What does randomness mean?
Randomness can describe where an outcome comes from, how a sequence is generated, or how a sample is selected. Those are related questions, but they are not interchangeable. A sequence can look irregular while still being generated by a predictable algorithm, and a physical source can contain measurable bias that needs correction.
The most useful classification for everyday computing separates physical randomness, ordinary pseudorandomness, cryptographically secure pseudorandomness, and quasi-random sequences. In statistics, the word random often refers instead to a defined sampling or assignment procedure.
What is physical or true randomness?
A physical random source measures an event that is not produced by a deterministic software formula. Examples include electronic noise, radioactive decay, photon behavior, or timing variation in physical systems. The measured signal is converted into bits that software can use.
Calling a source physical does not automatically make its output uniform or independent. Sensors can introduce bias, environmental changes can affect measurements, and hardware can fail. Practical systems therefore need health checks, conditioning, and a clear account of how the measurements become random bits.
Physical randomness is useful when unpredictability must not depend only on a software state. It can also seed a cryptographic generator, which then expands limited physical entropy into a larger stream of usable values.
What is pseudorandomness?
A pseudorandom number generator, or PRNG, uses an algorithm and an internal state to produce a sequence. The sequence may pass statistical tests and look unpredictable to a casual observer, but it is deterministic: the same algorithm started from the same state produces the same output.
Reproducibility is often a benefit. Simulations, software tests, procedural game worlds, and classroom demonstrations may need the same sequence again so results can be checked. A known seed makes that possible.
Ordinary PRNGs are not automatically suitable for passwords, session tokens, encryption keys, or other security-sensitive values. If an attacker can infer the state, later values may become predictable. JavaScript's Math.random() is intended for general-purpose randomness and does not promise cryptographic security.
What is cryptographically secure pseudorandomness?
A cryptographically secure pseudorandom number generator, or CSPRNG, is designed so that observing its output does not make past or future output feasibly predictable. The operating system gathers entropy from available sources, maintains secure internal state, and exposes random bytes to applications.
Web browsers provide this type of interface through Crypto.getRandomValues(). It is the appropriate browser API for generating security-sensitive random values. It is still a software interface, not a claim that every returned bit was measured directly from a physical event at that moment.
LetsRandomize's shared randomization engine uses browser cryptographic randomness when it is available and falls back to Math.random() for older environments. That approach improves everyday list shuffling and selection, but the tools are not independently audited systems for regulated drawings, gambling, or security key generation.
What are quasi-random sequences?
Quasi-random, or low-discrepancy, sequences are deliberately designed to cover a space evenly. They are deterministic and are not intended to be unpredictable. Their value is coverage: in some numerical integration and simulation problems, evenly distributed points can converge faster than independent random samples.
This is an important example of why "looks random" is not a complete definition. A low-discrepancy plot may look highly ordered compared with random points, yet it can be more useful for a calculation that needs even coverage.
What does random mean in statistics?
In statistics, random usually describes the selection or assignment mechanism. A random sample requires a defined population and a method that gives eligible units known selection probabilities. A varied-looking group is not necessarily a random sample if people volunteered, were easy to reach, or were filtered without accounting for the change.
Random assignment is different from random sampling. Sampling determines who enters a study; assignment determines which condition an included participant receives. Random assignment can reduce systematic differences between groups, but it does not by itself make the participants representative of a larger population.
Statistical tests can detect certain patterns, imbalance, or dependence in a sequence. No finite battery of tests proves that an unknown process is truly random. Evaluation needs both observed output and knowledge of the generator, implementation, state management, and threat model.
Which type should I use?
Use an ordinary seeded PRNG when reproducibility matters and there is no security threat, such as repeatable tests or simulations. Use a CSPRNG for passwords, tokens, keys, and situations where prediction would create harm. Use a physical source when the application specifically requires independent entropy and the hardware and extraction process can be evaluated. Use quasi-random sequences when even coverage is more important than unpredictability.
For casual choices such as shuffling a list or picking a classroom order, a browser-based randomizer can be convenient. For legal, regulated, financial, or high-value selections, follow the required process and use an independently reviewed system with appropriate records.