vibe-rules-collection — Windsurf Rules

• Goal: Realistically simulate strategy performance on historical data to estimate future potential, avoiding common pitfalls. • Data: • Use high-quality historical data (OHLCV, tick data if necessary). Account for adjustments (splits, dividends). • Ensure data covers various market regimes (bull, bear, sideways). • Be aware of survivorship bias (using data only from assets that survived). • Simulation Engine: • Process data chronologically, point-in-time. NO lookahead bias (using future information). • Accurately model transaction costs (commissions, fees, slippage). Slippage can be modeled as a fixed amount, percentage, or based on volatility/volume. • Model order execution realistically (e.g., assume market orders fill at next bar's open or based on slippage model; limit orders fill if price touches the limit). • Handle position sizing, margin calls (if applicable), and portfolio value updates correctly. • Metrics & Analysis: • Calculate standard performance metrics: Total Return, CAGR, Sharpe Ratio, Sortino Ratio, Max Drawdown, Calmar Ratio, Win Rate, Profit Factor, Average Win/Loss. • Analyze trade logs: distribution of returns, holding periods, drawdown periods. • Visualize results: Equity curve, drawdown chart, return distribution. • Avoiding Overfitting: • Test on out-of-sample data (data not used during strategy development/parameter tuning). • Perform sensitivity analysis on strategy parameters. • Consider walk-forward optimization instead of optimizing over the entire dataset at once. • Be skeptical of overly complex strategies with many parameters. • Common Pitfalls: Lookahead bias, survivorship bias, ignoring costs/slippage, overfitting, unrealistic execution assumptions.

• Lookahead Bias Prevention: When generating backtesting loops or logic, emphasize processing data sequentially. Add comments warning against using future data within a given time step's logic (e.g., // Ensure no lookahead bias here). • Cost & Slippage Modeling: Include placeholders or parameters in generated backtesting code for commission rates and slippage models. Remind the user to configure these realistically. • Example: trade_cost = commission + slippage fill_price quantity • Realistic Execution: Generate simulation logic that assumes fills based on the next available price after a signal, incorporating slippage, rather than assuming fills at the signal bar's close or ideal price. • Metrics Calculation: Generate code snippets for calculating key performance metrics (Sharpe, Drawdown, etc.) using libraries like quantstats, empyrical, or numpy/pandas. • Overfitting Warnings: When discussing parameter optimization or strategy results, include reminders about the danger of overfitting and the importance of out-of-sample testing or walk-forward analysis. • Data Handling: Remind the user about the importance of clean, adjusted historical data and potential biases (survivorship). • Code Structure: Suggest structuring backtesting code with clear separation between the data feed, strategy logic, execution simulation, and performance analysis components.