v2.4.1: Dice questions in the docs

The dice questions release.

Added

• Dice questions: a new page in the Docs lists every common dice question DiceTable answers, such as what 4d6 drop lowest averages or how much advantage is worth, each with its short answer and a link to the page that opens its rolls in a working table. Until now that list only showed on an empty table, so once you had rolls of your own there was no way back to it.
• Each question's own page links back to the list with More dice questions, next to Build it yourself.

Full detail in CHANGELOG.md.

v2.4.0: Modifier steps, Variants and recipe pages

The modifier steps release.

Added

• Modifier steps: a roll's total can now go through up to three steps, in order. Add, subtract, multiply or divide (dividing rounds down). Open a roll and use Modifier steps under the dice to say "double it, then add 3" or "roll it, then halve it", and move the steps earlier or later, since +3 then ×2 is not ×2 then +3. The table reads the steps in the Mod column (+3 ×2) and in the formula, where brackets show the order: (2d6 + 3) × 2. On a roll that counts successes the steps change the number of successes, and adding still means automatic successes. A single +2 still goes straight in the Mod box as before.
• Duplicate a roll, and put rows in order. Open a roll and press the Duplicate icon beside its formula, and a copy lands directly below it, named "Longsword (2)" and already open, ready to change one thing and compare. Move up and Move down sit in the same place. The order is what the chart draws first and which rolls head-to-head shows.
• Variants: make a whole family of a roll at once. Pick what to change (the number of dice, the die, how many to keep, one modifier step, or a pool's success number), pick the range, and up to seven new rolls appear directly below, named for what changed ("Longsword · d10", "2d6 · +3"). Seven, because the chart has eight colours and the roll you started from keeps one. Before you confirm, the panel says plainly what it is leaving out and why, and the message afterwards has an Undo. Checks can't make variants yet; duplicate one and change it by hand.
• Notes on rolls. A short note under the formula, for why a roll exists or what changed from the last version. The table marks a roll that has one, a phone shows it on the card, and notes travel with the table when you share it, export it or open it on another device.
• Recipe pages: ten pages that each answer one common dice question. What 4d6 drop lowest averages, how much advantage is worth, what Great Weapon Fighting adds to a greatsword, and more. Each one is the real table with the rolls already in it, so you can read the chart, change a die and try a target right there, and nothing you do there touches your own table. Open this in my table adds the rolls after the ones you already have. An empty table lists all ten under Common dice questions.
• A new recipe in Find a roll: Roll again on a 6 and add. A plain d6 next to an exploding one, for anyone wondering what exploding dice do to the average.
• The shape of a roll now shows where most of its results land. A faint band of the roll's colour covers the middle of its results, in the small shape beside each roll, in the chart you open from it, and in the comparison chart for the roll you point at. A narrower band is a steadier roll; a flat roll like a single d8 gets no band.

Changed

• Links and files you share now use a newer format. They open in this version of DiceTable or later. If an old copy of the app is still open in another tab, it asks you to reload instead of showing the wrong numbers. Older links and files still open as before.
• A copy or a variant goes directly below the roll it came from. The rolls after it move down to make room, and each takes the colour of its new place; Move up and Move down trade colours the same way.
• The homepage now says what DiceTable is for: building, comparing and balancing dice rolls, for the games you play and the homebrew you write.
• The footer credits Clark Perfecto by name and links to the GitHub profile.

Fixed

• A DiceTable link shared in a chat app or on social media now shows a short description under its title, instead of the title alone.
• Every page now carries a single title once it has loaded, instead of its own title followed by the site's default one.
• The Find a roll recipe that compares 1d20, 3d6 and 2d6 is now called More dice, fewer extremes. Its old name, Same range, different shapes, was wrong: the three rolls cover different ranges.

Full detail in CHANGELOG.md.

v2.3.0: Find a roll and What's new

The Find a roll release.

Added

• Find a roll: the Examples button is now a searchable library of ready-made rolls for the games people actually play. Type a game ("Blades", "Shadowrun", "L5R", "40k"), a rule ("advantage", "explode") or some dice ("3d6"), or narrow it with the chips for adding up, counting successes, best or worst of, head to head and beating a number. Each card says what the roll is also called, why it is built the way it is, and the average and range the row will show once you add it. Some recipes add two or three rows at once, so you can compare advantage with disadvantage, or a normal hit with a critical, side by side.
• Every recipe was checked against the game's own rules before it went in. Where a recipe leaves something out, the card says so: a Blades resistance roll does not count the critical that clears stress, and Daggerheart's total does not count doubles.
• What's new: after an update, the app opens a short list of what changed since your last visit. There is one section per version with the newest on top, and then it stays out of the way until the next one. The version number in the footer opens the same list whenever you like, going all the way back to the first release. A first visit never shows it, and opening a shared link holds it for your next visit, so a link always lands on the rolls it carries.

Changed

• The Examples button is now called Find a roll, and on a phone it shows as its own button. It sits in the same place as on a bigger screen instead of only inside the overflow menu, so you can find it without knowing it is there. An empty table offers it too, so the game recipes are there from your first visit. The dialog's "Load every example" is now "Load the starter set", and still adds the same eight rolls.
• The privacy page now names everything the app keeps in your browser. That includes the version you last opened and your light or dark choice, and it says plainly what the error screen's reset clears.

Fixed

• On a phone, the Target view's Edit buttons no longer run off the right edge of the screen. That covers both the target and the pool target, and the page no longer scrolls sideways there.

Full detail in CHANGELOG.md.

v2.2.0: Roll-and-keep with exploding dice

The roll-and-keep release.

Changed

• Some rolls that keep the best dice across parts now say "(too complex)" where they used to draw. The check that decides this was measuring the old way of working the odds out, and the app no longer works them out that way, so it had to be replaced rather than adjusted. The rolls affected all combine very large dice with keeping across parts, such as three d1000 keeping the best two, or three d20 exploding fifty deep. They drew in about a tenth of a second before, so if you were using one, it is a real loss rather than a slow roll being tidied away.
• A check whose critical rolls more dice than an ordinary hit is now measured on the dice the critical really rolls. That is more honest, and it means a few check rolls with large critical effects are refused where they used to draw, because they were only ever accepted on the size of the smaller roll.

Fixed

• Keeping some of a roll's dice while those dice can explode no longer freezes the page. Six dice keeping five took twenty-five seconds of a locked-up tab, and eight keeping seven would have taken about half an hour. Both are under a hundredth of a second now. Keep and Explode on the same group was reaching the maths through a route that grew with the number of faces a die could show, and an exploding d6 can show fifty-six of them rather than six. The check meant to catch that was reading the die as though it had never exploded; it reads the roll the app is actually going to make now.
• Rolls that kept dice are no longer turned away for work the app does not do. Keeping the best few of twenty d20s, of thirty d6s, or of a hundred d100s all said "(too complex)" and simply answer now, as does roll-and-keep at the size those systems actually use: twelve exploding d10s keeping six is nine milliseconds. A few of the largest take about a second rather than being instant, which is still a roll you can wait for rather than one that locks the tab.
• A roll that explodes on a low face is no longer refused for being wide. Exploding a d100 on a 1 was measured as though every extra die could land on 100, which made a roll reaching 150 look like one reaching 5,100.
• Dark mode no longer flashes light on every page load. The small script that sets the theme before the page paints was written straight into the page, and the site's own security policy does not allow that, so it never ran once in production: every visit painted light and then corrected itself when the app loaded. It is a file now, which the policy does allow.
• The warning about a roll that can never stop exploding now asks the right question. Picking the same face twice, or picking faces the die does not have, was read as exploding on everything and refused. Meanwhile a d6 that rerolls 1 to 3 and explodes on 4 to 6 really can never stop, and that passed, leaving a blank row and no reason for it.

Full detail in CHANGELOG.md.

v2.1.0: UI redesign

The UI redesign release.

Added

• The chart field: enlarging one chart on a wide screen draws every roll at once, however many there are. Past about thirty the curves stop being drawn in their own colour and dash and become a field: all of them at once and faint, with the one you point at lit up on top. Eight colours and eight dashes only ever make eight tellable-apart pens, so past that no cap and no extra canvas helps. Pick a name from the legend to light one, pick it again to put it back. A screen reader hears where the roll ranks instead, since a picture is nothing without sight.
• A way back to a page: the field is the default on that canvas, not the only option. A "20 at a time" button beside the view chips switches the enlarged chart back to twenty rolls in their own colours and dashes, with arrows to page through the rest. The field answers where one roll sits among all of them; the page answers what each roll looks like. The button only appears where the choice would change something.
• Advantage and Disadvantage say which way they move the curve, and each names the figure: on 1d20 that end of the chart goes from 5% to 9.75%.

Changed

• Head-to-head shows twenty-four rolls, up from twelve, and opens in about a third of the time it used to take at that size. The grid was building a separate tooltip per cell, which is row-count squared of them; there is one shared tooltip now, and every cell carries its sentence as its own label so a screen reader reads it whether or not anything is hovered.
• A big table gets a comparison chart again, and you can page through it. Past twenty rolls the chart used to go blank and say it was disabled, which is easy to land on in one paste of a shared link. It draws twenty at a time now and says which twenty ("Showing 21 to 40 of 75 rolls"), with previous and next beside it, so every roll can be reached. Totals and Successes page separately.
• Target hit's Curves can draw twenty rolls at a time. It still draws every roll by default, since reading one roll against all the others is the point of that view, but a hundred lines on one canvas is a wall rather than a comparison. The scale stays on the whole table whichever page you are on.
• Target and Pool target are each a single control, on every screen size. Each group is now one button that reads out what it is set to ("Target ≥ 10 14 18") and opens its editor when you press it: a small panel on a wide screen, a sheet from the bottom on a phone. The bar holds one line whatever you set.
• The eight row colours are one family now. They were a stock set that never got tuned with the rest of the palette. Every row now carries the same colour strength, so no roll looks more important because of the slot it landed in, and neighbouring rows stay separable for red-green colour blindness, which the old set did not manage. There is a set for light mode and a set for dark, because one set cannot clear the contrast floor against the palest light surface and the darkest dark one at once.
• Controls are finger-sized on a phone. Everything you tap is at least 40px there now, and goes back to its compact size once the table takes over on a wider screen.
• Clicking a name in the chart legend singles out that roll and keeps it singled out. Before, only hovering worked, which meant it did nothing on a phone and nothing from the keyboard even though it announced itself as a button.
• Pinning a roll to compare against no longer moves the table, in either direction. It used to widen the layout by about 125px, grow the header by a line, and raise a sideways scrollbar that was not there before.

Fixed

• The docs described a table that had moved on. The Head-to-head glossary entry said a big table draws its first twelve rolls while the view above it printed "Showing the first 24". Quickstart step 2 said to tap any token to edit it inline, which stopped being true when the notation became the way to open the distribution behind a roll. The P entry counted three chart views when there are four. Several more of the same kind, all corrected.
• A share link is about a quarter of the size. Every roll carried a hidden identifier that the recipient regenerated on arrival and never read.
• Opening a link made by a newer version says so and tells you to reload, instead of claiming the link is corrupt.

Full detail in CHANGELOG.md.

v2.0.0: Check rolls and share as image

The check rolls release.

Added

• Check mode: a third setting beside Sum and Pool for rolls that decide whether something happens. A check row rolls against a number you set, then applies an effect scaled by how it went, which covers attack-then-damage, save-for-half, and any if-then mechanic without typing a formula. The expanded row asks for the Check (die, modifier, and the number to clear, with a live success percentage), the Effect (its own dice and modifier), and the Outcomes (all of it, half rounded down, or none) on a success and on a failure. A check of exactly one die can also carry a Critical: chosen faces that always succeed and either double the dice, add one die, or add the highest the dice can show. The row's average counts the misses, so it reads as the result per attempt.
• Check rows are ordinary rows everywhere else: the Table, Target hit, Roll-off, Head-to-head, baseline deltas, sparklines, share links, and import all treat them like any other roll on the totals scale. Rows carry a Check badge, an orange band, and a "succeeds NN%" chip.
• Keep across parts: a rule on the whole roll rather than one part, so the trait-and-wild-die roll (keep the highest single die across a d8 and a d6) is finally expressible. It works across dice of different sizes, and turning it on switches off the per-part Keep chips, which explain why.
• Share as image: an Image button on every view renders a picture of whichever view you are on to a PNG, with an optional title. The picture follows the controls you set, not just the view you picked, and every card states on itself whatever it left out. Copying puts the picture and the share link on the clipboard together, so pasting into a chat gives the image and pasting into a text field gives the link. The picture is drawn from the computed numbers, not screenshotted.
• Docs: glossary entries (Keep across parts, Check, Succeeds when, Effect scale, Critical), two new "The Math" sections, a new Quickstart step 6 introducing Check mode, and an updated public/llms.txt.

Changed

• Target hit and Roll-off remember their own controls: the sub-view, the All / Sum / Pools filter, the grid's column sort, and the roll-off sort order are saved with the table instead of resetting whenever you leave the view, and the share image reads them so the picture matches what you set.
• The PMF chart no longer lets one row's misses flatten every other curve: a check row's "nothing happened" bar is capped and marked with its exact chance, still readable on hover and in the chart's description. Row sparklines do the same at their own scale, and tables with no check rows are unaffected.
• All probabilities stay exact. Checks are a weighted mix of the outcome distributions and keep-across-parts is an exact level walk; neither samples nor approximates.
• The exported image matches the chart it pictures: round axis percentages, one flat tread per result on the PMF view, and the same dash patterns that keep rows apart for color-blind readers. Narrow result ranges now reserve half a tread at each end, so the first and last steps sit inside the axis.
• The effect outcome controls read as one pick-one segmented control instead of three separate rule chips, so orange means "this is a check row" and blue means "this control is on", nowhere both.

Fixed

• The Curves sub-view no longer demands a target it does not use. Curves draw as soon as a summed roll exists, and the Grid / Curves / Bars chips stay on screen with no target set.
• The Roll-off headline no longer names a favourite when no roll can win outright. A table whose rolls always tie now reads "These rolls almost always tie."
• A check row that rolls more than one die no longer reports "(too complex)" when it carries a leftover critical rule.

Full notes in the CHANGELOG.

v1.9.0: First-run examples and clear all

The start-with-an-example release.

Added

• Start with an example: first visits begin with an empty table and a panel of 8 preset rolls, each card showing its real engine-computed average and range. "Use this roll" adds one, "Load every example" adds all 8, and "Start from a blank roll" works like Add roll. Deleting or clearing every roll returns to the panel.
• Examples dialog: once the table has rows, the same preset cards stay reachable from an Examples toolbar button. Adding one appends it to the table, and re-adding an example renames the copy, like "Weapon attack (2)".
• Clear all: a red button next to Add roll that empties the table after a confirmation dialog. Clearing also unpins the baseline. On phones the button reads "Clear".
• Quickstart: step 1 now mentions the example panel and the Examples button.

Changed

• Every Add roll button now disables at the 100-roll cap with an explanatory tooltip, including the one in the toolbar.

Full notes in the CHANGELOG.

v1.8.0: Docs section routes and a new icon

The findable-docs release.

Changed

• Docs sections at their own addresses: /docs/quickstart, /docs/glossary, and /docs/math. The section strip is real navigation (each section can be opened, bookmarked, and shared directly, with its own page title and description), and old /docs and /docs?tab= links redirect to the matching section.
• Machine-readable glossary: every term now carries a schema.org DefinedTermSet definition, and the sitemap and public/llms.txt point at the new section addresses. llms.txt also covers the baseline comparison and share features, and describes dice notation as displayed rather than typed.
• New app icon: a twenty-sided die sitting under a wooden table. The favicon, PWA icons, apple touch icon, and the social share card are all regenerated from it.

Full notes in the CHANGELOG.

v1.7.0: Roll-off and head-to-head views

The "who wins?" release.

Added

• Roll-off view: if every roll rolled once, each roll's exact chance of producing the single highest result — win bars scaled against the favorite, a headline sentence, per-roll tie notes, and a Win chance / Table order sort toggle. Ties are counted separately; nobody wins those outright.
• Head-to-head view: a matrix of one-on-one odds where each cell is how often the row roll strictly beats the column roll, ignoring every other row, with the tie chance on every cell. Scrolls sideways inside its own container on mobile.
• Both views compute exactly from the distributions already on hand (a new pure compare engine module, no simulation) and include Pool rolls on their success-count scale, with a caption explaining why cross-scale match-ups are usually lopsided.
• Comparison docs: glossary entries (Roll-off, Head-to-head, Tie), a "The Math" section on exact win and tie chances, a new Quickstart step introducing the four workshop views, and an updated public/llms.txt.

Full notes in the CHANGELOG.

v1.6.0: Baseline comparisons

The baseline release.

Added

• Baseline comparisons: pin any roll as the baseline with the pin button on its row. The pinned roll keeps its own numbers, wears a Baseline badge, and gets a tinted row with an accent band. Every other roll then shows how it differs instead of its own totals: labeled avg and spread delta lines with small direction bars, signed Hit % differences in percentage points, and a plain-language verdict under the name ("Averages 3.5 higher · swingier · hits 46% more often"). Green means better, red worse, and spread changes stay neutral. Rolls on a different scale (Pool vs Sum) keep their own totals and compare by Hit % only. Pinning another roll moves the baseline, tapping the pin again clears it, and the choice survives reload.
• Community dice vocabulary: the glossary's "Target ruling" entry now maps each ruling to its community name (≥ is "roll over" or "meet or beat" as in D&D and Pathfinder, ≤ is "roll under" as in Call of Cthulhu and GURPS).

Changed

• The ≥ and ≤ ruling tooltips (target chips, Hit % header, target hit view) name the same community terms.

Fixed

• TARGET view bar labels no longer overlap: labels drop a redundant trailing ".0", hide as a group when bars are too narrow to fit them, and the chart claims more width per bar.

Full notes in the CHANGELOG.