OCG-Industries Showcase · Cluster Science Featured
Two measurements disagree and the argument that follows is usually about who is more persuasive. Enter the observed count, the count the competing claim expects, the tolerance you are willing to state in advance, and the constraints each side declared, and this tool computes the integer margin between them, expresses it in basis points of the expected value, and routes a verdict against tolerance bands you fixed before looking. Re-running the same numbers reproduces the same verdict and the same hash, so the routing is a function of the stated inputs and not of who is arguing.
The tolerance bands are yours to set, and the point of setting them before you look at the numbers is that you cannot quietly widen the band once you see how far apart the counts are. This page will not do that widening for you. If you move the tolerance after seeing a result, that is a choice you made, not something the tool did.
Data: Integer margin comparison · no external data
This is an integer margin comparison against a tolerance you state. It is not a statistical test. It computes no p-value, no likelihood, and no Bayes factor, it models no uncertainty, and it does not know how the counts were obtained. Its whole claim is that the routing is reproducible from the stated inputs. Set the expected count to 0 to see the undefined case, where the margin cannot be expressed as a share of nothing.