Why “random” is hard for a computer
A computer follows instructions exactly, so by itself it cannot produce anything unpredictable. What programs usually call random is pseudo-random: a formula takes a starting number, the seed, and stretches it into a long sequence that only looks chaotic.
- Ordinary generators, such as JavaScript's
Math.random(), are fast and fine for animations. But after seeing a few results, a program can work out the internal state and predict the rest. - Cryptographically secure generators are built so that no amount of known output helps to guess the next value.
- Physical sources — electronic noise, radioactive decay — are unpredictable by nature but slow, so in practice they seed generators of the second kind.
Where our random numbers come from
In your browser, every result is produced with crypto.getRandomValues() — a cryptographically secure generator built into all modern browsers. It is seeded by the operating system, which keeps collecting entropy from events nobody can predict: the precise timing of key presses, disk and network activity, the hardware noise generator inside the processor. The same source is used to create encryption keys.
From random bytes to a fair die: the modulo bias
A random byte is a number from 0 to 255 — 256 values — and a die has six faces. The obvious shortcut is to divide by 6 and take the remainder. But 256 = 42 × 6 + 4: every face gets 42 values, and the four left over go to faces 1 to 4. Those faces would come up 43 times out of 256, the other two only 42. The tilt is small, but it is a tilt — the modulo bias.
The cure is rejection sampling: values that do not divide evenly — here 252 to 255 — are thrown away and a new byte is drawn. The remaining 252 values split exactly, 42 per face, so every face has precisely one chance in six. All generators on the site turn random bits into numbers this way.
Shuffled lists and the wheel
Lists are shuffled with the Fisher–Yates algorithm: it goes through the list once and swaps each item with a randomly chosen one among those not yet placed. With unbiased numbers, every possible order is equally likely; the random picker works this way.
The wheel decides nothing while it spins. The winner is chosen first, with a chance proportional to the size of its sector, and the animation only turns the wheel until the pointer rests on that sector.
Verified giveaways
In a verified draw the winning positions are chosen on our server by a cryptographic generator, and the result is saved as a protocol with its own link. The full list of participants never leaves your browser: the server receives its SHA-256 fingerprint instead — a short code that changes completely if a single character of the list changes.
The protocol also shows how many times that same list has been drawn, so re-running the giveaway picker until the “right” person wins does not go unnoticed.
What we never do — and how it is tested
- No seeds that can be replayed. A draw cannot be started from a chosen number to repeat a result or prepare it in advance.
- No paid or hidden influence. Nothing lets anyone buy better odds or nudge a result.
Fairness is checked automatically. One statistical test makes a million draws and uses the chi-square test to confirm that all outcomes come up as often as they should. Another checks that the wheel's pointer always stops on the chosen sector.
Honest limits
Our generators are cryptographically secure pseudo-random generators, not quantum devices. Their output is computed, yet no known method can predict it. It is the same grade of randomness that protects banking connections — more than enough for a giveaway, a board game or a password.
Frequently asked questions
Can anyone predict or rig the result?
No. Each result is produced by a cryptographic generator at the moment you press the button and does not depend on the previous ones. There is no seed to set and no switch that changes the odds — for visitors or for us.
Why did the same number come up twice in a row?
Because that is what randomness looks like. Draws are independent, so a repeat is as likely as any other outcome: a die shows the same face twice running about once in six rolls. If you need different numbers, turn on No repeats in the randomizer.
Are the odds the same as with real dice or a coin?
Yes — and a little more exact: a real die or coin is never perfectly balanced, while the online dice roller gives every face exactly one chance in six and the coin flip exactly one in two.
How can participants check a giveaway?
Send them the link to the protocol. Anyone who has it sees the date and time of the draw, the number of participants, the fingerprint of the list, the winners and how many times this list has been drawn.