The number people are worst at choosing
"Pick a number between 1 and 100" is the oldest guessing game there is, and it survives because people are so predictably bad at it. Asked to produce something random, we avoid the ends, skip the round numbers, dodge anything that reads like a date, and drift toward odd two-digit values in the upper middle. 37 and 73 are the famous ones. What almost nobody says out loud is 1, 50 or 100 — the exact values a real draw returns as often as any other.
That gap is the reason a tool beats a person here. If you are hiding a number for someone else to guess, your own taste narrows the search for them before they start. A generated number does not: it sits anywhere in the hundred with equal weight, round numbers and edges included.
One hundred even slices
A hundred values need seven bits of entropy, and seven bits produce 128 candidates. The 28 that overshoot get discarded and the generator pulls again, so a little over one attempt in five is thrown away before a usable number appears. The lazy alternative is a remainder operation, and at this range it is badly wrong rather than subtly wrong: folding 128 candidates onto 100 slots would give the first 28 numbers close to double the frequency of the other 72. We discard instead — the methodology page covers the details.
The result is a flat 1% per value across the whole hundred. Over a long session the distribution evens out; over ten presses it will look nothing like even, because ten samples spread across a hundred buckets almost never do. Both of those are the same tool behaving correctly.
What a 1 to 100 draw gets used for
The guessing game itself. Generate, keep the screen to yourself, and let the other side narrow it down. Because the source is not a human mind, the usual shortcuts — nobody ever picks 100 — stop working.
Raffles and giveaways by ticket number. A hundred entries is a common size for a small draw. Raise Count and clear Allow repeats to pull several distinct winners in one press, instead of running it repeatedly and throwing out collisions by hand.
Percent-style rolls. Tabletop games, homebrew rules and probability demos lean on a d100 constantly. This page is one, without the pair of ten-sided dice that normally stands in for it.
Spot-check sampling. Picking which of a hundred rows, files or seats to inspect, when checking every one of them is not realistic.
Smaller spans have dedicated pages — 1 to 3, 1 to 4, 1 to 5 and 1 to 10. For any other span, including negatives, the random number generator takes a minimum and maximum you type. To watch the draw land as a spinning wheel rather than a digit, use the wheel spinner.
All randomization on Spinness uses crypto.getRandomValues() — the browser's cryptographic random source. Learn how our randomness works.
Frequently asked questions
Why do people always guess 37 or 73?
Both are odd, two-digit, not round and not near the edges — the profile our brains reach for when asked to sound random. The habit is strong enough that it shows up whenever a large group is asked the same question informally. A generated draw has no such preference: 1, 50 and 100 come up exactly as often as 37 does.
Is every number from 1 to 100 equally likely?
Yes, 1% each. The generator discards any raw draw above 100 and pulls fresh entropy instead of wrapping it around, and that step is what stops the low numbers from collecting extra hits.
Can I draw several numbers at once without duplicates?
Yes. Raise Count and uncheck Allow repeats, and the page returns that many distinct values from 1 to 100 in random order. Keep Allow repeats checked and each draw is independent, so the same number can appear more than once.
Can I use this as a d100?
Yes, for anything that expects a flat 1 to 100 roll. One caveat: percentile dice read as 00 to 99 in some rule sets. If your table treats 00 as 100 the ranges line up exactly, and if it treats 00 as zero you want a 0 to 99 span from the random number generator instead.
Does my number get sent anywhere?
No. The draw happens in your browser and nothing leaves the device. There is no server behind this tool to receive a result, so nothing is logged, stored or shared.