How the Journal verifies a number: the method behind every figure on this site

What counts as a source

The first row is the one that distinguishes this Journal from most writing about casino games. A return figure on a typical review site is a number the operator supplied or the writer copied. Here it is an arithmetic result: for keno, every one of the forty paytables multiplied cell by cell against the exact hit distribution; for Plinko, every one of twenty-seven boards against the binomial path counts; for roulette, the observation that every bet kind multiplies out to thirty-six. The article shows the working, and the working can be repeated by anyone with the same source.

The verified block

Each article that quotes a game figure carries at least one block marked verified. It names the module, the constant or function, and where useful quotes the comment in the source. It is the audit trail for the paragraph above it, and it is deliberately dry. A reader who wants to check does not need the prose; they need the file name and the symbol.

Two limits are stated in every such block and are worth stating here. First, the client modules are what the browser verifier and the demo engine share; the house service that settles real money is described by those modules’ comments as mirroring them, and the Journal quotes that description rather than asserting it. Second, a verified block is a citation, not a certification. It says where a number came from. It does not say a regulator or an auditor agreed.

When the engine and the copy disagree

Reading source for every article has a side effect: it finds mistakes. The rule when that happens is that the engine is the fact and everything else is corrected to match, and the correction is reported to the team that owns the game rather than quietly edited around.

• Plinko, September 2026. Pricing all twenty-seven boards for the Plinko article showed that one nine-row high-risk table was asymmetric, paying ×1.6 on one edge bucket and ×0.6 on its mirror. The article described the asymmetry neutrally; the games team corrected the table, and the article was amended to say so.
• Bingo, September 2026. The bingo design note counted twelve pay lines in the engine where the game guide said seven. The guide was wrong; it was reported for correction and the article notes the discrepancy.
• The seed panel, September 2026. Writing the hash commitment article surfaced an ordering boundary in how a client seed change and a server seed rotation happen in one request. The article states the boundary plainly rather than omitting it.

A number that cannot be traced to the code that pays it is not a fact about the game. It is a fact about the person who wrote it down.the Journal’s working rule

What we do not claim

• No licence, certification or audit claims. The Journal never describes a game as certified, audited or approved by anyone. “Provably fair” is used in its technical sense, defined here, and never as a seal.
• No popularity or volume claims. No article says a game is popular, most played or trending. There is no play data behind such statements and they would not be verifiable if there were.
• No winning promises. Every article that touches strategy says the same thing: no sizing rule or system changes a negative edge. Where a game rewards correct play, as video poker and blackjack do, the article says how much and stops there.
• No medical advice. The mind pillar is science journalism plus pointers to professional resources, listed in the responsible play toolkit.

The tables the articles compute from are published as JSON and CSV at betkyo.com/data/, each file naming the engine module it was exported from and the article that derives it, under a Creative Commons attribution licence. Reuse them, cite them, or recompute them; the point of the method is that you can.

The same material is on GitHub under the organisation betkyo-open-labs, in three small repositories: odds-data mirrors the tables above, provably-fair-verifier is a dependency-free script that recomputes a round from the revealed server seed with the same byte-to-number construction the site’s panel uses, and odds-derivations holds the scripts that turn the tables into the return figures quoted in the articles. Each script prints the numbers its article prints; if they ever differ, one of them is wrong and we want to know.