About
About DiceCurv
A dice roller built around one idea: the number you rolled is less interesting than how likely that number was.
Why this exists
There is no shortage of dice rollers on the web. Almost all of them give you a total and stop there, which is fine for replacing a die you have lost down the back of the sofa and useless for the question people actually have at the table, which is some version of was that bad, or was that normal?
DiceCurv answers that. Every page carries the full probability distribution for its roll: the chance of each total, the odds against it, and the cumulative chance of hitting that total or better. So when a player groans at rolling 6 on 3d6 you can tell them it happens about 5% of the time, which is unlucky but not remarkable, and when someone rolls 17 you can tell them that is a one in seventy two event.
How the numbers are worked out
Every probability on this site is exact. Nothing here is simulated or sampled. The generator enumerates the complete outcome space for each notation and computes the distribution using exact rational arithmetic, so the figure you see is the true probability rounded for display rather than an estimate from a few million trial rolls.
That matters more than it sounds. A simulation of a million rolls of 8d6 will give you the average to two decimal places quite happily, and will tell you almost nothing useful about the chance of rolling 48, which is one in 1,679,616. Enumeration handles both cases identically.
When you change the roll in your browser, the distribution is recomputed there using the same method, so the chart always describes what you are actually rolling rather than what the page was built for. The browser code and the build code round identically, which is why the tooltip on a bar and the table underneath it never disagree.
How the dice are rolled
Rolls use crypto.getRandomValues with rejection sampling. The
naive approach, taking a random number modulo the number of faces, is biased
whenever the range does not divide evenly into the generator's output, and it
quietly favours the low faces. Rejection sampling discards the values that
would cause that skew and rolls again, so the faces really do come up flat.
Everything happens in your browser. No roll is sent to a server, nothing is logged, and the site works with the network disconnected once the page has loaded, which is worth knowing if your usual venue has poor signal.
Who runs it
DiceCurv is an independent project, not a company, and it is not affiliated with any game publisher. Where the site refers to specific game rules it is describing openly published mechanics in order to explain what a roll is used for.
Corrections are genuinely welcome, particularly on the maths. If a figure here disagrees with one you have worked out yourself, please get in touch and say which page and which number.