Settings: 1 to 10, integer, count 1.
Interpretation: each integer has a 1 in 10 chance.
Settings: 1 to 6, integer, count 1.
Interpretation: the output is one fair virtual die face.
Settings: 1000 to 9999, integer, count 1.
Interpretation: every result contains exactly four digits.
Settings: 1 to 30, count 30, unique.
Interpretation: every integer appears once in random order.
Settings: 0 to 1, 3 decimal places, count 5.
Interpretation: values come from 0.000 through 1.000.
Settings: 1 to 49, 6 unique values, 3 sets, reuse allowed.
Interpretation: each set has six different numbers; later sets may repeat earlier numbers.
For an inclusive integer range, the rule is X = min + floor(U × N), where N = max − min + 1 and U is a uniform value in [0, 1). Therefore every allowed value has probability 1/N. This implementation uses equivalent exact BigInt rejection sampling, which avoids modulo bias even when the domain is larger than 232.
Decimal precision converts values to a finite integer grid: at two decimal places, 1.25 becomes 125 and the default step is 1 on that scaled grid (0.01 when displayed). A custom step selects a wider evenly spaced sequence. Endpoint options then remove the minimum, maximum, or both when those values lie on the sequence. Exclusions remove more eligible points.
Repeatable sampling uses replacement, so each draw keeps the same probabilities. Unique sampling uses no replacement: after one of N values is selected, each remaining value has probability 1/(N − 1) on the next draw. Weighted modes deliberately change those probabilities and are not fair uniform draws.
BigInt values, rather than unsafe binary floating-point integers.crypto.getRandomValues(). It does not fall back to Math.random(); if Web Crypto is unavailable, generation stops with an accurate error.Maintainer: Starlight Robotics editorial team
Technical review: Starlight Robotics engineering team
Last tested: 17 July 2026
Processing: Client-side in this browser tab
Cost: Free to use
“Client-side” means the inputs, draws, and temporary history are processed in this page and are not uploaded by the generator. Web Crypto supplies cryptographically strong pseudorandom bytes, but this general-purpose interface is not an audited drawing system. Do not use it for regulated raffles, regulated gambling, or security-key, password, token, or seed generation.
Technical references: W3C Web Cryptography API: getRandomValues and MDN Web Docs: Crypto.getRandomValues().
RNG means random number generator: a system that selects values from a defined set using a source of randomness.
They are cryptographically strong pseudorandom results supplied by your browser’s Web Crypto API and ultimately seeded by the operating system. They are not measurements of a physical true-random process.
Yes by default. Advanced endpoint settings can exclude either or both endpoints. A maximum is included only when it lies on the selected stepped sequence.
Yes. Minimum and maximum can be negative, and Decimal mode supports 1 to 12 decimal places with an optional step size aligned to that precision.
Yes. Enter comma-separated values or ranges such as 3, 8-12, or -5--2. Every exclusion must be inside the selected stepped domain.
Unique results sample without replacement, so a value cannot repeat within a set. For multiple sets, you can allow reuse in the next set or remove selected values across every set.
In uniform mode, each of N eligible values has probability 1/N on the first draw. With unique results, the next draw has probability 1/(N − 1) for each remaining value. Weighted modes are intentionally biased.
Yes. Enter those bounds or use the 1–10 and 1–100 preset buttons to load the settings instantly.
No. Generation and temporary draw history stay in this page’s browser memory and are not uploaded. Reloading or closing the page clears the history.
It is suitable for informal games, classroom draws, tests, and non-regulated raffles. Do not use it for regulated drawings or gambling, and use a dedicated audited generator for passwords, keys, tokens, or other security material.