Independent test houses check that a game's random number generator is unpredictable, non-repeatable and uniformly distributed, and that the game implements the rules and probabilities it publishes. In Great Britain, operators must have new products tested by a test house from the Gambling Commission's approved list before release, and face an annual audit covering change controls and live RTP monitoring. Monthly RTP reports measure what games actually returned across all recorded play in a period, and state that they do not describe individual or future games.
On this page
- Approved testers
- A published, finite list of test houses named by the GB regulator, updated 2 March 2026Source 2
- Pre-release rule
- New products must be tested by an approved test house before releaseSource 1
- RNG certificate scope
- Unpredictable, non-repeatable, uniformly distributed output; component-level testing onlySource 6
- RTP audit method
- Total wins divided by total bets from game logs, generally monthlySource 5
- Annual audit
- Change controls, game list and live RTP monitoring, submitted to the regulatorSource 4
What "tested" means in a regulated market
Game testing is a documented process with a named tester, a scope and a date. It is not a badge.
In Great Britain the requirement sits with the operator. Licensees must ensure all new products have been adequately tested by an approved test house prior to release1. Where third-party testing is required by the testing strategy, licensees must appoint a test house from the Commission's published list2.
That list is public and finite. As published on 2 March 2026 it named eCOGRA Limited, several Gaming Laboratories International entities, BMM Testlabs, Gaming Associates Europe, Quinel, Global Lab and Nick Farley & Associates among others, with each entry showing the scope of work that house is approved for2.
What actually gets tested
Three layers are worth separating, because they are usually blurred together.
The random number generator
The random number generator is the engine that produces outcomes. In Great Britain it must satisfy RTS 7, which requires that for a software RNG it should be computationally infeasible to predict what the next number will be, that generation does not reproduce the same output stream, that output is uniformly distributed over the entire output range, that seeding and re-seeding do not introduce predictability, and that any scaling applied to the output maintains those qualities3. The mapping of random inputs to game outcomes must be in accordance with the prevailing probabilities3.
A published RNG certificate is the evidence of that check. A real example makes the scope concrete: one iTech Labs certificate states that the RNG was evaluated and complies with the relevant standards, that the numbers generated are unpredictable, non-repeatable and uniformly distributed, and that they passed Marsaglia's "diehard" tests for statistical randomness, with the evaluation carried out against the UK remote gambling technical standards and testing strategy6.
The same certificate carries the limitation that matters most: while it is not possible to test all possible scenarios in a laboratory environment, the laboratory conducted a level of testing appropriate for a component test of that type6.
The game around the generator
A correct RNG can still sit inside a game whose paytable does not match its rules. That is why testing covers the game as built, and why changes are classified. The Commission's strategy distinguishes major changes, which require external retesting, from minor ones that can be handled internally, and applies a risk-based approach to how much assurance it demands1.
Standards published by test houses describe the same territory in more detail. GLI-19, the standard for interactive gaming systems, was revised to version 3.0 in July 2020, and one of the stated aims of that revision was to distinguish technical requirements evaluated in the laboratory from operational controls and procedures evaluated on site after installation7.
That distinction is worth holding on to. Laboratory work checks the software. Site work checks how the business runs it.
The live operation
Testing does not stop at launch. Licensees who develop or update games are subject to an annual audit by an approved test house, which checks adherence to change controls, confirms the list of games made available to consumers under the Commission licence, and confirms that licensees have effective live return to player monitoring processes in place4. Results must be counter-signed by the relevant personal licence holder or specified person and submitted to the Commission4.
What an "RTP verified" report actually measures
Monthly RTP reports are a different exercise from certification, and they answer a narrower question.
iTech Labs describes its RTP audits as extracting total bets and wins for each game from game logs and calculating actual RTPs, meaning the percentage of total actual winnings to total bets for each game, with sanity checks on the consistency and completeness of the extracted data5. RNG audits compare observed probabilities for game events against theoretical probabilities, with 95% confidence limits calculated5. The audits are generally done monthly and the reports are published on the relevant gaming site5.
Two things follow.
First, an RTP report is a measurement of what happened across all recorded play in a period. The report itself notes that results do not reflect outcomes for individual games or future games5. It is a statement about a population of spins, not about a session.
Second, a measured figure and a designed figure are different numbers that converge only with volume. our RTP guide sets out that relationship, including why measuring a game with very little play produces a meaningless result.
Live dealer games and the limits of RNG testing
Not every game outcome comes from software. eCOGRA notes that RNGs do not apply to live dealer games, because those games use real dealers and physical equipment such as cards and roulette wheels to determine the outcome8.
Live tables are therefore assured in a different way, through studio procedure, equipment and surveillance rather than through an RNG certificate. how live casino games work covers how those studios operate.
What testing does not cover
- Whether an operator pays out. Certification concerns the product, not the balance sheet or the withdrawal queue.
- Which version of a game is in use. Where providers release several RTP versions, the choice is a commercial one made by the operator.
- Bonus terms. Wagering conditions are contractual, not mathematical certifications; see how wagering requirements work.
- Your session. No certificate makes an outcome due, and no audit changes the probabilities of the next spin.
- Legality where you are. A tested game is not a lawful game everywhere, a point that matters in restricted markets such as the Maldives.
What testing does cover is narrow and useful: that the generator behaves randomly under laboratory conditions, that the game implements the rules and probabilities it publishes, and that the operator has controls around changes and monitoring. Rules and likelihood information must in any case be made available to customers before they commit to gamble in Great Britain9.
The KiraBet approach to recording this is set out in our review methodology: we record a certificate where the tester or the provider publishes it, and we leave the field empty where nobody does.
- Random number generator (RNG)
- Hardware or software that generates the random numbers used to decide game outcomes. In regulated markets, RNG output is tested to show that it is statistically random and unpredictable.
- RTP
- RTP (return to player) is the percentage of all money staked on a game that it is designed to pay back as winnings, averaged over a very large number of plays. It does not predict the result of any single session.
- Provably fair
- A method, mainly used in cryptocurrency-based games, that lets players check a result using cryptographic hashes of values committed to before the round. It shows the outcome was not changed after bets were placed, but it does not change the house edge.