2020 年春，李迅雷给唐军派发了一个艰巨的课题：预测疫情拐点。 作为非流行病学专业人员，时任中泰证券金工首席的唐军有些困惑，这与投资有什么直接联系？但又很巧合， 大学期间的 唐军参加数学建模比赛 前 曾拿 2003 年非典练手，传染病传播模型 还有印象 。 唐军根据 封锁前 武汉迁出 到各省的 人数和 之后 各省上报 的确诊 人数，反推 当时 武汉感染人群数量，再根据疫情起始时间反推传播系数， 这一数据处理 上产生的效果远超模型精度本身。 最终经过测算，唐军预测对了感染高峰到来的时间。 有了经验后，美国疫情开始扩散，李迅雷又派发了一个艰巨的课题：美国能不能控住疫情？ 这一次，唐军意识到自己研究的课题，并不是与投资风马牛不相及。唐军知道人员流动性降到平时 30% 才能稳住疫情不扩散，「美国根本降不到，一定会 严重挤兑，即使医疗资源翻一倍也会挤兑。」 随后，唐军提示了做空原油，那年美油期货价格首次跌到负数。 这两次研究成果收录在李迅雷公众号，至今还能搜索到。三年后，唐军担任中泰资管组合投资部首席投资经理 ， 并开始管理公募 FOF 产品 ，李迅雷的指 导和这些与投资并不直接关联的研究，拓宽了他宏观的视野，潜移默 ...

Quantitative Models and Construction Methods 1. Model Name: Basic Risk Parity Model - **Model Construction Idea**: The model ensures that each asset in the portfolio contributes equally to the overall portfolio risk[20][23] - **Model Construction Process**: - Let the return vector of assets at time T be **r** and the weight vector be **w** - Covariance between assets is denoted as **Σ**, and the portfolio's return and volatility are: $$ \sigma(w) = \sqrt{w^T \Sigma w} $$ - Marginal Risk Contribution (MRC) and Risk Contribution (RC) for asset i are: $$ MRC_i = \frac{\partial \sigma(w)}{\partial w_i} = \frac{(\Sigma w)_i}{\sqrt{w^T \Sigma w}} $$ $$ RC_i = w_i \cdot MRC_i = w_i \cdot \frac{(\Sigma w)_i}{\sqrt{w^T \Sigma w}} $$ - Total Risk Contribution (TRC) is: $$ TRC = \sum RC_i = \sqrt{w^T \Sigma w} $$ - Risk parity requires: $$ RC_i = RC_j \; \text{for all} \; i, j $$[23][24][25] - **Model Evaluation**: The model is effective in balancing risk contributions but may lead to conservative portfolios when asset volatilities differ significantly[5][20] 2. Model Name: Risk Budgeting Risk Parity - **Model Construction Idea**: Adjusts the risk budget to allocate higher weights to riskier assets, making the model more flexible for different risk preferences[5][33] - **Model Construction Process**: - Adjust the relative marginal contribution of assets to the benchmark: $$ RC_i : RC = k_i \; \text{for all} \; i $$ - When assets are uncorrelated, the allocation becomes: $$ RC_i = \frac{k_i w_i^2 \sigma_i^2}{\sum k_i} $$ - Risk budget and actual weights are related quadratically: $$ \text{If actual weight is } n \times \text{basic weight, then risk budget is } n^2 $$ - Static and dynamic risk budgeting rules: - Static: Fixed risk budgets for equities, commodities, and gold - Dynamic: Adjust risk budgets based on Sharpe ratios over the past 6 months[37][39][41] - **Model Evaluation**: Provides higher returns but increases risk. Dynamic budgeting improves returns further but has mixed effects on risk metrics[41] 3. Model Name: Macro Risk Parity Model - **Model Construction Idea**: Allocates risk based on shared macroeconomic factors rather than individual asset risks, addressing overlapping risk contributions among assets[10][64] - **Model Construction Process**: - General asset pricing model: $$ r = 1^T \times I \times f_{base} + B \times I \times F + \varepsilon $$ - **I**: Dummy variable matrix indicating asset categories - **f_base**: Benchmark returns for major asset classes - **F**: Factor returns explaining intra-class differences - **B**: Factor exposures (sensitivity of assets to factors) - **ε**: Residual returns not explained by factors[64][66][68] - Systematic and idiosyncratic risk contributions: $$ RCF_i = w_{new,i} \cdot \frac{(\Sigma_f w_{new})_i}{\sqrt{w^T \Sigma w}} $$ $$ RCE_i = w_{new,i} \cdot \frac{(E w_{new})_i}{\sqrt{w^T \Sigma w}} = \frac{w_{new,i}^2}{\sqrt{w^T \Sigma w}} $$[74][75] - **Model Evaluation**: Effective in reducing leverage and addressing overlapping risks but requires precise macro risk modeling[12][115] --- Model Backtest Results 1. Basic Risk Parity Model - **Annualized Return**: 5.03% - **Maximum Drawdown**: -5.10% - **Volatility**: 2.58% - **Sharpe Ratio**: 1.90 - **Monthly Win Rate**: 71.11% - **Monthly Profit-Loss Ratio**: 3.44[28] 2. Risk Budgeting Risk Parity - **Static Risk Budgeting**: - **Annualized Return**: 5.80% - **Maximum Drawdown**: -9.30% - **Volatility**: 5.80% - **Sharpe Ratio**: 0.97 - **Monthly Win Rate**: 58.89% - **Monthly Profit-Loss Ratio**: 2.05 - **Dynamic Risk Budgeting**: - **Annualized Return**: 6.98% - **Maximum Drawdown**: -12.38% - **Volatility**: 6.29% - **Sharpe Ratio**: 1.07 - **Monthly Win Rate**: 63.33% - **Monthly Profit-Loss Ratio**: 2.30[46] 3. Macro Risk Parity Model - **Basic Asset Classes**: - **Annualized Return**: 5.03% - **Maximum Drawdown**: -5.10% - **Volatility**: 2.58% - **Sharpe Ratio**: 1.90 - **Monthly Win Rate**: 71.11% - **Monthly Profit-Loss Ratio**: 3.44 - **Expanded Sub-Asset Classes**: - **Annualized Return**: 7.35% - **Maximum Drawdown**: -11.49% - **Volatility**: 6.63% - **Sharpe Ratio**: 1.07 - **Monthly Win Rate**: 63.33% - **Monthly Profit-Loss Ratio**: 2.29[89] 4. Refined Asset Pool - **Asset Risk Parity**: - **Annualized Return**: 6.63% - **Maximum Drawdown**: -2.84% - **Volatility**: 2.83% - **Sharpe Ratio**: 2.27 - **Monthly Win Rate**: 75.51% - **Monthly Profit-Loss Ratio**: 5.99 - **Macro Risk Parity**: - **Annualized Return**: 8.03% - **Maximum Drawdown**: -3.59% - **Volatility**: 3.79% - **Sharpe Ratio**: 2.04 - **Monthly Win Rate**: 72.45% - **Monthly Profit-Loss Ratio**: 4.32[110]

Core Viewpoint - The report from CITIC Securities indicates a weak performance in A-shares in March, with a divergence in Hong Kong stocks, a decline in US stocks, a rise in gold, and a pullback in the bond market. It suggests that the current economic environment is characterized by a Kondratiev wave downturn, impacting various asset classes [1][2]. Global Macro Outlook - The report predicts that the peak year-on-year GDP for the US will be in Q1 2025, for Japan in Q2 2025, and for the Eurozone also in Q2 2025. It anticipates a temporary improvement in the yen's performance against the dollar and a strengthening of the euro against the dollar in the future. Additionally, it forecasts that gold priced in dollars will continue to strengthen [1][3]. Asset Price and Fundamental Outlook - According to analyst expectations, the forecasted ROE for the entire A-share market and non-financial A-shares in Q1 2025 is 7.38% and 6.42%, respectively, with slight adjustments from the previous month. The intrinsic value estimate for the CSI All A Index in Q2 2025 is projected to be 5,343 points. The report also notes that the ten-year Chinese government bond yield is deviating from historical cyclical patterns [3]. Industry and Style Rotation - The report identifies high economic sentiment in industries such as agriculture, non-ferrous metals, telecommunications, transportation, and non-bank financials. Currently, institutional focus is on non-bank financials and transportation, while interest in light manufacturing, automotive, consumer services, and comprehensive industries has decreased. Recent increases in institutional attention have been noted in the "petroleum and petrochemicals," "non-ferrous metals," "steel," "consumer services," and "real estate" sectors. The machinery sector is approaching a crowded indicator threshold, while the machinery, automotive, and food and beverage sectors are in a sustained crowded state [4].

Core Viewpoint - The article emphasizes the importance of a diversified asset allocation strategy, highlighting the performance of various asset classes, particularly gold and equities, in the current market environment [2][3][4]. Group 1: Asset Performance - COMEX gold ranked second in 2024 with a return of 27.65%, closely trailing the Nasdaq index [2]. - The asset allocation strategy of fund manager Tang Jun has evolved, with gold being a significant holding, peaking at nearly 18% mid-2024, but later decreasing to 13.71% by the end of Q4 2024 [3][4]. Group 2: Investment Philosophy - Tang Jun's investment approach is characterized by a "top-down" strategy that emphasizes low correlation among underlying assets, distinguishing it from traditional FOF managers [8][10]. - He focuses on the concept of "return streams," which involves optimizing risk and enhancing portfolio performance by reducing correlation among assets [10][65]. Group 3: Macro and Micro Analysis - The macroeconomic factors influencing asset allocation include monetary policy and credit expansion, which have become more significant than traditional economic cycles [44][46]. - Tang Jun's framework for asset allocation is structured into three levels: macroeconomic drivers, expectation differences, and low-correlation return streams [106]. Group 4: Tactical Adjustments - Tactical adjustments in asset allocation are made based on market conditions, with a focus on maintaining a disciplined approach to risk management [51][88]. - The strategy includes a dynamic adjustment of asset weights based on macroeconomic indicators and market sentiment, such as the performance of small-cap stocks and the behavior of retail investors [37][39]. Group 5: Future Outlook - The outlook for 2025 suggests that domestic asset allocation will depend on credit expansion, while international considerations will focus on U.S. fiscal policies and their impact on inflation and risk assets [101][103].