backtesting-frameworks
Build reliable trading backtests
Trading backtests often hide bias and overstate performance. This skill provides patterns and checks to design trustworthy backtests that handle look-ahead bias, survivorship bias, and transaction costs properly.
Download the skill ZIP
Upload in Claude
Go to Settings → Capabilities → Skills → Upload skill
Toggle on and start using
Test it
Using "backtesting-frameworks". How do I avoid survivorship bias in equity backtests?
Expected outcome:
- Use point-in-time constituent lists that include delisted securities
- Obtain historical data providers that maintain delisted symbol data
- Document the data source and its survivorship handling approach
- Test your universe against known historical index compositions
Using "backtesting-frameworks". What are the key metrics to evaluate a backtest?
Expected outcome:
- Sharpe ratio for risk-adjusted returns
- Maximum drawdown for worst-case loss
- Calmar ratio combining return and drawdown
- Win rate and profit factor for trading quality
Security Audit
SafeThis is a pure documentation skill containing only instructional content and Python code examples for building trading backtests. All 46 static findings are false positives. The scanner incorrectly flagged: ASCII diagram delimiters (backticks in markdown), dictionary keys (certificate/key files), financial terms like 'sharpe' (weak crypto), and legitimate function calls (dynamic constructor). No executable code, network calls, file access, credential harvesting, or data exfiltration patterns exist.
Risk Factors
🌐 Network access (4)
⚡ Contains scripts (1)
Quality Score
What You Can Build
Validate a new strategy
Apply bias checks and walk-forward splits before trusting performance estimates.
Compare alternatives
Use consistent cost models and metric standards across multiple strategy candidates.
Design backtest engine
Follow event-driven architecture patterns and execution modeling guidance.
Try These Prompts
Outline a basic backtesting workflow that avoids look-ahead bias and includes realistic transaction costs.
Compare event-driven and vectorized backtesting approaches for a daily equity strategy with 50 symbols.
Propose walk-forward train and test windows for 10 years of daily data and explain the rationale.
List Monte Carlo analyses and metrics to assess drawdown risk for a strategy returns series.
Best Practices
- Reserve a final test set that is never used for optimization
- Model commissions and slippage with realistic parameters based on your execution target
- Report drawdowns and risk-adjusted metrics, not only raw returns
Avoid
- Optimizing parameters on the full history without out-of-sample testing
- Ignoring delisted securities when building equity universes
- Assuming zero trading costs for high turnover strategies