Random Number Generator

Repeats, and why the default is off

By default every number you draw is different. That is what you want for a raffle, a lottery line, or picking five seats out of thirty - the same ticket cannot win twice.

Turn repeats on when each draw should be independent, like simulating dice or generating a PIN. The distinction catches people out: drawing six numbers from 1 to 49 with repeats allowed is not a lottery line, it is six separate draws that happen to have been made together.

Where the randomness comes from

Numbers are drawn using your browser's cryptographic random number generator, not Math.random(). The difference matters more than it sounds. Math.random() is a pseudorandom generator seeded from a small amount of state, fine for shuffling a playlist and not fine for anything where someone has an incentive to predict the next value.

The result is then reduced into your range using rejection sampling rather than a modulo. Taking a 32-bit random number modulo 100 makes the numbers 1 to 96 very slightly more likely than 97 to 100, because 2^32 does not divide evenly by 100. The bias is tiny and it is real, and it compounds across a large draw. Rejection sampling discards the values that would skew it and draws again.

Proving a draw after the fact

Every generation produces a receipt containing the range, the seed and the results, so anyone can reproduce it. For a prize draw that is the difference between announcing a winner and being able to show your working. How the receipts work.

Picking from a list instead

If the numbers are standing in for something - names, entries, options - it is usually better to put the actual list into the name picker or the wheel. The receipt then records what was actually being chosen between, rather than an index somebody has to map back by hand.