What Is RNG? Meaning, Testing and License Rules
RNG means random number generator, the code that decides each slot spin and card. How test labs check it, and what regulators require for a gaming licence.
ContentsHide contents
RNG stands for random number generator: the software, or occasionally hardware, that decides every slot spin, card dealt and roulette result in an online casino. For a licence, “random” is not a claim you make but a result an independent test laboratory has to demonstrate, and regulators from the UK to Curaçao want to see that evidence before your games go live.
- The RNG picks the outcome and the game maps it to a result. The paytable, not the RNG, sets the payout percentage.
- Most casino RNGs are pseudo-random, so they are only as unpredictable as their seed and their fresh entropy.
- Labs review source code, test final outcomes statistically and probe resistance to attack, at a 99% confidence level under GLI-19.
- Regimes differ sharply: the UK and Malta publish detailed rules, Curaçao accepts other regulators’ standards, and Anjouan states a one-line duty with no named lab or standard.
RNG meaning: what the generator actually does
When a player presses spin, the game asks the RNG for numbers and maps them to a result. On a single-zero roulette wheel it draws a number from 0 to 36 and the game shows that pocket. On a five-reel slot it draws one stop position per reel, and the paytable decides what the combination pays.
The RNG does not set the payout percentage. The game maths does. If a paytable returns 96% over the long run, then €1,000,000 staked returns €960,000 on average. The RNG’s job is to deliver each outcome with exactly the probability the maths assumes. As an illustration, in 3,700,000 spins of single-zero roulette each of the 37 numbers should appear about 100,000 times. A biased generator breaks that without anyone touching the paytable, so labs test the two separately.
The testing standard GLI-19 sorts RNGs into three types. Software-based RNGs derive their randomness from an algorithm. Hardware-based RNGs draw on small-scale physical events such as thermal noise or radioactive decay. Mechanical RNGs, such as wheels and shufflers in a live studio, use the laws of physics. NIST’s SP 800-90A describes the deterministic methods, built on hash functions or block ciphers, and SP 800-90B covers the entropy sources that feed them.
Seeding: where pseudo-random generators fail
A software RNG is deterministic: the same internal state produces the same numbers. Its unpredictability comes from the seed and from any fresh entropy added later. Two illustrations show why regulators care.
A weak seed. A day has 86,400 seconds. If a generator is seeded from the clock and an attacker knows roughly the hour, there are only 3,600 candidate seeds to try. GLI-19 therefore says the RNG shall not be seeded from a time value alone, and that games must not share an initial seed.
A small state. A 52-card deck can be ordered in 52! ≈ 8.07 × 10^67 ways. A 32-bit seed can start only 2^32 = 4,294,967,296 states, so almost every possible shuffle could never occur. A generator that can produce every shuffle needs at least 226 bits of state, since log2 of 52! is about 225.6. GLI-19’s available outcomes rule targets this: the RNG period must be large enough that all outcomes stay possible on every draw.
GLI-19 adds a third attack: if an attacker learns the state once, the RNG must periodically change it using external entropy, which limits how long the exploit lasts. The UK Commission’s guidance asks that seeding and re-seeding never introduce predictability. For hardware RNGs, GLI-19 requires dynamic monitoring that disables play when the output degrades.
What independent testing actually checks
Test houses work in layers, and the UK Commission’s testing strategy and GLI-19 describe a similar sequence.
Documentation and public weaknesses. The lab reviews the RNG documentation and researches the algorithm or hardware for known weaknesses.
Source code. GLI-19 requires review of the core randomness, scaling and shuffling algorithms, compared with published references where they exist. The lab looks for bias, errors, malicious code and undisclosed switches that could influence randomness.
Statistical tests on the final outcome. The tests run on output after scaling, shuffling or mapping, because that is what the player actually gets. The lab picks tests case by case to check the intended distribution and independence between draws. GLI-19 lists chi-square, overlaps, coupon collector, runs, interplay correlation, serial correlation and duplicates tests, evaluated collectively at a 99% confidence level. The UK Commission also asks for tests on the raw output.
Game maths and implementation. For games, the Commission’s strategy adds verification of the maths, rules and theoretical RTP, then testing in a live-like environment: simulation to compare actual and expected RTP, emulation of rare outcomes such as jackpot triggers, and manual play. GLI-19 also forbids a game from adapting its return to past payouts or swapping in a different losing outcome to fake a near miss.
Scaling is where bias hides. Illustration: map a raw value from 0 to 255 onto the 37 roulette pockets with a remainder. Since 256 = 6 × 37 + 34, pockets 0 to 33 each get 7 raw values (a 2.73% chance) and pockets 34 to 36 get only 6 (2.34%), against a fair 2.70%. A perfect generator behind a careless mapping still fails. GLI-19 requires scaling to be unbiased and allows discarding raw values, here anything from 222 upward, to remove the bias.
Standards and regulators: who asks for what
GLI-19 is a laboratory standard, not a law. It applies when a regulator or contract points to it. Version 3.0 (17 July 2020) is the latest final version we could verify. GLI has published a version 4.0 draft dated 7 July 2026, whose RNG chapter keeps the same 99% level and list of tests. Treat it as a preview.
| Regime | What the primary document requires | Who may test | Published technical standard |
|---|---|---|---|
| UK Gambling Commission | RTS 7: RNG and game results must be “acceptably random”, demonstrable through statistical analysis, with no adaptive behaviour. Test reports go to the Commission before release. | A test house on the Commission’s published approved list | Yes: the RTS and the testing strategy |
| Malta Gaming Authority | New games and RNG changes need an RNG test certificate. Online B2C games must pay out an average of at least 85% unless the licence says higher, and the licensee must monitor it. | Independent lab based in the EU or EEA, per the MGA form | Rules sit in the directive and forms; the documents we read set out no test methods |
| Curaçao Gaming Authority | Licence conditions: every remote game of chance, own or third-party, must be certified for soundness and integrity. | Labs holding an article 5.17 LOK certificate, with ISO/IEC 17025 accreditation covering RNG and RTP testing | None of its own; it accepts standards of New Jersey, Pennsylvania, Ontario, British Columbia, the UK and Malta |
| Anjouan Gaming | Its standards page says RNGs in gaming products must be certified by an approved independent testing facility. | Unnamed facility; we found no published lab list | None found published |
The UK is the most prescriptive. Its testing strategy lists what an RNG report must contain, including the RNG version, hardware or software type, digital signature, tests applied and any limits on use. Our UK licence requirements guide covers the rest of that regime.
Malta’s directive also lets the Authority accept certificates issued under another EU or EEA state’s standards, and says a B2C licensee need not resubmit a B2B licensee’s game certification. Our Malta gambling licence service shows where that sits in the file.
Curaçao needs care. The CGA says in its test-house guidelines that it has not yet established its own technical standards, so in practice a Curaçao file is tested against one of the recognised ones. Agree in writing with your lab which one before testing starts. The CGA’s certificate list of 31 August 2026 shows five test labs, and our Curaçao licence check guide explains how to read it. The Curaçao licence service page outlines the full application.
Anjouan is thinner still. Its licence conditions page says nothing about games, and its standards page gives one sentence on certified RNGs. We found no published technical standard, lab list or test method, so confirm each expectation with the licensor in writing. Our Anjouan requirements guide covers the wider document pack.
What a certificate does and does not prove
A certificate is a snapshot: a named lab tested a specific version against a stated standard and found no unresolved non-conformance on that date. Four limits follow.
It covers that version only. The UK Commission requires external retesting of updates that affect game fairness. When an RNG or a remote gaming server changes, the licensee must retest a representative sample of games spanning each game type and generation.
Statistics cannot prove safety alone. NIST calls its statistical test suite a first step and says no set of tests can certify a generator for a given application. That is why source review and attack analysis sit beside the numbers.
It does not guarantee live payouts. GLI-19 expects the operator to compare theoretical and actual RTP, investigate large variances and monitor RNG output. Malta requires B2C systems to monitor the average payout, and the UK testing strategy includes live RTP monitoring.
It is not an endorsement. The CGA says listing a test house is not an endorsement or guarantee. A certificate is likewise evidence, and your own controls have to keep it true.
What this means for your licence application
Expect the regulator or auditor to ask for:
- The certificate in the right name. Malta asks for an RNG test certificate in the name of the applicant or licence holder, issued by an independent EU or EEA lab. Curaçao requires certification by a CGA-approved lab and expects you to disclose third-party game suppliers on its portal. In the UK, the lab report goes to the Commission before release.
- The full lab report, not just the certificate. Regulators want the scope, tests and limits behind it.
- Game maths for every title. Keep documentation, such as PAR sheets, showing each game’s theoretical RTP, with a log of every change. GLI-19 treats a changed RTP as a new game for reporting.
- A live monitoring procedure. Show how you compare actual and theoretical RTP and who reviews variances.
- Supplier certificates, game by game and version by version. If you buy content, the studio’s certificates carry your file. Check each names a lab your regulator accepts. Our B2B supplier licence guide explains what suppliers must hold on their side.
- A change-control record. Show which updates trigger retesting and who signs them off.
GLI’s questionnaire shows what a lab asks a supplier up front: the core algorithm (options include AES, Java SecureRandom, Mersenne Twister and LCG), whether the RNG source compiles separately so digital signatures can be taken, and whether output is dynamically monitored. GLI-19 v3.0 expects documentation, credentials and access to a production-equivalent test environment, and the 2026 draft adds source code.
The technical file is half of the pack; the other half is compliance, covered in our primer on iGaming AML and KYC. Our iGaming certification page covers the testing side.
Crypto casinos add a second fairness mechanism that players can check themselves, covered in provably fair casinos explained.
If you are not sure whether your platform’s RNG, or a supplier’s certificates, will be accepted by the regulator you are targeting, that is a scoping call, and it is free.
Frequently asked questions
Select a question to open the answer.
What is RNG in gambling?
RNG stands for random number generator. It is the software or hardware that picks the outcome of every spin, card and roll in an online casino. The game then maps the number to a symbol, a card or a roulette pocket, and the paytable decides what that result pays.
Is a casino RNG really random?
Most are pseudo-random: a deterministic algorithm whose output depends on a seed and an internal state. That is acceptable if the generator is cryptographically strong, seeded properly and refreshed with outside entropy. Hardware generators use physical noise instead, and the testing standard GLI-19 requires dynamic monitoring of their output so play stops if they degrade.
Who tests an RNG and what do they check?
An independent test laboratory reviews the source code, runs statistical tests on the final outcomes after scaling and shuffling, and checks that the generator resists attack. For games it also verifies the maths, the paytable and the theoretical RTP. Which labs a regulator accepts differs by regime.
Do I need an RNG certificate to get a gambling license?
If you offer RNG-based games, yes in the regimes we checked: the UK, Malta, Curaçao and Anjouan all expect independent certification. If you run only third-party games, you rely on each studio's certificates, but you still have to supply them and keep them current.
What is the difference between RNG and RTP?
The RNG decides which outcome happens on a single play. RTP, return to player, is the long-run percentage of stakes the game's maths pays back. A fair RNG delivers the outcomes with the probabilities the maths assumes, so RTP is a property of the paytable and the RNG together.
Does an RNG certificate prove a game is fair?
It proves that a named lab tested a specific version against a stated standard and found no non-conformance on that date. It does not cover later versions, does not guarantee actual payouts in live play, and statistical tests alone cannot prove a generator is safe. Regulators therefore also expect change control and live monitoring.
Sources
Sources (14)Hide the sources
- GLI — GLI-19 Standards for Interactive Gaming Systems, v3.0 (17 July 2020)
- GLI — GLI-19 v4.0 draft (7 July 2026)
- GLI — RNG submission questionnaire
- UK Gambling Commission — RTS 7, Generation of random outcomes
- UK Gambling Commission — Testing strategy, procedure for testing
- Malta Gaming Authority — Approval of New Games, enclosure requirements
- Malta Gaming Authority — Player Protection Directive (Directive 2 of 2018, v3 January 2023)
- Curaçao Gaming Authority — Online gaming licence conditions, indefinite term
- Curaçao Gaming Authority — Guidelines for the Certification of Independent Gaming Test Houses, v1.0
- Curaçao Gaming Authority — List of certificate holders under article 5.17 LOK (31 August 2026)
- Anjouan Gaming (licensor) — Regulatory standards
- NIST — SP 800-90A Rev. 1, Random number generation using deterministic random bit generators
- NIST — SP 800-90B, Entropy sources used for random bit generation
- NIST — SP 800-22 Rev. 1, Statistical test suite for random and pseudorandom number generators
Related service iGaming certification →
This article is for general informational purposes only and is not legal, tax or financial advice. Consult a qualified professional before acting.
Stay ahead of the rules.
Licence fees, rules and deadlines change through the year. Subscribe and we'll email you Vantegris licensing updates.
Licence, done right.
300+ operators licensed across 40+ jurisdictions. Book a free consultation.
-
Iryna H.Gaming Licensing -
Dariia Y.Junior Associate, Crypto Licensing -
Aleksandra A.Paralegal, Gaming Licensing -
Maksym D.Paralegal, Banking & Payments -
Diana K.Paralegal, Compliance & Legal -
Myroslava V.Business Development Assistant