Slot Game Certification: GLI-19, RNG Testing and the Package a Studio Prepares

A finished slot is not a product until someone independent has confirmed that it does what its maths sheet says. In regulated markets that someone is a test laboratory working to a published standard — most often GLI-19 for online games — and the process is a mix of document review, source inspection, statistical testing and a lot of waiting. This is what certification actually involves, what the lab asks for, and how a studio prepares so the wait is weeks rather than months.

Who certifies and against what

Three names cover most of the industry: Gaming Laboratories International (GLI), BMM Testlabs and iTech Labs, with regional labs such as eCOGRA, Trisigma and Quinel active in specific markets. The standards they test against are either the lab’s own (GLI publishes its series; GLI-19 is the Interactive Gaming Systems standard, GLI-11 covers gaming devices) or the regulator’s technical requirements (the UK Gambling Commission’s RTS, Malta’s technical standards, Italy’s ADM rules, Ontario’s iGO standards). A game certified for one jurisdiction is not automatically valid in another; each regulator decides which reports it accepts.

Note the distinction with platform review. A quality gate such as Stake Engine’s pre-release rating is not a regulatory certification; it checks that the game behaves as documented on that platform. A studio targeting licensed European operators needs the lab report on top.

What the lab tests

1. The random number generator

RNG certification is the foundation. The lab examines the algorithm, seeding, scaling (how raw numbers are mapped to reel positions without bias), reseeding and the handling of the number stream between games. Statistical batteries — chi-square, runs tests, the Diehard and NIST suites or the lab’s own — are run on large samples. If the RNG is the platform’s rather than the studio’s, its existing certificate can be referenced, which is one reason platform choice matters.

2. RTP and the maths

The lab reproduces the PAR sheet: it recomputes base-game combinations from the reel strips and paytable, checks feature contributions by simulation, and compares its results with the studio’s. Every bet mode is checked separately. If the game has a configurable RTP (say 94%, 96% and 97% builds), each configuration is verified and listed on the certificate.

3. Game rules and display

Everything the player is told must match the maths: paytable pages, feature rules, the displayed RTP, max win, and — in some markets — the odds of the top prize. Labs read the help screens line by line.

4. Session and error handling

What happens on disconnect mid-feature, on a double-click of spin, on a currency change, on a malformed server message. Standards require that no interruption can change the outcome of a round or lose a player’s money, and the lab tries to make it happen.

5. Responsible-gambling and jurisdiction rules

Spin speed minimums, autoplay limits or bans, reality checks, bonus-buy restrictions, and display requirements that differ by market. These are usually implemented as configuration flags per jurisdiction, and each flag set is tested.

What the studio prepares

The submission package, in the order labs tend to ask for it:

1. Game description and rules — the design document reduced to what the player is told.
2. PAR sheet / maths specification — reel strips, paytable, combination tables, feature analysis, summary with RTP, hit frequency, volatility and max-win probability, per bet mode.
3. Simulation report — number of spins simulated (tens of millions at minimum; we run billions), empirical RTP with confidence interval, feature frequencies.
4. Source code or build — the maths module and the client, with a fixed version hash. Any change after submission means re-testing.
5. RNG documentation — algorithm, seeding, scaling; or the certificate of the platform RNG.
6. Jurisdiction configuration matrix — which flags are set for which market.
7. Test harness — a way for the lab to force outcomes (specific reel stops, feature triggers) to test rules and edge cases without waiting for randomness.

The single most valuable item is the last one. A forcing tool that lets the lab set the RNG output turns a two-week rules review into two days.

Timelines and money

A clean submission for one jurisdiction typically takes four to eight weeks from receipt to report, longer in busy seasons. Costs sit in the low-to-mid thousands of dollars per game per jurisdiction for RNG-plus-RTP testing, and rise with the number of markets and with every resubmission. The expensive failures are avoidable: a paytable screen that disagrees with the sheet, a feature contribution that was estimated rather than simulated, a client that loses state on reconnect.

How to design for certification from day one

• Write the compliance section of the design document before art begins: markets, standards, flags.
• Keep the maths module separate from presentation so the lab can test it in isolation.
• Version everything; the hash you submit is the hash you ship.
• Build the forcing tool during development, not at submission time.
• Simulate more than the lab will: billions of spins cost hours of compute and buy credibility.

Certification support — preparing the package, answering lab questions, and building the harness — is part of full production at Giro Games and available for games we did not build. It is listed under HTML5 slot game development and process and terms on the services page.

Background reading (German, Italian): our earlier article on Germany’s PTB type-approval pipeline, and Italian coverage of ADM certification requirements for online games — both describe how national rules diverge from the lab standards.