How-to guides
Compare search settings reproducibly
Complete a task or diagnose a specific problem.
1. Keep seed, target, gravity, wind, and sizes fixed while varying a single search parameter. Reproducing a generation requires the same environment as well as seeded randomness.
2. Compare hit rate, best minimum distance, solved streak, and ghost paths. Increasing population size also increases rollout work, so compare cost alongside convergence.
3. After moving the target, interpret old genomes as starting material rather than evidence that the new scene is solved. Distinguish animation replay controls from optimizer state.