strategy-generate

Config-driven backtesting requires a config.json file defining instrument codes, time ranges, and entry/exit rules, alongside a pandas-based SignalEngine implementation generating normalized signal series for the built-in backtest evaluation loop.

The workflow normalizes instrument codes and parses time ranges across equities, crypto, and other instruments, applying consistent data and validation rules before invoking the backtest tool for evaluation.

The process parses high-level trading intentions, normalizes instruments and dates, designs signal logic via a five-question checklist, implements a pandas-based SignalEngine, and writes executable definitions to config.json for backtesting.

Refining trading strategies involves examining artifacts and metrics like trade_count and equity from the built-in backtest, iterating on signal_engine.py edits, and re-evaluating until trade_count, equity, and review criteria achieve confidence.

No, the workflow automatically parses requirements, normalizes instrument codes, and infers or requests missing time ranges before writing config.json, ensuring data and validation decisions are enforced prior to backtesting.