RNG and fairness in crash games

Key takeaways

• UK Gambling Commission standard RTS 7 requires outcomes to be acceptably random, unpredictable and free of adaptive behavior. It applies to games offered under a British license.
• Under the UK Gambling Commission’s testing strategy, an approved test house tests a new game’s math, rules, RTP and software implementation, and the report goes to the Commission before release.
• GLI-19, a voluntary lab standard, has statistical tests evaluated at a 99% confidence level and sets a default minimum theoretical payout of 75% for house-banked games unless the regulator specifies otherwise.
• A certified RNG is evaluated by a lab and overseen by a regulator; provably fair is a per-round check the player can run. Neither changes the return set by the game’s math.

What an RNG does in a crash game

An RNG supplies the one thing in a crash round that nobody chooses: the multiplier at which the round ends. The rising curve, the countdown and the cashout buttons are all built around that single value, which exists before the multiplier starts to move. The page on how crash games work shows where that step sits in a round.

Two steps are involved. The RNG produces raw random numbers, and a scaling step converts each one into a game result, here a crash point such as 1.37x or 24.60x.

How a particular game does that conversion is part of its math. One game, bustabit, publishes its formula; the Spaceman Game Help describes the end point of each round as random and predetermined and gives no formula. What testing is meant to establish is that the whole chain, from random number to displayed result, behaves the way the rules say it does.

What “acceptably random” requires

A clear public statement of what a fair RNG has to do comes from the UK Gambling Commission (UKGC). Its technical standard RTS 7 covers the generation of random outcomes, and each requirement maps onto something concrete in a crash round.

Requirement	What it means	What it rules out
Acceptably random	Randomness has to be demonstrable by statistical analysis, not just asserted	Crash points that follow a detectable sequence or cycle
No adaptive behavior	The game may not compensate for what has happened before	A game that “owes” a high round after low ones, or tightens after a large payout
Unpredictable output	Predicting the next value should be computationally infeasible without knowing the algorithm and the seed	Working out the next crash point from past results
Fair implementation	The game has to run the way its published rules describe	Rules that describe one game while the software runs another

The second row is the regulatory answer to two common beliefs: that a round is “due”, and that the game reacts to a player’s bet. Both are covered on the page about crash game myths.

The third row is, in our reading, why predictor apps have nothing to work with. An RNG that meets the standard gives no usable signal about its next output.

How crash game RNG testing works

Where a British license is involved, these requirements are checked by a third party before launch. The UKGC’s testing strategy sends new games and RNGs through an independent test lab, which the Commission calls an approved test house.

1. Submission A new game or RNG goes to a test house the Commission has approved. The Commission publishes the list of approved test houses.
2. Testing The test house examines the math, the rules, the RTP (return to player) and the software implementation.
3. Report The test report is provided to the Commission before the game is released.

What a lab looks at

One published lab standard is GLI-19: Standards for Interactive Gaming Systems, version 3.0 (17 July 2020), from Gaming Laboratories International (GLI). It is a voluntary standard: regulators decide whether to adopt it. Its requirements include the following.

• Source-code review. The lab reviews the source code of the randomness and scaling algorithms.
• Statistical tests. The tests are evaluated at a 99% confidence level. That describes how the tests are evaluated; it is not a win rate or an RTP figure.
• Cryptographic strength. The RNG must be cryptographically strong, with future values computationally infeasible to predict from past ones.
• No adaptive behavior. Outcomes must not be modified by adaptive behavior.
• A payout floor. House-banked games have a default minimum theoretical payout of 75%, unless the regulator specifies otherwise. It is a lower limit, not a description of any particular game.

Certified RNG vs provably fair

A crash game can back up the claim that results are not manipulated in two different ways, and the two models answer different questions.

Question	Certified RNG	Provably fair
Who checks	An independent test lab, with a regulator overseeing	The player, round by round
How	Lab evaluation, such as source-code review and statistical tests	Cryptographic commit-and-reveal: a commitment before the round, the underlying data after it
What it shows	The game and its RNG were evaluated against a standard	A result was not altered after it was committed
What it does not give	A per-round check the player can run	The RTP or house edge, which the check does not establish by itself
What varies	The regulator and the standard it adopts	The implementation, from game to game

Neither model changes the math of the game. The mechanics of hash chains and seeds are explained on the page about provably fair crash games.

The limits of certification, and what you can check

What certification does not do

• It does not change the return. A certified game is tested against its own published math; where a provider publishes an RTP, the house edge follows by arithmetic. Fair and favorable are different things.
• A company-level statement is not a game-level certificate. A provider saying its RNG is certified describes its portfolio. It does not identify a test report for one title.

What a player can check

• The operator’s license. Regulators such as the UKGC publish registers of the businesses they license. The entry that matters is the operator running the game; a provider’s supplier license does not make any particular site licensed.
• The rules inside the game. For games offered under a British license, UKGC standard RTS 3 requires rules to be easily available before a customer commits to gamble. Information on the likelihood of winning, such as RTP or house edge, and on potential prizes must also be available.
• The stated RTP. The figure in the game’s own help text is the one that applies where the game is being offered.
• Whether a per-round check exists. Some crash games offer a provably fair check. Whether a given game does is a question for the information inside that game.

Where Spaceman fits

Pragmatic Play’s March 2022 press release calls Spaceman an RNG title, but neither it nor the Game Help describes the RNG algorithm, the distribution of crash values or a test lab for the game. At company level the provider names the labs that test and certify its RNG, which is a statement about its portfolio; we found no certificate specific to Spaceman. That record, with the provider’s licensing and the provably fair question, is set out on the page Is Spaceman fair?

Sources

• UK Gambling Commission — RTS 7: Generation of random outcomes
• UK Gambling Commission — RTS 3: Rules, game descriptions and the likelihood of winning
• UK Gambling Commission — Testing strategy: procedure for testing
• UK Gambling Commission — Approved Test Houses
• Gaming Laboratories International — GLI-19: Standards for Interactive Gaming Systems, version 3.0 (17 July 2020)
• Pragmatic Play — About Us
• Pragmatic Play — Safer Gambling
• Pragmatic Play — Pragmatic Play blasts off with Spaceman (press release, 24 March 2022)
• Pragmatic Play — Spaceman Game Help, version 1.7 (change log to 10 July 2024), as distributed to licensed operators (cited by name, not linked)
• bustabit — published documentation (cited by name, not linked)

Frequently asked questions

Are crash games rigged?

A crash game offered under a license that applies standards such as the UK Gambling Commission’s RTS 7 has to produce acceptably random outcomes and may not adjust them in response to earlier results. That concerns randomness, not return: where a provider publishes an RTP, the house edge follows by arithmetic, and random does not mean favorable.