True random number generator

How it works

A true random number generator, sometimes abbreviated TRNG, measures a messy physical process and turns that measurement into bits. Atmospheric radio noise is one public example. Other designs watch electronic jitter, photon arrival, or similar unpredictable physical events. The point is that the next bit is not computed from the previous bits by a published recipe you could replay. If you could replay it from a seed, you are not looking at this class of tool.

Those bits are then shaped into the form you asked for: an integer in a range, a coin, a shuffle of a list. Shaping has to be done carefully. A sloppy “take a big number and remainder N” can favor some residues if the source range is not a multiple of N. For a classroom raffle you will not see that. For a protocol you should not ignore it. Inclusive bounds still matter. If you asked for 1 through 6, 1 counts and 6 counts. A physical source does not repair a range you typed wrong.

Independence is the usual default for a stream of integers: each draw does not remove the last value, so repeats can appear, the way separate dice rolls can match. A sequence without repeats is a shuffle of a finite set, which is a different request. Say which you wanted. A true source that returns 4, 4, 4 in a short 1-to-6 run can still be honest. A generator that secretly avoids the last face is no longer a memoryless die.

What “true” is not

“True” is not a moral upgrade on a party picker. For a board-game turn, a classroom volunteer, or a writing prompt, a decent pseudorandom function is already hard to guess by eye. You reach for a true random number generator when someone will argue about the source, when you need bits that should not be reproducible from a seed, or when the job is cryptographic and the generator is built and reviewed for that job. A widget labeled “true” on a random website is a claim. Read how it collects entropy. If it does not say, treat it as unknown, not as physics.

It is also not automatic fairness. Fairness is the range, the list, and the replacement rule. Atmospheric noise will happily pick the person who should have been ineligible if you left them on the list. It will happily pick 31 if you typed 1–31 and meant 1–30. Publish the setup with the result if the group needs a record: source, range, count, timestamp, output. A screenshot of a different app is not this draw.

Name the range or the list, including both ends. State if a value may appear more than once. Use a generator whose entropy source you can actually describe. Keep the output with that description if anyone else must trust it.

A radio can pick a number you will not replay. It cannot pick the range you meant if you typed the wrong one.

When a formula is the right tool anyway

Simulations, games, and tests often need a stream you can replay. A seed plus an algorithm gives you that. That is a feature. A true random number generator is a poor fit when a bug report must recreate the same shuffle, when a unit test must be stable, or when you want the same classroom warm-up tomorrow. Use the physical source when unpredictability to an observer is the point. Use the seeded formula when repeatability is the point. Mixing the two and then being surprised is a category error, not a hardware failure.

Do not harvest a public true-random page by scraping it in a tight loop. Services that offer noise-based integers meter heavy use and publish an API for a reason. A handful of numbers in a browser is the ordinary case. A silent script that hammers the form is how hosts get blocked. If you are writing software that needs cryptographic randomness, use the operating system’s CSPRNG, not a toy page, and not a homemade mix of mouse wiggles you have not had reviewed.