衍生品量化策略周报

Quantitative Models and Construction Methods - Model Name: OLS Model Construction Idea: Linear regression-based model to predict commodity returns using historical data[159][158] Construction Process: 1. Data preprocessing: Adjust frequency, handle missing values, remove outliers, and normalize data[155][157] 2. Feature extraction: Use PCA to reduce dimensionality and avoid overfitting[155] 3. Model training: Train OLS model using rolling five-year data (2015-2020) and predict T+1 to T+6 VWAP returns[158] 4. Strategy construction: Build market-neutral and directional long-short strategies based on predictions[158] Evaluation: Directional long-short strategies outperform market-neutral ones in terms of annualized returns and Sharpe ratio[159] Formula: Not explicitly provided - Model Name: XGB Model Construction Idea: Non-linear regression model leveraging gradient boosting for commodity return predictions[163][158] Construction Process: 1. Data preprocessing: Similar to OLS model[155][157] 2. Feature extraction: PCA for dimensionality reduction[155] 3. Model training: Train XGB model using rolling five-year data (2015-2020) and predict T+1 to T+6 VWAP returns[158] 4. Strategy construction: Build market-neutral and directional long-short strategies based on predictions[158] Evaluation: Directional long-short strategies show higher stability and returns compared to market-neutral ones[166] Formula: Not explicitly provided - Model Name: Composite Model Construction Idea: Combine predictions from OLS and XGB models using equal weights[167][158] Construction Process: 1. Data preprocessing: Same as OLS and XGB models[155][157] 2. Feature extraction: PCA for dimensionality reduction[155] 3. Model training: Combine OLS and XGB predictions with equal weights[158] 4. Strategy construction: Build market-neutral and directional long-short strategies based on combined predictions[158] Evaluation: Market-neutral strategies outperform directional ones in 2024, showing higher annualized returns and Sharpe ratio[167] Formula: Not explicitly provided Model Backtesting Results - OLS Model: - Market-neutral strategy: Annualized return 3.08%, Sharpe ratio 0.44[159][170] - Long-short strategy: Annualized return 0.59%, Sharpe ratio 0.07[159][170] - XGB Model: - Market-neutral strategy: Annualized return -7.74%, Sharpe ratio -1.25[166][165] - Long-short strategy: Annualized return -5.36%, Sharpe ratio -0.41[166][165] - Composite Model: - Market-neutral strategy: Annualized return 3.08%, Sharpe ratio 0.44[167][170] - Long-short strategy: Annualized return 0.59%, Sharpe ratio 0.07[167][170] Quantitative Factors and Construction Methods - Factor Name: Carry Factor Construction Idea: Exploit price differences across contract maturities[137][138] Construction Process: 1. Identify futures contracts with significant price spreads[137] 2. Construct portfolios based on long-short positions across maturities[137] Evaluation: Stable performance with annualized return 8.4%, Sharpe ratio 1.72[134][137] - Factor Name: Trend Factor Construction Idea: Capture momentum in commodity prices[137][138] Construction Process: 1. Use historical price data to identify upward or downward trends[137] 2. Construct portfolios based on trend-following signals[137] Evaluation: Moderate performance with annualized return 6.3%, Sharpe ratio 1.05[134][137] - Factor Name: Value Factor Construction Idea: Identify undervalued or overvalued commodities based on historical price metrics[137][138] Construction Process: 1. Compare current prices to historical averages[137] 2. Construct portfolios based on valuation signals[137] Evaluation: Lower performance with annualized return 3.2%, Sharpe ratio 0.58[134][137] Factor Backtesting Results - Carry Factor: Annualized return 8.4%, Sharpe ratio 1.72[134][137] - Trend Factor: Annualized return 6.3%, Sharpe ratio 1.05[134][137] - Value Factor: Annualized return 3.2%, Sharpe ratio 0.58[134][137] Commodity Strategy Tracking Results - CWFT Strategy: Annualized return 10.2%, Sharpe ratio 1.69, Calmar ratio 1.16[152][153] - C_frontnext & Short Trend Strategy: Annualized return 13.0%, Sharpe ratio 1.95, Calmar ratio 1.94[152][153] - Long CWFT & Short CWFT Strategy: Annualized return 14.5%, Sharpe ratio 1.54, Calmar ratio 1.11[152][153] - CS XGBoost Strategy: Annualized return 9.6%, Sharpe ratio 1.63, Calmar ratio 1.31[152][153] - RuleBased TS Sharp-combine Strategy: Annualized return 12.3%, Sharpe ratio 1.54, Calmar ratio 1.49[152][153] - RuleBased TS XGB-combine Strategy: Annualized return 12.8%, Sharpe ratio 2.11, Calmar ratio 2.84[152][153] - CS Strategies EW Combine: Annualized return 15.9%, Sharpe ratio 2.15, Calmar ratio 2.15[152][153]