How randomness works here.

The short version

1. Your browser asks the operating system for cryptographically secure random bytes via the crypto.getRandomValues API. We do not use Math.random.
2. Those bytes are turned into a uniform integer using rejection sampling, so every name has exactly the same chance of winning. No off-by-one or modulo bias.
3. We store the exact bytes used (the "seed") alongside the winner. Anyone with the seed can replay the draw.
4. When you visit a result page, your browser independently recomputes the result from the seed and confirms it matches.

Why rejection sampling

The naive way to pick a number 1..N from a 32-bit random word is word % N. That works perfectly when N divides 2³², and is slightly biased otherwise — early numbers in the range get picked a tiny bit more often. The bias is tiny but it is real.

We avoid it by computing a threshold equal to the largest multiple of N that fits in 32 bits, and rejecting any drawn word at or above the threshold. The accepted word is then taken modulo N. The probability of needing to reject anything is microscopic for any reasonable list size, but the result is provably uniform.

The algorithm, in code

function pickIndex(n) {
  const max = Math.floor(0xFFFFFFFF / n) * n;
  const buf = new Uint32Array(1);
  let x;
  do {
    crypto.getRandomValues(buf);
    x = buf[0];
  } while (x >= max);
  return x % n;
}

We record every 32-bit word drawn (in big-endian order) so the loop can be replayed deterministically by anyone holding the seed.

What you can verify on a result page

• The seed bytes used (shown as a hex string)
• The full list of entries that were in the draw
• The algorithm name (so any future change is explicit)
• A "Recompute" button that runs the algorithm in your browser and confirms the recorded winner