How random number generators work in online games

A random number generator, commonly shortened to RNG, supplies values used in producing game outcomes. A game also needs rules that turn those values into results and payouts. Understanding that distinction helps explain why random does not mean every prize has the same probability or that an individual will receive a certain return.

This is an educational account of the concepts, not a technical audit. The Gambling Commission's remote technical standard RTS 7 is a Great Britain example. No named game, supplier or operator is certified by this guide.

Separate the generator from the game rules

An RNG supplies a value; the game's mapping and rules determine what that value means. A simple fictional draw could map several possible values to no payout and fewer values to a prize. Equal treatment of the underlying draw values would not make the prize categories equally likely.

That illustration is not an implementation specification. Real products may have more complicated rules and state. To understand the published probabilities, you need the exact game's rules and configuration, not merely a statement that it uses an RNG.

What Great Britain's RTS 7 says

RTS 7A requires random outcomes within its scope to be acceptably random, demonstrated using statistical analysis. The retrieved standard also prohibits adaptive or compensated behaviour in that context. These are requirements in a particular regulatory framework, not evidence that every accessible online game satisfies them.

Do not expand this summary into a claim that every possible prize is equally likely or that all products operate identically. A technical requirement is distinct from proof of compliance by a specific implementation. Other jurisdictions may have their own standards, testing arrangements and product distinctions.

Randomness and return are different properties

RTP concerns an expected relationship between stakes and payouts under the rules. Randomness concerns the generation and behaviour of outcomes. A random process can have unequal payout probabilities and a negative expected net outcome for the customer.

Imagine a hypothetical game that pays nothing for most possible draw values and a larger amount for a smaller set. Random selection does not make that distribution financially favourable. Nor does it promise that a customer's next few results resemble a long-run average. Fairness terminology must not obscure the risk of loss.

Past outcomes do not create a payout entitlement

Under an independent model with unchanged probabilities, a sequence of losses does not make the next payout due. Independence is an assumption that must fit the model; it should not be asserted blindly for every feature or product. The important point is that a recent sequence alone is not a reliable basis for claiming a future win.

An apparent pattern can occur in random data. A short screenshot or a personal session history cannot certify an RNG or establish manipulation. Conversely, the word random in advertising is not sufficient proof that the implementation meets its regulator's requirements.

Ask for relevant evidence, not a decorative badge

Check the regulator responsible for the account and product first. Then read the game's rules and any specific technical information. A badge or supplier logo without an identifiable scope, date and source is not a complete verification record.

If certification is claimed, ask what exactly was tested, which version or configuration was covered and who issued the evidence. This guide has not checked any such claim for a real product. It should never be cited as proof that a website is certified, secure or endorsed by a regulator.

Keep a technical concern distinct from a complaint

If an outcome appears inconsistent with the published rules, save the game identifier, date, transaction reference, relevant rules and available account history. Use the operator's formal complaint procedure and the local independent route where appropriate. Do not try to diagnose an RNG by increasing play or stakes.

Sources were reviewed on 17 October 2026. RTS 3 is cited for making relevant rules available before gambling, and RTS 7 for the specific random-outcome requirements summarised above; both concern Great Britain. Randomness is not protection against financial harm, a prediction method or a guarantee of winnings.