The Number We Deleted

What we said

The claim came out of the 2020 course recording, where it sat as an elaboration on a broader point about house edge. It was never measured. It was the kind of figure that gets repeated because it sounds about right and nobody has a reason to check it.

It was on our own material, in our founder's own voice, for years.

What happened when we checked

In 2026 we built a simulation engine to derive this system's numbers from first principles rather than inherit them. The first thing it did was disagree with us.

The measurement ran over eighteen million rounds and returned a house edge with a confidence interval attached. The published figure fell outside that interval. Not marginally — outside it, which means the difference is not something the measurement's own uncertainty can account for.

So the claim was deleted. Not corrected quietly, not softened, not left up with a footnote: removed from the registry entirely, so that no surface in this system can render it again. The registry entry that replaced it carries the measured figure and the interval, and it is re-derived every time the software is built.

Why this page exists

A house that publishes numbers has two options when one of them fails. It can fix the number quietly, which is what almost everyone does and which nobody can audit. Or it can publish the failure, which costs something and can be checked.

We think the second is the only version that means anything. A number you can verify is worth more than a number you are asked to trust, and the clearest evidence that a claim is verifiable is watching one get withdrawn.

This is also the honest answer to a question people ask about anyone teaching this subject: how would I know if they were wrong? Here is how. The same machinery that would catch us is the machinery running on every figure we publish, and it runs whether or not anyone is looking.