#Evidence — Parity

Reference-versus-runtime comparison for hold-and-win-reference.

Two distinct claims live here:

1. Reference/runtime parity — one recorded corpus in libs/mathengine/parity, executed from both the reference engine and the game-engine spin orchestrator. Run with make test-parity.
2. Step-mode equality — persisted step-by-step execution produces the same final state and award as RunHoldAndWinToCompletion for the same seed and initial state. This is what makes the simulation numbers applicable to production play.

A failing vector means the two sides disagree or behaviour moved. It is decided, never re-recorded.

#Contents

Empty. Populated at plan Task 9.