证券研究报告 金工 黄金超买风险或得到一定的释放 2025 年 11 月 23 日│中国内地 量化投资周报 期限结构模拟组合创新高，近期持仓做多工业金属板块 商品融合策略近两周上涨 0.97%，今年以来上涨 3.16%。在三个子策略中， 商品期限结构模拟组合近期表现较好，近两周上涨 2.31%，今年以来上涨 7.46%，并于上周五（2025-11-24）创下回测以来的净值新高。期限结构模 拟组合中，近两周收益贡献靠前的品种是玻璃、甲醇、PVC，收益贡献分别 为 0.46%、0.25%、0.16%，近期玻璃跌幅较大，期限结构组合因做空玻璃 而获得较多正收益；近两周收益贡献靠后的品种是橡胶、白糖、乙二醇，收 益贡献分别为-0.08%、-0.09%、-0.17%。期限结构模拟组合最新持仓中， 主要做多工业金属板块，主要做空能源化工板块，和前期持仓相比，工业金 属的多头仓位和能源化工的空头仓位均有所上升。具体品种看，截面仓单模 拟组合在铁矿石、菜油、玉米、沪铜、沪铝上配置了较高比例的多头仓位， 而在橡胶、螺纹钢、PTA、塑料上配置了较高的空头仓位。 黄金周度 RSI 指标回落至 70 以下，超买风险或得到一定的释放 ...

Quantitative Models and Construction Methods Model 1: Commodity Term Structure Model - **Construction Idea**: This model captures the state of commodity contango and backwardation using the roll yield factor, dynamically going long on commodities with high roll yields and short on those with low roll yields[23][24] - **Construction Process**: - Identify the roll yield for each commodity - Rank commodities based on their roll yields - Go long on commodities with the highest roll yields and short on those with the lowest roll yields - **Evaluation**: The model has shown good performance recently, particularly in the industrial metals and agricultural products sectors[23][24] Model 2: Commodity Time Series Momentum Model - **Construction Idea**: This model captures medium to long-term trends in domestic commodities using multiple technical indicators, dynamically going long on assets with upward trends and short on those with downward trends[23][24] - **Construction Process**: - Use technical indicators to identify trends in commodity prices - Rank commodities based on their trend strength - Go long on commodities with the strongest upward trends and short on those with the strongest downward trends - **Evaluation**: The model has underperformed recently, with significant losses in the black and energy chemical sectors[33][35] Model 3: Commodity Cross-Sectional Inventory Model - **Construction Idea**: This model captures changes in the domestic commodity fundamentals using the inventory factor, dynamically going long on assets with decreasing inventories and short on those with increasing inventories[23][24] - **Construction Process**: - Identify inventory levels for each commodity - Rank commodities based on their inventory changes - Go long on commodities with the largest inventory decreases and short on those with the largest inventory increases - **Evaluation**: The model has shown mixed performance, with significant losses in the agricultural products sector[39][41] Model Backtesting Results Commodity Term Structure Model - **Recent Two-Week Return**: 1.69%[26] - **Year-to-Date Return**: 3.09%[28] - **Top Contributors**: Glass (1.27%), PVC (0.32%), Rubber (0.31%)[30] - **Top Detractors**: Sugar (-0.16%), PTA (-0.24%), Methanol (-0.25%)[30] Commodity Time Series Momentum Model - **Recent Two-Week Return**: -1.22%[26] - **Year-to-Date Return**: -3.17%[33] - **Top Contributors**: Soybean Oil (0.26%), LPG (0.16%), Soybean Meal (0.07%)[37] - **Top Detractors**: Rebar (-0.28%), Soda Ash (-0.30%), Cotton (-0.33%)[37] Commodity Cross-Sectional Inventory Model - **Recent Two-Week Return**: -0.56%[26] - **Year-to-Date Return**: 3.42%[39] - **Top Contributors**: Corn (0.54%), Polypropylene (0.27%), Nickel (0.22%)[43] - **Top Detractors**: PVC (-0.26%), Cotton (-0.39%), Soybean Oil (-0.46%)[43]