How-to guides
Investigate a population that stops improving
Complete a task or diagnose a specific problem.
1. Check whether the target and field changed, whether the best candidate is a near miss or an actual hit, and whether trajectories have collapsed into a narrow family.
2. Review mutation, elite share, and the random reset tail. Elitism preserves successful knowledge; mutation and resets address different exploration scales.
3. Inspect rollout termination and frame budget before blaming selection. Floor settling and repeated bounces can end attempts, and the two-number genome cannot learn mid-flight control.