What provably fair means
A regulated casino's fairness rests on a chain of trust: an accredited laboratory certifies the random number generator and the game's return, the regulator makes compliance with its technical standards and testing requirements a licence condition, and the customer trusts the regulator. A crypto casino serving customers in many countries under one offshore licence needed a substitute, and it found one in a cryptographic scheme that lets the customer verify each result themselves.
The mechanism is a commitment. Before a bet, the operator generates a secret server seed and publishes its hash, a fingerprint from which, for a secure hash algorithm, it is computationally infeasible to recover the seed. The customer supplies or is assigned a client seed. Each bet also carries a nonce, a counter. The result of the bet is computed from the server seed, the client seed and the nonce through a published function, for example a keyed hash such as HMAC with the server seed as the key. When that server seed is retired, on many platforms when the customer rotates to a new seed pair, the operator reveals it, and the customer can check that it matches the hash they were shown before betting and recompute every result they received. If the operator had changed the seed after seeing the customer's bets, the hash would not match.
What this proves is narrow and real. It proves the operator did not alter outcomes after the customer committed, and that results were derived from the published function. It does not prove the function is fair (the customer has to read it, and even the Crypto Gambling Foundation's white paper on the scheme leaves the uniform distribution of outputs out of scope), it does not prove the advertised return is the actual return (that is a property of the function, which the customer can compute but rarely does), and it does not prove anything about games the operator does not run itself. Third-party slots on a crypto casino are as provably fair as they are at any other casino, which is to say they are certified by a laboratory or they are not.
Why it mattered
Provably fair was the answer to the question every unlicensed casino faces: why should a customer believe the games are honest? It gave the crypto segment a credible answer without a regulator, it became a marketing term, and it shaped the product. Games built to be provably fair are simple, because the function has to be readable: a multiplier that rises until it crashes, a dice roll over or under a number, a ball dropped through pegs, a grid of hidden mines. These are the "originals", and they are the crypto casino's defining product.
Regulators have not accepted it as a substitute for certification, and they are right not to. Britain's technical standards, for example, require game results to be "acceptably random", demonstrated to a high degree of confidence through statistical analysis, with independent testing where the Commission's testing strategy requires it. A provably fair scheme with a badly designed function, a disclosed return that the function does not deliver, or a server seed generated from a predictable source is provably fair and unfair at the same time. Certification checks what provably fair does not; provably fair checks what certification does not (that this particular result was not tampered with). The strongest position, which BC.Game, for one, now advertises as provably fair systems alongside an audited RNG, is both.
The originals
Crash. A multiplier starts at 1.00 and rises; the customer cashes out at any moment and wins their stake times the multiplier, or the game "crashes" first and they lose. Round length is seconds, the social element is a shared round with every player's bets and wins visible, and the house edge is set by the distribution of crash points: 1 per cent in BC.Game's published algorithm, 3 per cent in Spribe's Aviator, which publishes a 97 per cent return. Because a crash round is shared, the per-customer seed scheme is adapted: BC.Game pre-generates a chain of ten million SHA-256 hashes and releases them in reverse order, so each round's crash point exists before anyone bets. Crash is one of the segment's signature formats and one that regulated suppliers have taken into mainstream online casinos: Spribe, licensed by Britain's Gambling Commission, says the format became popular in crypto casinos and that it was among the first to bring it to regular online casinos. The Aviator disputes covered in this site's news pages are about that crash game.
Dice. Roll over or under a chosen number; the customer sets the win chance and the payout adjusts. The purest expression of provably fair, with a high published return: BC.Game, for example, publishes a 98 per cent return and a 2 per cent house edge for its Classic Dice, against the 1 per cent edge of its crash game. Its simplicity suits very high-volume play and, lesson five notes, laundering patterns.
Plinko. A ball dropped through a triangle of pegs into buckets with multipliers. Adjustable risk and row count; visual and fast, and in supplier versions such as Spribe's it comes with an autoplay option.
Mines. A grid of hidden tiles, some holding mines; each safe tile revealed raises the multiplier, and the customer stops when they choose. Skill-adjacent in appearance, pure chance in fact.
Others. Limbo (choose a target multiplier), wheel, keno, hilo, towers, and house versions of blackjack, roulette and baccarat. Each platform's originals are its identity, and specialist suppliers such as Spribe now build games in the same formats, Plinko among them, for other operators.
The originals share properties that explain the segment's numbers: published returns of about 97 to 99 per cent, round lengths of seconds, minimum stakes of fractions of a cent (Stake's minimum bet on its originals is $0.0001), and a design that makes it easy to bet again immediately. A customer playing dice at 99 per cent return cycles their bankroll many times over in a session, and faster still with the automatic betting modes many originals offer, which is why wagered volume on crypto casinos is so large and why revenue is a small fraction of it.
Third-party content
The regulated slot catalogue arrived on crypto casinos through the same aggregators that serve regulated markets, and the same suppliers. For a supplier the revenue was attractive and the exposure was real: a studio licensed in Britain or Malta serving an operator that takes customers from Britain or Malta without a licence is a supplier the regulator will ask about. The Sweepstakes course describes the same retreat by suppliers from that segment; in crypto the picture is more mixed, because the largest crypto casinos hold offshore licences of their own (BC.Game's, for example, is issued in Anjouan, in the Comoros), and suppliers differ on how much weight to give them. Live dealer studios face the same question with fewer places to hide, since a live table is streamed from a licensed studio in a named jurisdiction.
The commercial terms differ from regulated markets. Revenue share on crypto casinos is calculated on balances that may be in tokens, settlement can be in stablecoins, and the supplier's own compliance function has to decide what it knows about where the operator's customers are. Suppliers that serve the segment generally rely on contracts, the operator's licence and its geo-blocking representations; whether that reliance is reasonable is the question regulators and courts have started to ask. In January 2025 Britain's Gambling Commission warned that games from licensed suppliers were appearing on unlicensed sites open to British customers, in some instances through third-party resellers, and that licensees who allowed it may be placing their own licence at risk. In September 2026 a Brasília court ordered Spribe to suspend Aviator's integrations with operators not authorised in Brazil.
Return, volatility and the customer
Crypto casinos publish the return or house edge of their originals, and the figures are high: 97 to 99 per cent on the pages cited above. The customer experience is nonetheless one of rapid loss for many players, because the round frequency multiplies the house edge by the number of rounds, and a 1 per cent edge taken sixty times an hour at a constant stake is an expected loss of 60 per cent of that stake every hour, which for a customer staking a tenth of their bankroll each round is 6 per cent of the bankroll an hour. Sixty rounds an hour is slow for games whose rounds last seconds: at one round every five seconds, the minimum Britain allows for casino games other than slots, the same customer's expected loss is 72 per cent of the bankroll an hour. The Casino Game Mathematics course on this site covers the arithmetic. The point for this course is that "99 per cent RTP" is true and does not mean what customers take it to mean, and that responsible gambling controls on originals (session limits, speed limits, loss limits) are the ones a regulator would insist on first. Britain's technical standards already do this for licensed games: at least 2.5 seconds between slot game cycles and 5 seconds for other casino games, a fresh press of the start button for each cycle, no turbo or quick-spin features and no playing several games at once.
What to take from this lesson
Provably fair is a commitment scheme that proves the operator did not alter a result after the customer committed; it does not prove the game is fair or the return is real. It shaped the product toward simple, fast, high-return originals (crash, dice, plinko, mines) that define the segment and explain its volume figures. Third-party content reaches crypto casinos through the same suppliers as regulated markets, who carry the exposure. A 99 per cent return at sixty rounds an hour is a fast way to lose money, and it is the games' speed, not their edge, that a regulator would control first, as Britain's minimum game-cycle times already do.