Breeding calculators look simple from the outside: choose two parents and display a child. The harder engineering problem is trust.
If the game changes, how does a player know which data a result came from? If a public dataset has a gap, should the calculator guess? And if automated tests pass, is it fair to say the results were verified in the game?
1. Treat game data as a versioned dependency
The first design decision was to stop treating breeding data as anonymous JSON.
Every generated dataset carries a small manifest:
