strategy-backtester

Strategy Backtester

Objective

Without a dedicated tool_backtest_* under the current capabilities, complete lightweight backtest orchestration and performance analysis output based on market data and indicator tools.

Input

• User strategy description
• Historical market data
• Technical indicator data

Output (fixed structure)

1. Strategy specification and backtest window
2. Return and risk metrics
3. Trade statistics and parameter sensitivity
4. Constraints and next-step experiments

Evidence table (required)

• List tool_fetch_market_data、tool_calculate_technical_indicators and other calls and quality_status; returns/drawdowns etc. must come from tools or be explicitly declared as MVP derived calculations with formulas attached.

Counterevidence and limitations (required)

• Out-of-sample intervals, survivorship bias, and the impact on conclusions when costs/slippage are not included.

Mandatory rules

• Only through manifest / tool_runner call dependent tools; direct connection to plugins.data_collection。
• If there is insufficient historical data, output insufficient_evidence。
• If there is no dedicated backtesting tool, must declare MVP mode and capability boundaries.
• It is prohibited to output buy/sell points, position ratios, or leverage recommendations.
• Parameter search range is read from config/strategy-backtester_config.yaml .

Dependent tools

• tool_resolve_symbol(L2; instrument resolution)
• tool_fetch_market_data
• tool_calculate_technical_indicators
• tool_l4_valuation_context(L4-data; when the instrument includes individual stocks, optional parallel valuation fact fields)

General output fields

• strategy_spec
• backtest_window
• performance
• risk_metrics
• trade_stats
• parameter_sensitivity
• limitations
• next_experiments