backtest

Start from a JSON strategy file, run the backtest tool, interpret results as factual statements + benchmark comparison + hypothesis falsification judgment, and hand findings back to create-strategy when revision is needed.

This skill is the CLI wrapper and result interpretation guide in the hypothesis-to-falsification workflow: the tool returns deterministic results, and the skill explains what those results mean. Never recalculate or estimate the result values.

Detailed guides are split into references/:

• Result interpretation principles and report format: references/result-interpretation.md
• Benchmark comparison method: references/benchmark.md
• Timeframe-specific default periods: references/period-defaults.md
• Official market-rule sources for manual checks: references/market-rules.md

Load skills/references/glossary.md when explaining metric names, indicator names, or strategy terms in Korean.

Data Root and Execution Mode

At the start of work, identify the mode and resolve {data_root} as an absolute path.