# Future Behavior Target Light Audit

Light-only check. Full target stability audit was not started.

- seeds: `[1, 2, 3, 4, 5, 6, 7, 8, 9, 10]`
- bootstrap replicates: `50`
- residue/stride: representative checks only
- features: record_age, above_start_dwell, cumulative_log_drift
- targets: future_net_log_change, future_record_update, future_max_log_drawdown, future_above_start_dwell, future_drop_below_start

Main files:

- `light_observation_target_metrics.csv`
- `light_observation_target_summary.csv`
- `light_target_summary.csv`
- `light_foldwise_model_metrics.csv`
- `light_bootstrap_minimal.csv`

Coarse standard-run priorities are in `light_full_run_recommendations.csv`.

High-priority candidates:

- `future_max_log_drawdown`
- `future_net_log_change`

Medium candidates:

- `future_record_update`
- `future_above_start_dwell`

Low/defer:

- `future_drop_below_start`

Estimated standard run time: about 2-4 hours for the current full script; light findings can narrow targets first.