- base OOS SR 1.47 ret +1272% dd -34%
- conviction OOS SR 1.68 ret +4345% dd -33%
Policy + normalizer saved to models/checkpoints/ (consumed by production/conviction_runner.py).
"Frozen conviction stack (trained < 2016, OOS 2016+)"
2026-09-28
·2026-09-28
·Policy + normalizer saved to models/checkpoints/ (consumed by production/conviction_runner.py).