A seed in a slot game’s random number generator (RNG) is source data used to initialise the generator’s internal state. It is not a hidden winning number assigned to each spin. In a software RNG, the state changes as the system generates values; game logic then maps those values to outcomes under the game’s mathematical rules.
Exact implementations differ and are often proprietary. The practical point for a player is narrower: a conforming regulated system is designed so that previous results and button timing do not reveal the next result.
Seed, state, output and game outcome are different things
These terms are often collapsed into one idea, which makes otherwise simple explanations misleading.
| Term | What it means | What it does not prove |
|---|---|---|
| Seed | Source value or values used to initialise an RNG’s state. | It does not by itself define the paytable, RTP or fairness of the whole game. |
| RNG state | Variables in memory representing the generator’s current condition. State changes as the generator operates and may also be mixed with new entropy. | It is not normally visible to the player. |
| RNG output | A generated value or set of values. | A raw value is not automatically a reel display or a win. |
| Mapping and game logic | Rules that convert generated values into symbols, features and outcomes under the intended distribution. | Good output cannot compensate for biased or incorrect mapping. |
| Game math | The paytable, available outcomes, prize rules and theoretical long-run behaviour of the configured game. | It does not predict the next spin. |
GLI-19 defines seeding as the initialisation of RNG state variables and defines the seed as the source value or values used for that initialisation. That is more precise than saying a slot simply “chooses a seed for every spin.”
A simplified path from seed to displayed result
This is a conceptual path, not a diagram of any named provider’s architecture. A software RNG can be deterministic internally: given identical starting state and identical processing, it follows the same sequence. Security comes from using a suitable algorithm, unpredictable initialisation, protected state and appropriate state updates—not from calling every output “random” without testing it.
The game still needs correct mapping. GLI-19 requires the final outcome output to match the intended distribution and requires independence between draws. UK Gambling Commission RTS 7 similarly requires acceptable randomness, unpredictability and fair implementation in accordance with the published rules and payouts.
Does a slot use a new seed for every spin?
Not necessarily. Seeding, state progression and generating an output are separate operations. GLI-19 says an interactive gaming RNG should not be seeded from a time value alone, games should not share the same initial seed, and the state should be modified periodically with external entropy to limit the duration of a state-compromise attack.
That does not establish one universal schedule. Public standards describe required properties; they do not tell us that every provider reinitialises state for each round, at a fixed interval or in the same way. Unless a provider or test report documents an implementation, the honest answer is that the exact mechanism is private.
Can spin timing or previous results reveal the next outcome?
A compliant system is designed to prevent that. GLI-19 requires that knowledge of numbers chosen in one draw not provide information about future draws. UK RTS 7 says the next value should be computationally infeasible to predict without complete knowledge of the algorithm and seed value, and that seeding or reseeding must not introduce predictability.
That is why changing when you press Spin is not a useful strategy. The request still occurs at a particular time, but a player does not have the protected internal state and cannot turn button timing into a dependable forecast.
Random sequences can still contain streaks
Independence does not mean wins and losses must alternate neatly. Clusters, long losing runs and repeated symbols can appear in random data. A visible streak is not evidence that a game has become “hot” or “cold,” and it does not make the opposite result due next.
The game’s RTP, prize rules and volatility describe its mathematical design over many outcomes. They do not reveal the next generated result.
A reproducible toy seed-and-counter example
The table below hashes the UTF-8 text <demo-seed>|<counter> with SHA-256 and displays the first 16 hexadecimal characters. Anyone using the same text and method can reproduce the values. Changing either input changes the output.
On a small screen, scroll the table sideways to compare both demo-seed columns.
| Counter | demo-seed-A | demo-seed-B |
|---|---|---|
| 0 | 7025ad22f6bbbc0f | 146420e735d09e7b |
| 1 | 5d30ec227a5ef0b8 | 822ac4af9fdc1789 |
| 2 | 37be8a51408e91ff | cce68f23acdac53c |
| 3 | 727cccfba63f239e | 82e3b559640a1c84 |
| 4 | a93de8630a3d4795 | c587a245a7811113 |
SHA-256 is used here as a transparent teaching tool, not as a claim about a live gambling system. We deliberately do not map the values to reels, prizes or wins because doing so would create a fake slot model rather than evidence about a real one.
What public standards actually require
Two useful public benchmarks show why checking only the seed is insufficient.
UK Gambling Commission RTS 7
- RNG and game outcomes must be demonstrably “acceptably random,” including through generally accepted statistical tests.
- Outputs should be unpredictable, streams should not synchronise or reproduce the same cycle, and seeding or reseeding should not introduce predictability.
- Mapping from random inputs to game outcomes should follow the prevailing probabilities, paytables and game rules.
- Adaptive intervention that changes outcome probabilities during play is not permitted, except for disclosed game features operating under their own rules.
GLI-19 v3.0 for interactive gaming systems
- An independent test laboratory reviews core randomness, scaling, shuffling and outcome-selection code for bias, errors, malicious code and undisclosed parameters.
- Statistical analysis checks intended distribution and independence between draws.
- The outcome-determining RNG should be cryptographically strong and resistant to prediction from past values.
- A time value alone is not an adequate seed; games should not share the same initial seed; external entropy should periodically modify state.
These are testing and implementation requirements. They do not say a player will win, that every jurisdiction uses identical rules, or that one certificate covers every future game version and casino configuration.
Provably fair seeds are a separate transparency model
Some crypto-oriented games expose inputs and commitments so a player can verify a completed round. That is different from a conventional online slot whose internal RNG is assessed by a test laboratory and regulator rather than reconstructed by the player.
BGaming’s published workflow is one concrete example. It says the system calculates a result and secret before the bet, sends the player a SHA-256 hash of that combination, accepts a client seed, and reveals the result plus secret after the round so the commitment can be checked. Its slot-specific process then applies the client seed to the prepared result.
This can provide post-round verification of the documented commitment. It does not let the player know the result before play, and it should not be treated as the universal design for all games described as provably fair.
Where the public evidence stops
- Public standards do not identify the exact algorithm, seed material or state-update schedule used by every commercial provider.
- A laboratory standard describes what should be evaluated; it is not itself a certificate for a particular game build.
- A certification or licence is not a strategy for winning. Slots remain negative-expectation games unless a specific offer changes the calculation, and even then outcomes remain uncertain.
- An unlicensed or unverifiable implementation should not inherit trust merely because it uses words such as RNG, seed or provably fair.
Key takeaways
- A seed initialises RNG state; it is not the displayed result.
- Generating a value, updating state, reseeding and mapping a value to an outcome are different operations.
- Fair implementation depends on unpredictability, distribution, independence, correct mapping, code review and testing—not on the seed alone.
- Previous spins and visible streaks do not provide a reliable forecast of the next result in a conforming system.
- Provably fair verification is implementation-specific and should be checked against the provider’s exact published method.
Frequently asked questions
Is a new seed selected for every slot spin?
Not as a universal rule. An implementation may initialise state, advance that state as values are produced, and mix in external entropy under its security design. Public standards require secure properties, not one schedule shared by every provider.
Can knowing a seed help someone predict a slot?
A secure interactive-gaming RNG is designed so prediction remains computationally infeasible even for an attacker who knows the algorithm. A seed alone is not the same as full current state, mapping logic or authorised access. In a provably fair design, verification information is normally revealed so a completed result can be checked, not forecast before play.
Do losing streaks mean the seed is bad?
No conclusion follows from a short streak alone. Independent random outcomes can cluster. Concerns should be based on licensing, published game rules, verifiable test information and actual technical evidence—not a sequence that merely feels unusual.
Does the seed set a slot’s RTP?
No. The seed initialises state. The configured game mathematics and mapping define the theoretical outcome distribution and payouts. Read the RTP guide for why a long-run percentage does not predict a session.
How can a player check whether a slot is trustworthy?
Check the casino’s licence in the relevant regulator’s public register, open the game’s own information screen for its rules and configured RTP, and look for test information tied to the actual provider and game. Treat a generic fairness badge without a verifiable source cautiously. If gambling is causing harm, stop and use the support routes in our responsible gambling guide.
Sources
- UK Gambling Commission — RTS 7: Generation of random outcomes. Last updated 7 May 2024; accessed 15 July 2026.
- Gaming Laboratories International — GLI-19 v3.0: Standards for Interactive Gaming Systems. RNG requirements on PDF pages 26–28; seed/state definitions on pages 112–113; accessed 15 July 2026.
- BGaming — Provably Fair workflow. Workflow and slot example; accessed 15 July 2026.
Found an error or a source that materially changes this explanation? Contact SlotDecoded so we can review it.