How To Create A Random Number

How to Create a Random Number (and Why It's Trickier Than It Sounds)

You pick a number between 1 and 10. Because of that, easy, right? Now try doing that a million times in a row without a pattern creeping in. Harder than you'd think. And that's the whole reason we don't actually leave randomness up to humans — we use machines. But how machines make "random" numbers is a story worth knowing, because not all randomness is the same, and picking the wrong method can quietly mess up simulations, games, encryption, and even research.

What "Random" Actually Means Here

When people talk about creating a random number, they usually mean one of two things:

• True randomness — generated from a physical process you can't predict, like atmospheric noise or radioactive decay.
• Pseudo-randomness — generated by a mathematical formula that looks* random but is technically deterministic. Given the same starting point, it'll produce the same sequence every time.

Most of the random numbers you encounter day to day are pseudo. And that's usually fine. But for cryptography, lotteries, or anything where predictability is dangerous, you need the real stuff. Knowing which one you actually need is the first step, and most people skip it.

Why People Care About Random Numbers

You'd be surprised how much rides on a coin flip.

Online games use random numbers to decide what loot drops. Scientists use them to select samples for studies. Day to day, cryptography uses them to generate keys that keep your messages private. Statisticians use them to model uncertainty. Even a basic playlist shuffle on your phone is running some kind of random selection under the hood.

The thing is, when randomness fails, the failure is often invisible. In real terms, a biased random number generator doesn't crash. It just slowly skews results, and nobody notices until something goes subtly wrong — a game that feels rigged, a survey that over-represents one group, a security vulnerability nobody caught.

How Random Numbers Actually Get Made

There are a few common approaches, and each has its place.

1. Using a Programming Language

Almost every language has a built-in random function. In JavaScript, Math.Practically speaking, in C++, you'd use functions from the <random> library. random(). randint(1, 100). In Python, it's random.These are fast, convenient, and good enough for most everyday tasks.

But here's the catch: most of these are pseudo-random. They start from a seed* value (often the current time, internally) and run a formula to produce a stream of numbers that pass basic statistical tests. For a dice-rolling app or picking a random winner from a list of 100 emails, that's perfectly fine. Don't overthink it.

2. Hardware Random Number Generators

These are the gold standard for true randomness. They pull entropy from physical phenomena — thermal noise in a circuit, timing variations between processor cycles, or even dedicated chips that sample quantum effects. Linux has /dev/random and /dev/urandom, and on some systems, the OS feeds real hardware entropy into the mix.

You don't see this directly unless you're working in cryptography or systems programming, but it's there in the background, powering things like TLS connections and disk encryption keys.

3. Online Random Number Tools

Websites that generate random numbers for you usually fall into one of two camps. Some are pure JavaScript, running a pseudo-random algorithm in your browser — fine for casual use, not fine for anything sensitive. Others pull from a server-side source that's closer to true randomness, often using atmospheric noise or hardware sources.

If you're just picking a number for a board game or deciding who buys lunch, the casual tools work. If you're using it for anything that needs to be unpredictable to an attacker, look for a tool that explains where* its entropy comes from.

4. Physical Methods

Old school still works. Humans are bad at being random ourselves (we unconsciously avoid repeating patterns), but physical objects don't have that bias. Dice, coins, drawing slips of paper from a hat. A six-sided die is genuinely close to uniform. Spinning a well-balanced wheel, drawing from a shuffled deck — these are surprisingly good, as long as the physical process is fair.

Common Mistakes People Make With Random Numbers

This is where things usually go sideways.

Using Math.random() for Security

In JavaScript, Math.random() is not cryptographically secure. It uses a PRNG (pseudo-random number generator) that's fast but predictable. If you're generating a password reset token, a session ID, or anything an attacker could exploit, use crypto.getRandomValues() instead. Same goes for most languages — there's usually a "secure" variant of the random function specifically for this. Using the wrong one is one of the most common bugs in beginner crypto code.

For more on this topic, read our article on what time will it be in 16 hours or check out how many days until may 9th.

Reseeding Without Knowing Why

Some PRNGs accept a seed. Setting it to a fixed value (like 42) makes the sequence reproducible — which is great for testing but terrible for security. A common mistake is hardcoding a seed in production "just to make debugging easier" and forgetting to remove it.

Assuming Uniform Distribution

Most basic random functions give you uniform distribution — every number is equally likely. A common pattern: take several random values and average them to get a bell curve. But if you want, say, a random number that skews toward* higher values, you have to do extra work. That's why it doesn't. In real terms, people sometimes assume random() already does this. Or use a specific distribution function if your language offers one.

Reimplementing the Algorithm

Please don't. Writing your own random number generator is a fun exercise, but the well-tested algorithms (Mersenne Twister, PCG, ChaCha20 for crypto) have been analyzed and improved over years. Yours will have subtle biases. Use the library.

Forgetting the Edge Case at Zero

In some languages, random functions can return 0 (or the upper bound). If your code does something like array[random_index], and random_index is 0, that's fine — but if you later divide by it or use it as a multiplier, you'll get a bug. Small thing, but it bites people regularly.

Practical Tips That Actually Help

Pick the Right Tool for the Job

The single biggest factor in "how do I create a random number" is what you're using it for. Picking a raffle winner at the office? On top of that, generating an API key? Now, choice()in Python. Practically speaking, running a Monte Carlo simulation?secrets.Think about it: token_hex(32). On the flip side, random. Numpy has fast vectorized random generation that outperforms a Python loop by orders of magnitude.

Seed Reproducibly When Testing

If you're running a simulation or a randomized test suite, set the seed to a known value. That way, when something goes wrong, you can reproduce the exact sequence and debug it. Just remember to remove the fixed seed before deploying.

Visualize When You're Not Sure

If you're generating a lot of numbers and want to verify the distribution looks right, plot a histogram. Biases show up fast visually even when they're not obvious numerically. For a quick sanity check, generating 10,000 numbers and bucketing them should look flat. If it doesn't, something's off.

Don't Confuse "Random" With "Unique"

Sometimes people ask how to generate a random number when they really mean "how do I pick a unique value from a set?Random selection with replacement can give you duplicates. " Those are different problems. If you need uniqueness — like generating a list of distinct IDs — you need a different approach: shuffle a list, sample without replacement, or use a UUID.

Use UUIDs When You Need Identification

If you need a "random" identifier — say, a unique order number or a database row ID — reach for a UUID or a similar scheme. They're designed to be unique across systems, not just within yours. A plain random number has too high a chance of collision once you're generating millions of them.

FAQ

Is there a truly random number?

Yes, but only if you use a physical source of entropy. Day to day, atmospheric noise, quantum measurements, and radioactive decay are all genuinely random as far as physics can tell. Anything generated by a deterministic algorithm is technically pseudo-random, no matter how good it looks statistically.

What's the difference between random and secrets in Python?

random uses a Mersenne Twister PRNG. Here's the thing — secrets uses the operating system's secure entropy source, which is unpredictable to outside observers. It's fast and good for simulations, games, and non-sensitive use. Use secrets for anything related to passwords, tokens, or security.

Can a computer ever be truly random?

Not purely through software, no.