Backtest Expert

⚠️ **Trading warning / Avertissement trading** : informational only, not investment advice. A strategy that survives backtest stress tests can still lose money live; past simulated performance never guarantees future results. Curated by Skill Harbor: a professional backtesting methodology whose goal is inverted from the usual one: find strategies that "break the least" under pessimistic conditions, not strategies that "profit the most" on paper. The workflow forces discipline at every step: state the edge as a one-sentence hypothesis (if you cannot, stop), codify entry, exit, sizing, filters, and universe with zero discretion, run the initial test over at least 5 and preferably 10+ years across bull, bear, and high and low volatility regimes with realistic commissions and conservative slippage, then spend most of the time stress testing: stop and target parameters varied from 50% to 150% of baseline looking for stable plateaus instead of narrow spikes, entry and exit timing shifted, slippage inflated to 1.5 to 2 times typical estimates, and worst-case fills modeled. A dedicated bias-prevention pass hunts curve-fitting, look-ahead bias, and survivorship bias, the three classic ways backtests lie. An evaluation script (Python 3.9+) scores user-provided metrics; no API keys and no external data are required, you bring your own data and results. From the tradermonty/claude-trading-skills repository (MIT). Honest caveats: this is methodology and evaluation guidance, not a backtesting engine; you still need your own data, code, and platform to generate the numbers it judges. Its pessimism is deliberate and will kill most exciting-looking strategies, which is the point. Skill Harbor never reviews the code, review it yourself before use.