How Random Winner Selection Works

The goal: equal chances

A fair random draw has one defining property: before the draw, every eligible entry has exactly the same chance of winning. With 200 entries and one prize, each entry has a 1 in 200 chance. With 200 entries and three prizes, each entry has a 3 in 200 chance of being among the winners, because the winners are drawn without putting names back.

Everything in a good selection method serves that goal. The method must not favour the top of the list, the first commenter, the longest comment or anyone the host likes. It should produce a result that no one, including the host, can predict in advance.

Turning a list into a draw

A picker begins with your list. Each line becomes one entry, and blank lines are dropped. Optionally, repeated entries are removed. From there there are two common ways to choose winners: pick random positions one at a time, or shuffle the entire list and read winners from the top. Shuffling is neat because the same step works for one winner or ten, and the order of the shuffled list even gives you natural reserves.

The shuffle used by the tool described here is the Fisher-Yates algorithm. It walks through the list and swaps each position with a randomly chosen position among those not yet fixed. Done properly, it makes every possible ordering of the list equally likely, which means every entry is equally likely to land first.

Where the random numbers come from

The shuffle needs a supply of random numbers. In a web browser, the standard source for strong randomness is crypto.getRandomValues(). It fills an array with cryptographically strong random values, and it is designed for security work such as key generation, so it is more than adequate for a prize draw.

Using it also means the sequence cannot realistically be guessed from earlier outputs. That matters when a prize is at stake, because a draw that could be predicted could in theory be gamed. The longer question of how this source differs from physical randomness is covered in a separate article on pseudo-random and true random numbers.

The key point for hosts is simple: the source is built into the browser, it is free to use, and the draw does not need a server to decide the result.

Avoiding hidden bias

Randomness can be spoiled by small implementation mistakes. A well-known one is modulo bias: if a program takes a random number from a large range and reduces it with the remainder operator to fit a smaller range, some results come up slightly more often than others, unless the range divides evenly. The usual fix is to discard values that fall outside the even part of the range and draw again.

Another mistake is a naive shuffle that swaps every element with a randomly chosen element from the whole list. It looks reasonable, but it produces some orderings more often than others. This is the reason the Fisher-Yates method is the standard choice: it avoids that skew by design.

How duplicates change the odds

The draw treats every line as a separate entry. If a person appears three times in your list, they hold three tickets and have three times the chance of a person who appears once. That is neither right nor wrong in itself; it is a rule, and you must choose it knowingly.

When the promise is “one chance per person”, remove duplicates before the draw. When the promise is “more comments, more chances”, keep them and say so publicly. The problem arises only when the rule you announced differs from the list you used.

What randomness does not promise

Random does not mean evenly spread, and it does not mean “fair outcomes” in the sense people sometimes expect. A random process can pick someone who won last month, pick two friends in a row, or choose an entry near the top of the list. None of that indicates a fault, because each draw is independent of the earlier ones.

It also does not check eligibility. The tool draws from whatever list you give it, including entries that should have been excluded. Randomness guarantees fairness of the selection step, not of the whole giveaway. For that you need clear rules and a clean list.

If you want to see this for yourself, run a draw on a short list of ten names many times. You will notice streaks and repeats that look odd to the human eye, but over many runs each name appears about equally often.

Making the result believable

Most people cannot inspect an algorithm, so trust usually comes from visible process. A few habits help:

• Publish the entry rules and the closing time before the giveaway starts.
• State how many valid entries were in the final list.
• Run the draw once and record the screen while you do it.
• Explain in one sentence how the winner was chosen.

Because the tool runs in your browser and does not upload the list, you stay in control of the data throughout. If anyone asks questions, you can describe exactly what was drawn from what, and that is the most convincing answer a host can give.