Certified RNG in Virtual Horse Racing Explained

What “Certified RNG” Really Promises

The phrase “certified RNG” lives on more help pages than I can count, and it almost never means what the marketing department thinks it means. After eight years auditing how virtual racing products talk about their own machinery, I have learned to read it as exactly one thing: the source code that produces race outcomes has been examined by a UKGC-approved test house, and the operator displaying that code’s output runs under a Combined Remote Operating Licence. That is a real promise — but it is also a specific, bounded one.

“Certified” does not mean every round is fair to your bet. It means the long-run statistical behaviour of the engine matches what the supplier declared, that no operator-side code can override an outcome once the seed is set, and that an independent body has stamped a piece of paper saying so. UK software providers must pass testing at one of two main labs — eCOGRA or Gaming Laboratories International (GLI) — which UKGC officially approves for the role. Without that certification, there is no operating licence, and without the licence, there is no legal virtual racing product to sell in the UK.

What I care about more, when I look at a new product, is what the certification regime cannot deliver. It does not promise the favourite will win as often as a real-race favourite. It does not promise a particular punter will see normal variance over their first hundred rounds. It does not promise the published RTP applies to their personal bet mix. Those are statistical statements about populations, not about an individual session.

The reason this distinction matters in 2026 is that several operators applying for or holding a UKGC Combined Remote Operating Licence have made virtual sports a major slice of their portfolio. BetConstruct’s UK licence covered eight virtual products at launch, including Horse Racing and Greyhound Racing. The phrase “certified RNG” appears on each of their product pages. It is true, and it is incomplete. The job here is to explain what is inside the certificate.

How an RNG Produces a Race Result

The cleanest way I have found to explain this is by walking through a single virtual round. Imagine the supplier’s server is about to generate a six-runner Flat race. The engine does not “pick a winner” in any normal-language sense; it pulls a string of pseudo-random numbers from a seeded generator, maps those numbers against a probability table the supplier built, and lets the resulting outcomes drive the animation that the punter sees.

A pseudo-random generator is not, strictly, random. It is deterministic — a long, complex arithmetic recipe — but seeded with enough environmental entropy at startup that no observer can predict the next output. Hardware RNGs (which lean on physical noise) and software RNGs (which lean on cryptographic primitives) both qualify, and the UKGC’s Remote Technical Standards do not mandate one or the other. They mandate behaviour: unpredictability, non-repeatability across reasonable sequences, statistical uniformity in distribution, and so on.

The next layer is where it gets interesting. The raw RNG output is uniform across some numeric range. But virtual horse racing does not want uniform outcomes — it wants outcomes that reflect the realistic spread of a six-runner race with one short-priced favourite and three longer-priced rivals. That requires a probability model on top of the RNG. The supplier assigns weights to each runner; the RNG output is mapped into those weighted bins. A favourite at 6/4 wins more often than an outsider at 25/1 because the bin assigned to it is bigger, not because the RNG is biased.

Critically, the seed and the weighted mapping are locked before the round shows on the punter’s screen. There is no “watch them place their bet and then nudge the outcome” pathway in a certified engine. Auditing focuses heavily on this: when does the seed get drawn, when does the outcome get computed, what is the time gap between outcome computation and outcome display, and is there any operator-side hook that could rewrite anything in between? That is the boring, technical heart of what gets signed off.

The animation, paradoxically, is the part of the round that has nothing to do with fairness. The graphics engine renders horses moving toward a finish line that was decided milliseconds earlier. Some products even pre-compose the visuals from a fixture pool — the pre-recorded model, used by Mohio Gaming and discussed in more depth in our piece on RNG labs and what they actually test. Either way, the visible race is downstream of the result, never upstream of it.

Where the Probability Weights Come From

This is the question I get asked least and that matters most. Auditors check that the RNG produces uniform outputs and that the mapping table behaves as declared. They do not — at least, not in the public-facing way — opine on whether the underlying probability weights model “realistic” horse racing.

That is by design. Probability weights are a content decision, the same way the symbols on a slot reel are a content decision. The supplier picks what kind of race shape it wants to portray. A Flat product might use weights modelled on UK Flat short-distance handicap data; a Sprint product compresses the tail and gives the favourite a higher win rate. The supplier chooses; the RNG executes; the auditor verifies execution.

What punters often expect — and what the regime is not built to enforce — is that virtual racing weights mirror real Flat or Jump statistics. They sometimes do, sometimes do not. Suppliers tend to engineer the weights so the displayed odds look like a sensible real-race book, but the win rates can be subtly different. An 8/1 second-favourite on a virtual card is paid 8/1 if it wins, but the supplier might have weighted it to win 1 time in 12 rather than 1 time in 9 to recoup margin. That is where the overround lives, hidden in the weights themselves.

I have made this point in print before and I will keep making it: the displayed odds are honest about payout, not about probability. The certified RNG ensures the engine respects the probability table; the probability table is the supplier’s commercial product. If you treat the displayed price as a fair-probability estimate, you are reading the wrong column.

The certificate also covers the seasonality of the seed. Most regulated virtual racing engines use the operator’s underlying secure entropy source for seeds, refreshing per round, with sufficient entropy bits that the next output cannot be reconstructed from observing previous outputs. That is the boring part. The interesting part is that the audit log of seed draws is what regulators subpoena when they want to investigate a complaint. Without it, no one could prove an individual round was generated cleanly.

The Labs: eCOGRA, GLI Europe BV, iTech Labs

There are three names I see most often on the UK side of virtual sports certification. eCOGRA was founded in 2003 by industry stakeholders, has been UKGC-approved as a test house since the licensing regime came in, and is headquartered in London. Its remit covers RNG verification, game-fairness audit, and ongoing surveillance of payout percentages.

Gaming Laboratories International — GLI — is the larger global operator, originally US-founded, with European operations run through GLI Europe BV. Mohio Gaming, the pre-recorded virtual racing supplier I mentioned earlier, runs its certification through GLI Europe BV. iTech Labs is the third major lab, Australia-headquartered but accredited for UK work, and tends to appear on smaller B2B providers’ certificates.

All three labs work to roughly equivalent standards in the UK context — the UKGC’s Remote Technical Standards 7 covers RNG, and the labs run the same tests against it. What differs is the depth and frequency of follow-up. eCOGRA’s name appears on a public seal that some operators display on their site; GLI’s certificates are more often referenced indirectly via supplier press releases.

One trap that catches punters: the seal on the website is not a per-product certificate. It is a corporate-level audit credential. To know that the virtual horse racing product specifically passed testing, you generally need to find the supplier’s technical documentation — which the supplier shares with the operator, not with the public. The opacity is structural. Regulators see the certificates; punters do not.

The regulatory rhythm matters too. RNGs are not certified once and then forgotten. The lab’s relationship is ongoing: re-tests run on a defined cadence, and significant code changes trigger a re-audit before deployment. The frequency depends on the supplier’s release pipeline and the regulator’s inspection profile. For an established product like Inspired’s Flat racing or Coral’s Playtech-driven virtual book, re-audits land roughly annually, with patch-level reviews more often.

What a UK Punter Can Actually Check

This is the section where I usually disappoint people. The honest list of things a UK punter can verify, from outside the operator and supplier ecosystem, is short.

They can verify that the operator displays a current UKGC licence number — searchable in the public register on the regulator’s website. They can verify that the operator’s terms reference an RNG certification regime, naming eCOGRA, GLI or iTech Labs. They can verify that the operator publishes an RTP figure for the virtual racing product, even if they cannot independently confirm it.

What they cannot do is read the supplier’s certificate, watch a round’s seed get drawn, or test whether the probability weights match the displayed odds. Those checks live inside the regulated chain. The punter’s role in the chain is to pick a licensed operator and trust the regime, which is exactly what the regime was designed for.

I find it useful to frame it this way: certified RNG does not promise you will win. It promises that if you lose, you lost honestly, by the published numbers, and that the loss was recorded in an audit log a regulator can examine. That is a meaningful guarantee. It is also, on its own, not enough to make the product safe to play without a deposit limit and a time-out tool live in the background.

Prepared by the Horse Racing Bet Game editorial staff.