How Shuffled works

One verified shuffle per day. Here is exactly what happens when you press the button - and why every claim on this page is literally true.

1. The randomness

The deck is shuffled on the server, never in your browser - nothing about the order is decided on your device, so nothing can be tampered with there. Every random draw comes from Node's crypto module, the cryptographically secure random number generator that underpins TLS encryption and key generation. There is no seed, no timestamp trick, no pattern to predict.

2. The shuffle

We use the Fisher-Yates shuffle, walking the deck from the last card to the first. At each step, the current card swaps places with one chosen by crypto.randomInt, which uses rejection sampling so every position is chosen without bias. Fisher-Yates is mathematically proven to make all 52! possible arrangements precisely equal in probability - the gold standard for unbiased shuffling. A machine shuffle like this is actually more random than a human one: studies show a deck needs seven or more riffle shuffles before it approaches true randomness.

3. The fingerprint

The moment your deck exists, the server computes a SHA-256 hash of the exact card order - a digital fingerprint. Change even one card and the hash changes completely. Because the recipe is stated here, you can recompute it yourself and confirm the cards you see are the cards the server dealt. The shuffle and hashing code is public, verbatim, at github.com/Gradmania/shuffled-verification - run it yourself with Node.

4. The comparison

Your deck is then laid alongside every shuffle ever recorded here, one by one, position by position. A match means the same card in the same slot: if your 14th card is the Queen of Hearts and so is a stranger's, that is one match. Your result is the closest relative your shuffle has in the whole archive - and your shuffle stays in the archive, so a future stranger may match against it days or years from now.

The odds of two decks sharing N positions

Between any two random decks, shared positions follow the fixed-point distribution: the expected number is exactly 1, and the chance of at least one is 63.2%. The per-pair odds for each exact count:

One honest caveat, because the archive is bigger than any single pairing: your result is measured against every recorded shuffle, and pairs multiply fast - a thousand shuffles make nearly half a million possible pairs; a million make five hundred billion. Just as fifty people in a room are almost certain to include two who share a birthday, a large enough archive is almost certain to hide close matches somewhere inside it. The odds above are about two shuffles meeting - not about how common such a result becomes across everyone.