Before writing a line of this website we read the public sites of nine other game providers and catalogued what each one publishes about its games, its markets and its licences. We are not naming them: the exercise was about what a game page owes a reader, and a studio does not need to be named to make that point.
The spread is wider than we expected. The most forthcoming of the nine publishes twelve labelled figures on a game page, including a hit frequency and a max-win probability. Others publish a return and nothing else, and several publish no figures at all and keep everything behind a client login.
What decided how we built ours was smaller and more awkward. On one of those game pages, the prose says the game has fifteen paylines and the specification block three lines below says nineteen. Both were typed into a content management system by a person, and only one of them was corrected. That is an ordinary mistake, and it happens whenever the page and the maths are two separate documents maintained by two different people.
Ours have a single source. Every figure on one of our game pages is read from that game's certification pack, which the measurement run writes itself. No number is copied across at any step, and a check in our build fails when a published figure drifts from the maths that produced it. The cost is that our pages can only say what we have measured, so each one also prints what its pack could not answer.
The other thing we took from the exercise: nobody publishes a confidence interval. A return of 96.58% means very little without the precision it was measured to. Measured to half a percentage point against a half-point tolerance, it is a coin flip presented as a fact. Ours carry both the interval and the method: enumerated exactly, bounded, or sampled.