Compliance and validation¶
How we know the engines are right, and how far to trust them. No single check is enough, so each layer sees something the others cannot.
flowchart LR
X["XMage"] -->|"verdicts, games"| G["gorge"]
F["Forge"] -->|"games"| G
S["Hand scenarios"] --> G
R["Ratchets, goldens"] --> G
Z["cardfuzz"] --> G
G <-->|"lockstep"| D["dirac"]
Agreement is evidence, not proof. A ruling that gorge is wrong rests on the printed card or the Comprehensive Rules, never on a vote of engines.
-
Do Forge, XMage and gorge play the same game? Does one card do what its printed text says?
-
Does dirac play every decision as gorge does?
-
Ratchets, golden heads, CR coverage, known approximations and cardfuzz.
-
Focused tests while you work, the full suite after merge and in CI.
-
Write a rules engine? Replay gorge's games in it and see where the two disagree.