History and rules
Prohibition, licensing, standards, disclosure
Regulation of games of chance developed in recognisable stages, and each stage was a response to a failure the previous stage could not address.
Stage one: prohibitionGames are banned outright, usually on public-order grounds. Play continues unsupervised, and the ban removes any means of checking that equipment is sound or that prizes are paid.
Stage two: exceptionsNarrow exemptions appear for particular formats, often draws raising money for public works or for charitable purposes. The exemption establishes the principle that the activity can be permitted if it is supervised.
Stage three: licensingPermission becomes conditional and revocable. Operators are identified, premises are approved, and the licence itself becomes the enforcement mechanism, because it can be withdrawn.
Stage four: technical standardsAttention moves from who runs a game to whether the game works. Equipment tolerances, shuffling procedures, random number generation and audit trails are specified in writing.
Stage five: independent testingCompliance with those standards is verified before deployment by laboratories separate from both the operator and the manufacturer, which examine the design documents rather than the finished cabinet.
Stage six: disclosureReturn figures, odds and rules become publishable information rather than internal documents, on the reasoning that a player who cannot see the mathematics cannot assess the proposition.
The regulatory history of games of chance is unusually legible, because the same sequence recurs across many jurisdictions and centuries. Each stage exists because the previous one left a specific problem unsolved, and the stages accumulate rather than replace one another.
Why prohibition kept failing
The oldest approach was to ban the activity. Prohibitions on dicing and on card play are recorded repeatedly across the medieval and early modern periods, usually justified by public disorder, by concern about soldiers and apprentices, or by the ruin of households.
They failed consistently, and the interesting part is the mechanism of the failure. A ban does not remove demand; it removes supervision. Under prohibition the games continue with unexaminable equipment, no recourse when a prize is not paid, and no way to distinguish a fairly run game from a rigged one. The activity becomes more harmful precisely in the respects the ban was meant to address, which is the argument that eventually produced licensing.
Why licensing was not sufficient either
Licensing solved the question of who. An identified, approved operator holding a revocable permission has a great deal to lose from misconduct, and the licence is a far more effective instrument than a criminal prohibition that is never enforced.
What licensing could not settle was whether the games themselves behaved as described. A licensed venue can operate a machine whose real return is nothing like its apparent one, and no amount of scrutiny of the operator's character detects it. As soon as outcomes were produced by mechanisms a player could not inspect, and above all once electronic machines with virtual reels replaced purely physical ones, the regulatory question had to move from the operator to the equipment.
Technical standards
The result was a body of written technical requirements. Dice must be within stated dimensional tolerances. Wheels must be maintained, rotated and checked for bias. Random number generators must meet defined statistical properties, must not be predictable from observed output, and must be seeded from an adequate entropy source. Machines must retain logs sufficient to reconstruct what happened. Paytables and their underlying weightings must be documented in a par sheet that can be produced on demand.
These requirements are simply the mathematics on the rest of this desk converted into obligations. The theory assumes a uniform sample space and an honestly stated paytable; the standards are what force those assumptions to hold in the physical world.
Independent testing
Written standards need someone to check them against, and self-certification is a weak instrument when the person certifying benefits from the answer. The answer was independent testing: laboratories, separate from both operator and manufacturer, that examine design documents, source code and par sheets before a game is approved for use.
The distinctive feature of this model is that it inspects the design rather than the deployed device. A finished machine reveals very little from the outside, and observing it in operation for any feasible length of time cannot distinguish a 92 per cent return from a 96 per cent one, for exactly the reasons set out on the law of large numbers page. The only practical way to know what a game does is to read the document that defines it.
Disclosure
The most recent stage treats the mathematics as information a player is entitled to. Odds, rules and return figures move from internal specification to published fact, and advertising rules develop alongside, on the reasoning that a proposition whose terms are misdescribed is defective regardless of whether the equipment is sound.
Disclosure has a real limitation that is worth stating. Publishing a return figure tells a player what the long-run arithmetic is; it does not make that arithmetic favourable, and it does not compress the timescale over which the arithmetic asserts itself. It is an information remedy applied to a subject where the information is genuinely difficult to interpret, which is why the underlying derivations, rather than the summary percentages, are the useful thing to understand.