The break-even hit chance is a clean formula — the power attack wins when your normal hit chance exceeds 25% plus 2.5% per point of average damage. At 15 average damage that is a 62.5% hit chance; at 20 damage, 75%. Advantage, which lifts the mid-range hit chance by 25 points, flips the trade from marginal to mandatory, which is why the barbarian's reckless attack and the feat travel together.

The d100 arrives in two dialects. The single true hundred-sided die — a golf ball of a thing that rolls for a suspiciously long time — is one click here. The traditional pair is two ten-sided dice: one marked 00 to 90 as the tens die, one marked 0 to 9 as the units die, read together as a number from 1 to 100. The tool supports both, and shows the pair separately in percentile mode so nobody has to squint at anyone else's dice tray.

And now, the argument. When the tens die shows 00 and the units die shows 0, what did you roll? The 5e rules settle it: 100. The reasoning is that a result of zero is impossible — the range is 1 to 100 — so the double-zero must mean the top of the scale. Half the tables you will ever sit at have been playing it as 10 since 1989 and will look at you like you have grown a second head. You can now point at this paragraph. We accept your thanks in advance.

Every dice roller is a trust tool before it is a math tool, so here is the whole fairness story. Each die is generated independently from the browser's cryptographic-quality randomness source where available — the same underlying mechanism that secures your connections — rather than a hand-rolled PRNG with discoverable patterns. Faces are equiprobable by construction. Rolls have no memory: the roller does not track what you 'owe' and does not smooth streaks, because streaks are real randomness doing real random things.

The visible-breakdown design is the second half of the trust story. Every kept die, dropped die, and modifier is displayed and recorded, so the tool's arithmetic is auditable by anyone at the table in real time. When the DM rolls three natural 20s in a boss fight, the history will show all three — and the table will groan in public, which is the correct and traditional venue for groaning.

Because the roller highlights which mode you used, it doubles as a teaching tool for the mechanic's edge cases. The condensed decision guide the table arguments keep re-deriving:

That third row is the one worth tattooing somewhere: multiple sources of advantage do not stack, multiple disadvantages do not stack, and one of each cancels to a flat roll. The number of tables that have argued this mid-boss-fight is a number we politely decline to estimate.

Honest limits, because trust is the product. A digital roller cannot be inspected the way physical dice can — you cannot float-test a browser — so the fairness claim rests on architecture and open arithmetic rather than on physical properties. It cannot reproduce the specific pleasure of a dice tower, the sound a full fist of d6s makes in a metal cup, or the ritual weight of declaring a roll before you make it. Some tables ban digital rolling at physical sessions for exactly these reasons, which is a defensible position about ritual, not about math.

The tool also stops at the dice themselves: it does not apply your character sheet's situational bonuses, track spell slots, or remember that the paladin gets to reroll 1s. It rolls dice and shows its work; the game's bookkeeping stays with you, your sheet, or the character generator you built it in. That boundary is deliberate — a roller that starts making rules calls is a roller that starts rules-lawyering, and nobody wants that at 11 p.m.

• Remote tables rolling in shared view, where physical dice are unverifiable by design
• Players between dice-bag ownerships — new players, travelers, the jacket-forgetful
• DMs generating NPCs and random encounters mid-session without a 30-second dice hunt
• Solo journaling and actual-play writers, who roll more dice than any group could feed
• Optimizers studying probability — the history plus the advantage table above is a small statistics course
• Warhammer and d6-pool players, welcome raiders of the notation system

The roller is a browser page: no app install, no permissions, no account, and no roll ever leaves the device. Every generation is local computation, the history lives in the tab's memory and dies with it, and the whole thing works on a train, a plane, or the campground with one bar of signal. The only thing we could do with your dice is wish them well.

The dice roller pairs with the D&D Character Generator for full sheets rather than single stats, the general-purpose Dice Roller for non-RPG rolling, and the Random Number Generator for ranges that are numbers rather than dice. Same fairness architecture across all of them, same everything-in-the-tab philosophy, and the same price the dice have always deserved: nothing.