This package contains parameter search and walk-forward validation logic for single-strategy research.
| Module | Purpose |
|---|---|
optimizer.py |
Optuna-driven parameter optimization for a single strategy |
wfv_optimizer.py |
walk-forward orchestration over repeated in-sample and out-of-sample windows |
fold_generator.py |
rolling/purged fold construction for time-series validation |
objective.py |
composite optimization score and penalties |
validation.py |
guardrails for valid optimization inputs |
This layer is about selecting and validating strategy parameters. It is not the place for:
- CLI parsing
- artifact browsing
- dashboard route logic
- portfolio allocation logic
- optimization runs ultimately evaluate strategies through
BacktestEngine - walk-forward orchestration is triggered from
services/wfo_run_service.py - cache validation and strategy loading happen before entering this package
- any reusable execution-friction estimation belongs to
execution/cost_model.py