Regulators do not care how polished your reel animation looks if they cannot reconstruct how a spin was decided. At Lapland Lucky we treat the random number generator as a pipeline: entropy in, outcomes out, signatures in between, and the browser as a presenter that never re-rolls the dice.
This article explains the architecture we ship for Canadian and wider APAC operators who need GLI-11 traceability without adding perceptible latency to the spin button.
Why server-side outcomes still matter in HTML5
Client-only RNGs are fine for social demos, but regulated real-money play expects outcomes generated on infrastructure you control. Our default pattern uses an HSM-backed AES-CTR generator on the game server, producing a complete spin payload — reel stops, feature flags, win amounts — before any animation frame renders.
The payload is hashed and signed. The client verifies the signature, caches the outcome, and only then drives PixiJS timelines. If verification fails, the round aborts and support receives a structured incident record. That single decision removes an entire class of dispute where players claim the animation "changed" the result.
Entropy and reseeding policy
We combine OS crypto.getRandomValues() stir-ins
with hardware entropy on the host. Reseeding happens on a
counter basis, not on wall-clock intervals, and every reseed
event is logged with a monotonic sequence number auditors can
correlate to build IDs.
- Never seed from timestamps, session IDs or player nicknames.
- Run NIST SP 800-22 batteries on staging weekly, not only before lab submission.
- Document the RNG module boundary in your technical file — reviewers look for it first.
Bridging math and presentation
Math engineers publish outcome schemas; client engineers consume them as immutable structs. Bonus wheels, pick-em layers and free-spin counters all arrive in the first payload so the UI cannot invent intermediate prizes. When we add a new feature, the schema version bumps and older clients refuse incompatible rounds.
"If your QA team cannot replay a spin from JSON alone, your pipeline is not ready for Toronto go-live." — Priya Nair, CTO
What we hand to test labs
Labs receive seed-change procedures, source snapshots, statistical output dumps and API examples showing how operators request outcomes. Packaging this early avoids the second-round fees that sting Canadian studios operating on tight launch calendars.
Need an RNG review before certification? Our math team offers fixed-scope audits — get in touch.