Random Number Generator

Uncheck "Allow duplicates" for lottery numbers

The first thing to know, because the default will surprise you and the mistake is easy to miss.

In the Multiple Numbers tab, "Allow duplicates" is ticked by default. Ask for six numbers between 1 and 49 and you can get 17 twice — which is correct behaviour for six independent draws, and wrong for a lottery line, where each number is drawn from a shrinking pool.

So: untick it for anything where the same result cannot occur twice — lottery numbers, raffle positions, a set of distinct picks, assigning people to slots. Leave it ticked for genuinely independent events, such as simulating dice rolls or generating test data.

Confusingly but correctly, the Random from List tab has the opposite default: duplicates are off, because drawing the same person twice from a hat is almost never what anyone wants. Each tab's default suits its usual job, which means the one thing to do is glance at the checkbox on whichever tab you are using.

When duplicates are off, the numbers are drawn by properly shuffling the whole range and taking from the top — the same method as dealing from a shuffled deck, which is the correct way to do it and not what every generator uses.

One limit worth knowing: if you ask for more unique values than the range contains — ten distinct numbers between 1 and 5 — you get as many as exist rather than an error. Check the count you got back.

This uses real cryptographic randomness

A technical point with practical consequences, and it is checkable.

Most online random number generators call the standard JavaScript random function. It is fast, it is fine for a game, and it is not designed to be unpredictable — it produces a sequence from an internal state, and given enough outputs that state can be worked out and the rest of the sequence predicted.

This tool uses the browser's cryptographic random source instead, the same one used to generate encryption keys. It draws from genuine system entropy and its output cannot be predicted from previous values.

For everyday use neither matters at all. Picking a restaurant, choosing a number between 1 and 100, rolling for a board game — any generator will do. It starts to matter when someone might have a reason to game the result: a giveaway with a prize worth having, a draw where a participant could object, anything where "how do I know it was fair" is a question that might actually be asked.

One caveat, stated precisely because the point of saying any of this is accuracy. The random value is reduced into your chosen range with a remainder operation, which makes the lowest values very slightly more likely than the highest. At ordinary ranges the effect is around one part in a hundred million — utterly undetectable, and irrelevant for any practical purpose. It only becomes measurable if you ask for a range spanning billions. For a giveaway, a lottery line or a dice roll, it makes no difference whatsoever.

When to use RANDOM.ORG instead

Worth naming, because for one job it is genuinely the better tool and you should know that before a draw rather than after.

RANDOM.ORG generates its numbers from atmospheric noise rather than from a computer, and — the part that matters — it can issue a signed, timestamped record of a draw that anyone can verify afterwards.

That is the difference between a result being fair and a result being provably fair, and they are not the same thing. If you are running a public prize draw, a competition with real value at stake, or anything where a losing participant might reasonably ask for evidence, a result that cannot be independently checked is a weak position no matter how good the randomness behind it was.

This tool has no audit trail. There is no seed, no record and no certificate — the result appears on your screen and that is all. For a small giveaway among people who trust you, that is entirely fine, and recording your screen while you draw is a reasonable informal safeguard. For anything contested, use a service built to be audited.

What this does better is everything else: it is instant, it needs no account, it runs offline once loaded, it puts five modes in one place, and nothing you type — including a list of names and email addresses — leaves your device.

The five modes

Single Number takes a minimum and maximum, defaults to 1 to 100, and keeps a history of what it has produced so you can see the previous results. Tick "Include decimals" for a fractional value rather than a whole number.

Multiple Numbers generates up to a thousand at once, with the duplicates setting above and an optional sort. Leave the sort off while drawing and apply it afterwards if you want a tidy list — sorting first makes it harder to see the order things were actually drawn in, which is the part that matters for a raffle.

Random from List is the giveaway tool. Paste names, one per line, set how many winners you want, and draw. This is the mode most people arrive looking for, and it handles the whole job — paste a column straight out of a spreadsheet and it works. If your list might have the same person entered twice, Duplicate Line Remover will clean it first, which matters more than it sounds: duplicate entries are extra chances to win.

Dice Roller covers d4, d6, d8, d10, d12, d20 and d100, up to ten at a time, showing each die and the total. That is the full tabletop set, so a lost d20 mid-campaign is not a problem.

Coin Flip shows the result and keeps a running history of the last twenty with the tally. The history is the interesting part — watch it long enough and you will see runs of five or six heads, which is exactly what real randomness looks like and exactly what people wrongly take as evidence something is broken.

A note on what random looks like

Since this tool makes it easy to see: genuinely random sequences contain clusters, streaks and repeats. Six heads in a row will happen roughly once in every sixty-four attempts. Two identical numbers in ten draws from 1 to 100 is unremarkable.

Sequences that look evenly spread, with no repeats and no runs, are the ones that have usually been tampered with — evenness is what people produce when asked to fake randomness. So if a draw looks suspiciously clumpy, that is generally a point in its favour.

Everything runs in your browser. Nothing is transmitted, which is the reason to use it for a draw involving real people's names: a list of entrants is personal data, and it has no reason to be on someone else's server.

Related tools

Password Generator uses the same kind of randomness for passwords and passphrases, where unpredictability is the entire point. Duplicate Line Remover cleans a list of entrants before you draw from it. Text Case Converter tidies a pasted list that has come out in inconsistent capitalisation.