大宗商品中观轮动系列（二）：从信念到模型验证：估值与周期双轮驱动

1. Report Industry Investment Rating - Not provided in the report 2. Core Viewpoints of the Report - The research aim of commodity meso - rotation is to combine the "subjective + quantitative" concept and put it into practice, reducing the specificity of factors, parameters, and models in the strategy, and focusing on interpretability and attributability [3][81] - The report constructs a monthly - frequency variety cluster meso - rotation model. The in - sample average annualized return rate is 17.79%, the Sharpe ratio is 1.44, the drawdown is - 5.70%, and the monthly win - rate is 68.98%. From January to November 2025, the out - of - sample total return is 15.43%, the drawdown is 1.09%, the monthly win - rate is 70%, and only three months record negative returns [3][4][83] 3. Summary by Relevant Catalogs 3.1 Commodity Rotation Mechanism and Variety Cluster Division - In the previous report, it was proposed that during the upward phase of the inventory cycle, the real - side is dominant, manifested as fundamental valuation; during the downward phase, the expected - side is dominant, manifested as macro - valuation. A research framework for the rotation of fundamental and macro - valuation under the cycle phase was put forward, and the meso - targets were implemented at the variety cluster level [6] - 16 variety clusters were selected, including 3 in the black sector, 3 in the non - ferrous sector, 5 in the energy and chemical sector, 3 in the agricultural products sector, and 2 in the precious metals sector, starting from January 1, 2019 [7] 3.2 Bottom - up - Fundamental Perspective 3.2.1 Fundamental Valuation Index Construction - The construction of fundamental valuation indexes in meso - rotation is similar to but different from traditional fundamental quantitative analysis. Due to differences in data among varieties, a special method is needed. First, pre - process the original data of inventory, profit, and inventory - to - consumption ratio, including pre - screening, filling missing values, 3σ standardization, and seasonal adjustment. Then, construct the variety cluster diffusion index. Finally, use the inventory diffusion index as the base diffusion index and design "logic gate" adjustment rules [10][13][14] 3.2.2 Back - testing Results - The strategy is rebalanced monthly. By adjusting the number of long and short variety clusters, it is found that the overall return is strongly correlated with the number of variety clusters. The average annualized return rate of the full - parameter group is 9.88%, the Sharpe ratio is 0.52, the drawdown is - 10.58%, and the monthly win - rate is 57.96%. The ls_4_4 group is selected as the strategy benchmark [22] 3.3 Top - down - Macro Perspective 3.3.1 Variety Clusters Expressing Macro Views - Through principal component analysis of Wind's five major sector indexes, three principal components are obtained. PC1 is the combined effect of growth and interest rate factors, with precious metals having a significant negative exposure and the other four sectors having significant positive exposures; PC2 is the combined effect of inflation structure and monetary policy expectations, with energy and chemical and agricultural products having positive exposures and the other three sectors having negative exposures; PC3 is the influence of RMB exchange - rate depreciation, with black and energy - chemical sectors having negative exposures and precious metals, non - ferrous, and agricultural products having positive exposures [28][29][31] 3.3.2 Macro - valuation Index Construction - Select growth, inflation, interest rate, and exchange - rate as macro indicators and construct monthly - frequency indicators. For growth and inflation factors, select proxy indicators, pre - process, seasonally adjust, filter, and synthesize them; for interest rate and exchange - rate factors, calculate them from high - frequency asset data and then reduce the frequency. The macro - valuation intensity index is constructed by multiplying the factor exposure after rolling regression by the factor momentum, summing them up, and then multiplying by the confidence indicator [40][41][49] 3.3.3 Back - testing Results - The strategy is rebalanced monthly. By adjusting the number of long and short variety clusters, it is found that the overall return decreases as the number of variety clusters increases, and the drawdown and volatility ease. The average annualized return rate of the full - parameter group is 10.13%, the Sharpe ratio is 0.91, the drawdown is - 11.17%, and the monthly win - rate is 66.94%. The ls_4_4 group is selected as the reference group [59] 3.4 Cycle Timing 3.4.1 Inventory Cycle Index Construction - Construct an inventory cycle index based on enterprise accounts receivable and inventory, which is the ratio of the increment of enterprise finished - product inventory to the increment of enterprise revenue. After data selection, cleaning, and calculation, the inventory cycle index is standardized to the [0,1] interval and lagged by one month [62] 3.4.2 Inventory Cycle Inflection Point Identification - First, determine the dynamic threshold; then, identify the initial inflection points; finally, filter the inflection points for the second time. After the second filtering, 5 adjacent inflection points are removed. From March 2012 to the end of 2024, there are 11 effective inflection points, and the average inventory cycle running time is 2 years and 1 month. After subjective adjustment, the average running time is 41 months [65][66][70] 3.5 Variety Cluster Meso - rotation Model - Based on the inventory cycle's up - and - down phases, conduct a binary rotation of the valuation model. In the inventory up - phase, the weight of fundamental valuation is 100%; in the down - phase, the weight of macro - valuation is 100%. The average annualized return rate of the in - sample full - parameter group is 17.79%, the Sharpe ratio is 1.44, the drawdown is - 5.70%, and the monthly win - rate is 68.98%. The out - of - sample total return from January to November 2025 is 15.43%, the drawdown is 1.09%, the monthly win - rate is 70%, and only three months record negative returns [74][83] 3.6 Summary and Outlook 3.6.1 Summary - The report builds a variety cluster fundamental valuation rotation model from a bottom - up perspective and a variety cluster macro - valuation rotation model from a top - down perspective. It also constructs an inventory cycle index and conducts a binary rotation of the valuation model based on the inventory cycle [81][82][83] 3.6.2 Outlook - Adjust the variety cluster division method and include active varieties such as new - energy silicon and lithium - Consider factors such as the currency and hedging attributes of precious metals and the geopolitical attributes of oil products - Construct a variety cluster trend state identification model based on volume - price characteristics and evaluate the trend confidence with valuation levels - Deploy a monitoring system from sentiment analysis and news for mid - cycle strategies to avoid risks [84]