Kelly Mears

Seeded Randomness

Pseudorandom generation from an explicit seed, so any output is reproducible.

Graphics & Rendering2 min read209 words10 out · 22 in
also calledSeedDaily puzzle

Seeded randomness means using a pseudorandom generator initialized from an explicit value, so the same seed always produces the same sequence. Everything described as "random" in software is this; the only question is whether the seed is recorded.

Recording it is what makes randomness compatible with engineering. A generated world, a shuffled deck, or a procedural layout becomes reproducible, so a bug in it can be reported, tested, and fixed. Without a seed, a defect in generated content is a story rather than a case. See Determinism and Reproducible Case.

The pattern also has a direct product use: deriving the seed from the date gives everyone the same puzzle on the same day, which is the mechanic behind the daily-word-game genre. It creates a shared experience with no server, no accounts, and no coordination — the date is the coordination. Offering several difficulties per day is the same seed varied by a difficulty label.

Two cautions. A generator's quality matters for simulation and not much for games, but its distribution does — a naive modulo of a random integer into a range is subtly biased. And seeded does not mean secret: if the seed is derivable, so is the outcome, which matters the moment anything competitive rests on it.

See also8

Related2

Nearby in the graph rather than deliberately chosen. Looser, sometimes surprising.

Linked from22