Group 1 - The core viewpoint of the article is that leading quantitative private equity firms are shifting from a long redemption notice period to optimizing liquidity to enhance customer experience [1][4]. - On March 30, 2023, Pansong Asset announced a significant change in its redemption rules, reducing the notice period from T-5 trading days to T-2 trading days, thereby improving liquidity [3][4]. - Pansong Asset, established on June 29, 2022, has rapidly grown its assets under management (AUM), surpassing 20 billion yuan in July 2023, reaching 50 billion yuan in March 2024, and exceeding 100 billion yuan by July 2024 [3]. Group 2 - The adjustment in redemption notice time is seen as a landmark move for liquidity optimization in the industry, reflecting the confidence of quantitative firms in their strategy capacity and capital stability [4]. - In March 2023, many leading quantitative private equity firms experienced a significant decline in average returns, with losses ranging from 9% to 12% for their strategies [6]. - The private equity market has seen a surge in new quantitative long strategies, with a notable increase in the number of registered products, which doubled year-on-year and month-on-month [6][7].

Core Viewpoint - OpenClaw demonstrates the ability to construct investment strategies and provide signal alerts, showcasing its unique advantages in market tracking and task execution [1][4][36]. Group 1: OpenClaw Usage Scenarios - OpenClaw can autonomously deploy a quantitative environment, extract data, and construct various quantitative strategies, highlighting its operational efficiency [1]. - The platform is integrated with Tushare's database via API, allowing for seamless data extraction and strategy development [1][6]. Group 2: Investment Strategy Construction - The initial phase involves data extraction, where complete index data is stored locally for subsequent calculations [3]. - OpenClaw successfully completes the full process from data extraction to strategy construction and analysis, showing good reasoning capabilities in strategy design [4]. - The platform can monitor constructed strategies daily and provide alerts on whether to open positions [4]. Group 3: Initial Investment Calculations - OpenClaw conducts initial investment calculations by comparing one-time investments, monthly investments, and basic intelligent investment strategies [7]. - The importance of aligning time dimensions for accurate comparisons between different investment strategies is emphasized [7][10]. Group 4: Internal Rate of Return (IRR) Analysis - OpenClaw initially misinterprets the comparison of returns between investment strategies, favoring simple return rates over internal rates of return (IRR) [10]. - After guidance, OpenClaw successfully incorporates IRR as a core metric for evaluating investment strategies, demonstrating its learning capabilities [10][14]. Group 5: Complex Strategy Development - OpenClaw autonomously constructs six complex intelligent investment strategies using various technical indicators, showcasing its technical application abilities [15]. - The strategies include multi-tiered investment mechanisms based on different market conditions, enhancing flexibility and adaptability [15][17]. Group 6: Performance Comparison of Strategies - The performance of different strategies is compared based on their IRR, with the baseline intelligent investment strategy showing the highest IRR at 6.49% [22]. - The RSI oversold strategy emerges as a strong performer, particularly in risk control and drawdown management [22][34]. Group 7: Risk Assessment of Strategies - OpenClaw develops risk indicators tailored to investment strategies, assessing metrics such as the proportion of time below cost price and maximum drawdown [28][30]. - The 60-day moving average strategy is identified as the most balanced in terms of performance, despite its lower frequency of trades [31]. Group 8: Future Considerations - OpenClaw's initial user experience is hindered by a lack of foundational market knowledge and issues with model hallucinations, which can lead to errors [36][37]. - Continuous improvement in OpenClaw's learning capabilities and error correction mechanisms is essential for enhancing its efficiency in investment research [36][37].

1. Report Industry Investment Rating No relevant content provided. 2. Core Viewpoints of the Report - Last week, the bond futures showed a differentiated trend, with the 30 - year main contract rising by 0.45%, the ten - year main contract falling by 0.01%, the five - year main contract rising by 0.01%, and the two - year main contract falling by 0.01%. The market risk preference gradually weakened, activating the hedging attribute of bond futures. The downward trend of bond futures is not likely to reverse when the long - term bullish logic of the stock market remains unchanged and the coupon income of bonds is not very attractive. Only when the expected return of equity or risk assets declines marginally, the short - term hedging trading attribute of the bond market becomes more obvious [6]. - Last week, domestic commodities remained highly volatile. Energy and chemical products showed a differentiated trend, with significant declines in crude oil and fuel oil, but sharp increases in some chemical products. The market's expectations regarding the situation in the Iranian region are still the focus of divergence and game. Most commodity factors declined, with significant drops in term - structure factors and slight declines in volatility, basis, and warehouse - receipt factors. Long - term trend and term - structure factors remain effective, and the gradual recovery of fundamental factors is worthy of future attention [13][15]. 3. Summary by Directory 3.1 Treasury Bond Futures Quantitative Strategy 3.1.1 Treasury Bond Futures Market Review - Last week, each bond futures contract showed a differentiated trend. The 30 - year main contract rose by 0.45%, the ten - year main contract fell by 0.01%, the five - year main contract rose by 0.01%, and the two - year main contract fell by 0.01%. The basis of each variety also showed a differentiated trend. The CTD bond of the ten - year bond was 250025, with a basis of about 0.06 yuan on the 27th, in line with the historical average. The CTD bond of the 30 - year bond was 210014, with a basis of 0.34 yuan on the 20th, higher than the historical average [6]. - The market risk preference gradually weakened, activating the hedging attribute of bond futures. The downward trend of bond futures is not likely to reverse when the long - term bullish logic of the stock market remains unchanged and the coupon income of bonds is not very attractive. Only when the expected return of equity or risk assets declines marginally, the short - term hedging trading attribute of the bond market becomes more obvious [6]. - For the ten - year treasury bond, in the out - of - sample period (from January 1, 2021, to the present), the annualized return, Sharpe ratio, and maximum drawdown of the portfolio under single - leverage are 2.71%, 1.27, and 2.04% respectively. Since the release of the report (from November 1, 2025, to the present), the annualized return, Sharpe ratio, and maximum drawdown of the portfolio under single - leverage are 2.62%, 1.61, and 0.67% respectively [6]. 3.1.2 Unilateral Strategy Performance - The unilateral strategy's performance is shown in the table. The annualized return, annualized volatility, annualized Sharpe ratio, maximum drawdown, and Calmar ratio in the full - sample period are 2.71%, 2.13%, 1.27, 2.04%, and 1.32 respectively. Since the report was released, these indicators are 2.62%, 1.63%, 1.61, 0.67%, and 3.88 respectively [9]. - The large - category factors include basis, intraday technical, intraday volume - price, high - frequency capital flow, member positions, and risk assets. The signal is generated by equal - weighting within each large - category factor and then taking the average, with the sign of the average as the long - short signal. The back - test details are that the strategy uses the VWAP of the first ten minutes of the next - day's opening as the trading price and buys with single - leverage [11]. 3.2 Commodity CTA Factor and Strategy Performance 3.2.1 Commodity Factor Performance - Last week, domestic commodities remained highly volatile. Energy and chemical products showed a differentiated trend, with significant declines in crude oil and fuel oil, but sharp increases in some chemical products. The market's expectations regarding the situation in the Iranian region are still the focus of divergence and game. Most commodity factors declined, with significant drops in term - structure factors and slight declines in volatility, basis, and warehouse - receipt factors. Long - term trend and term - structure factors remain effective, and the gradual recovery of fundamental factors is worthy of future attention [13][15]. 3.2.2 Tracking Strategy Performance - The performance of each tracking strategy is as follows: - CWFT strategy: Annualized return of 9.4%, Sharpe ratio of 1.62, Calmar ratio of 1.07, maximum drawdown of - 8.81%, recent one - week return of - 0.22%, and year - to - date return of 2.84% [14]. - C_frontnext & Short Trend strategy: Annualized return of 11.2%, Sharpe ratio of 1.70, Calmar ratio of 1.66, maximum drawdown of - 6.72%, recent one - week return of - 1.19%, and year - to - date return of 1.42% [14]. - Long CWFT & Short CWFT strategy: Annualized return of 12.6%, Sharpe ratio of 1.41, Calmar ratio of 0.96, maximum drawdown of - 13.07%, recent one - week return of - 0.84%, and year - to - date return of 5.56% [14]. - CS XGBoost strategy: Annualized return of 4.8%, Sharpe ratio of 0.78, Calmar ratio of 0.22, maximum drawdown of - 21.64%, recent one - week return of - 0.44%, and year - to - date return of - 5.86% [14]. - RuleBased TS Sharp - combine strategy: Annualized return of 11.4%, Sharpe ratio of 1.49, Calmar ratio of 1.38, maximum drawdown of - 8.26%, recent one - week return of - 0.75%, and year - to - date return of - 0.05% [14]. - RuleBased TS XGB - combine strategy: Annualized return of 10.8%, Sharpe ratio of 1.88, Calmar ratio of 2.18, maximum drawdown of - 4.95%, recent one - week return of - 1.08%, and year - to - date return of - 3.30% [14]. - CS strategies, EW combine strategy: Annualized return of 12.6%, Sharpe ratio of 1.80, Calmar ratio of 1.71, maximum drawdown of - 7.38%, recent one - week return of - 0.84%, and year - to - date return of 2.86% [14]. - Among the above six strategies, the CWFT strategy performed the best last week with a return of - 0.22%, and the Long CWFT & Short CWFT strategy performed the best year - to - date with a return of 5.56%. The equal - weighted composite strategy of the above cross - sectional strategies (equal - weighted weekly returns) has an annualized return of 12.6%, a Sharpe ratio of 1.80, a Calmar ratio of 1.71, a maximum drawdown of - 7.38%, a recent one - week return of - 0.84%, and a year - to - date return of 2.86% [34].

Investment Rating - The investment rating for the company is "Buy-A/Buy-H" with a current price of CNY 15.31 and HKD 13.20, and a fair value of CNY 19.58 and HKD 16.88 [3]. Core Insights - The company has shown steady growth in profitability, with a reported revenue of CNY 24.972 billion for 2025, representing a year-on-year increase of 19.53%, and a net profit attributable to shareholders of CNY 12.350 billion, up 18.91% [8][13]. - The company is focusing on wealth management transformation and enhancing its institutional business, with significant growth in brokerage and asset management services [8][23]. - The company is expected to maintain robust earnings, with projected net profits of CNY 14.109 billion in 2026 and CNY 16.409 billion in 2027, supported by a stable capital market environment [62]. Summary by Sections 1. Steady Profit Growth - Revenue for 2025 reached CNY 24.972 billion, a 19.53% increase year-on-year, while net profit attributable to shareholders was CNY 12.350 billion, up 18.91% [13]. - The company's leverage ratio slightly decreased to 4.34, with a return on equity (ROE) of 9.94% [14]. 2. Balanced Business Recovery and Wealth Management Transformation - The brokerage business saw a revenue increase of 45% in 2025, with a slight decline in market share to 1.85% [24]. - The company expanded its client base significantly, with a 55.41% increase in private equity trading assets [28]. - The asset management business generated stable income, with net revenue of CNY 8.73 billion, a 21.8% increase [40]. 3. Investment Business and Underwriting Recovery - The investment business reported a net income of CNY 85.79 billion, a 2.5% increase year-on-year [52]. - The underwriting business generated CNY 10.28 billion in revenue, a 20% increase, with significant growth in IPO underwriting [50]. 4. Earnings Forecast and Investment Recommendations - The company is projected to achieve net profits of CNY 14.109 billion in 2026 and CNY 16.409 billion in 2027, with a fair value estimate of CNY 19.58 per share [62].

Group 1: Report Industry Investment Rating - No relevant content provided Group 2: Core Viewpoints of the Report - OpenClaw can reproduce research reports and conduct regular monitoring, which is a typical scenario suitable for OpenClaw in investment research work [1] - OpenClaw has strong reading and summarizing abilities for research reports, but the process of reproducing models requires repeated error correction [1] - The follow - up tracking and regular prompting of the model can better demonstrate the advantages of OpenClaw, such as timing tasks and automation of complex chain work [1] - Currently, OpenClaw still has a certain gap compared with specialized code platforms in report reproduction, but it has unique interaction modes and advantages in monitoring and regular prompting [1] Group 3: Summary According to the Directory 1. OpenClaw's Reproduction of Shenwan Hongyuan's Technical Pattern Research Report 1.1 OpenClaw's Reading and Summarizing of the Research Report - By sending the PDF version of the research report to the OpenClaw dialogue window through Feishu, OpenClaw can read and summarize the report. Its summary of the report is complete and accurate, and the sorting process is consistent with the original report [4][7] 1.2 Repeated Error Correction in the Reproduction Process - OpenClaw has some uncertainties in understanding the details of the logic. After confirming the details, the written code still has errors. By asking OpenClaw to describe the code logic in words, possible errors can be quickly found [12][15] 2. Results of the Research Report Reproduction 2.1 Statistical Results of the Identified Divergence Pattern Returns - After repeated error correction, the complete model of the research report is reproduced. After the divergence occurs, the subsequent performance of stocks with top divergence shows an obvious decline, while stocks with bottom divergence show an obvious increase. The winning rate of 10 trading days can reach about 70%. After confirmation, the returns change significantly, and the divergence occurrence is more meaningful for investment [34][36] 2.2 Distribution Statistics of the Identified Divergence Pattern Returns - The return distribution of top - divergence stocks shows an obvious long - tail effect, and there is a possibility of missing big - bull stocks. The return distribution of bottom - divergence stocks also has a long - tail effect, but the "tail up" effect is not obvious, and there is a possibility of continued decline [39][46] 3. Regular Monitoring of the Divergence Pattern - After completing the calculation, a regular monitoring system for stock divergence patterns is set up through OpenClaw, which can automatically extract market data, identify patterns, and prompt divergence signals [49][60] 4. Thoughts and Summaries - OpenClaw shows good reading and summarizing abilities for research reports, but there are many small problems in model reproduction. It is expected that its usability will be improved in the future. Although it has a gap with specialized code platforms in report reproduction, it has unique interaction modes and advantages in monitoring and regular prompting [61][62]

Report Industry Investment Rating - Not provided in the content Core View of the Report - OpenClaw can reproduce research reports and conduct regular monitoring. It shows strong ability in reading and summarizing research reports but has issues in model reproduction that require repeated error correction. The subsequent tracking and regular prompting of the model can better demonstrate OpenClaw's advantages. Although OpenClaw currently lags behind specialized code platforms in research report reproduction, it has unique interaction modes and advantages in monitoring and regular prompting [3]. Summary According to the Table of Contents 1. OpenClaw's Reproduction of Shenwan Hongyuan's Technical Pattern Research Report - **1.1 OpenClaw's Reading and Summarization of the Research Report**: By sending the PDF version of the research report to the OpenClaw dialogue window via Feishu, OpenClaw can read and summarize it. Its summary of the research report is complete and accurate, with a consistent recognition process and correct calculation and understanding of thresholds [7][10]. - **1.2 Repeated Error Correction in the Reproduction Process**: In the process of code writing and model reproduction, OpenClaw encountered difficulties. There were uncertainties in understanding logical details, and errors occurred in the written code. To save time, OpenClaw was asked to describe the code logic in words to quickly check for errors. After a long - term error - correction process, the final code logic was consistent with the original text [15][18][34]. 2. Reproduction Results of the Research Report - **2.1 Statistical Results of the Identified Divergence Pattern Returns**: After repeated error correction, the research report model was reproduced. Market data of CSI 300 constituent stocks from 2016 to March 23, 2026, were extracted. After divergence occurred, stocks with a top divergence tended to decline, and stocks with a bottom divergence tended to rise, with a win rate of around 70% in 10 trading days. However, when only considering the performance after confirmation, the returns changed significantly, and the divergence confirmation had little investment value [35][37][38]. - **2.2 Distribution Statistics of the Identified Divergence Pattern Returns**: The return distributions of both top and bottom divergence showed a long - tail effect. For top divergence, there was a possibility of missing big - bull stocks. For bottom divergence, stocks had a high probability of rising, but some stocks continued to fall [40][47]. 3. Regular Monitoring of Divergence Patterns - After completing the calculations, a regular monitoring system for stock divergence patterns was set up through OpenClaw. OpenClaw was asked to regularly extract market data, identify patterns, and prompt for divergence signals, demonstrating its unique features compared to other code platforms [50]. 4. Thoughts and Summary - OpenClaw shows good ability in reading and summarizing research reports but has many minor problems in model reproduction. It currently lags behind specialized code platforms in research report reproduction. However, it has unique interaction modes and advantages in monitoring and regular prompting [63][65].

Core Viewpoint - The recent decline in small-cap stocks has been significant, with the Wind Micro Cap Index dropping 7.12% in one week, and a total pullback of 14.6% over nine trading days from March 11 to March 23, indicating a rapid contraction in market risk appetite [2] Group 1: Market Performance Comparison - The current round of small-cap stock declines is less severe compared to the February 2024 downturn, with the Wind Micro Cap Index and the CSI 2000 Index experiencing weekly declines of approximately one-third of those seen in February 2024 [2][3] - In February 2024, the Wind Micro Cap Index fell by 21.69%, while in March 2026, it only fell by 7.12%, showcasing a significant improvement in market stability [3] Group 2: Quantitative Product Performance - Quantitative investment products have shown relative stability during the recent downturn, with drawdowns significantly lower than those experienced in February 2024, where mainstream index products like CSI 1000 and CSI 500 saw declines around -5% [6] - The performance of various quantitative strategies has improved, with the small-cap index strategy experiencing a drawdown of -5.90% in March 2026 compared to -11.48% in February 2024 [7] Group 3: Excess Return Distribution - The distribution of excess returns has improved significantly, with the overall excess returns for various index products converging around zero in March 2026, indicating a more balanced and stable performance compared to the extreme negative excess returns seen in February 2024 [10] - In March 2026, the small-cap index strategy achieved a positive excess return of 1.23%, contrasting with a negative excess return of 10.21% in February 2024, reflecting a notable improvement in strategy effectiveness [11] Group 4: Risk Management Enhancements - The recent downturn has prompted stricter risk management measures among quantitative private equity managers, including limiting small-cap stock positions, enhancing liquidity controls, and optimizing quantitative models to better predict and respond to extreme market conditions [13] - These upgrades in risk management practices have contributed to the resilience of quantitative strategies during market fluctuations, reducing the likelihood of extreme drawdowns [13]

Core Viewpoint - The A-share market is experiencing significant downward pressure due to a combination of internal and external factors, leading to a sharp decline in major indices and raising concerns about the sustainability of the current bull market [2][3][4]. External Factors - The U.S. Federal Reserve's hawkish signals and geopolitical tensions have resulted in foreign capital outflows, putting pressure on growth stock valuations [3][4]. - The ongoing tensions in the Strait of Hormuz and rising inflation expectations are contributing to global market volatility, with potential implications for A-shares [3][4]. Internal Factors - The tightening liquidity at the end of the quarter has led to increased selling pressure from institutional investors, particularly in high-valuation technology growth sectors [4][5]. - The market's initial decline triggered stop-loss selling from quantitative funds, exacerbating the downward momentum and resulting in a significant drop in trading volumes [5]. Market Sentiment - The market is currently characterized by heightened fear, with a large number of stocks declining and a significant number hitting their daily limit down [2][3]. - The capital market's self-reinforcing mechanism is leading to a "panic selling" scenario, where investors who were previously bullish are now forced to reduce their positions due to short-term losses [5]. Diverging Opinions on Market Outlook - Some international investment banks, like Goldman Sachs, express caution, warning that the current asset pricing does not adequately account for the negative impact of high energy costs on global economic growth [6]. - Conversely, some market analysts believe the current downturn is merely a "pause" in a longer bull market, supported by strong underlying fundamentals such as policy support and capital inflows into undervalued Chinese assets [6][7]. Investment Strategies Post-Correction - After the adjustment, key investment directions are identified: resource stocks benefiting from geopolitical premiums, AI infrastructure driven by policy support, and renewable energy aligned with national energy transition goals [7].

1. Report Industry Investment Rating - No relevant content provided 2. Core Views of the Report - In the Treasury bond futures market, last week, each maturity of bond futures showed differentiation. The 30 - year main contract fell 0.35%, while the 10 - year, 5 - year, and 2 - year main contracts rose 0.03%, 0.02%, and 0.05% respectively. The market risk preference weakened, activating the hedging attribute of bond futures. The downward trend of Treasury bond futures is not easy to reverse when the long - term bull market logic of the stock market remains unchanged and the coupon income of Treasury bonds is not very attractive. Only when the expected return of equity or risk assets declines marginally, the short - term hedging trading attribute of the bond market is more obvious [5]. - In the commodity CTA market, due to the expected non - short - term end of the Iranian regional conflict, the prices of upstream and downstream varieties in the energy and chemical industry chain continued to rise last week. The sudden increase in inflation expectations reduced the market's expectation of the Fed's interest rate cut this year, and precious metals were among the varieties with relatively large declines. Going long on volatility still has a certain winning rate, and in the commodity bull market, going long on volatility can be an optimal strategy after unexpected events. The returns of spot basis - related factors and trading volume and position ranking factors have recovered, and the warehouse receipt factors have a slight increase. However, the long - term returns of fundamental - logic - driven factors are generally mediocre, and continuous observation is needed [11][13]. 3. Summary by Relevant Catalogs 3.1 Treasury Bond Futures Quantitative Strategy 3.1.1 Market Review - Last week, each maturity of bond futures showed differentiation. The 30 - year main contract fell 0.35%, the 10 - year main contract rose 0.03%, the 5 - year main contract rose 0.02%, and the 2 - year main contract rose 0.05%. The basis of each variety also showed differentiation. The CTD bond of the 10 - year bond was 250025, and the basis on the 20th was about - 0.02 yuan, lower than the historical average; the CTD bond of the 30 - year bond was 210014, and the basis on the 20th was 0.19 yuan, also lower than the historical average [5]. 3.1.2 Quantitative Strategy Performance - For the 10 - year Treasury bond, from 2021/01/01 to the present, the annualized return, Sharpe ratio, and maximum drawdown of the portfolio under single - leverage are 2.71%, 1.27, and 2.04% respectively. Since the release of the report (2025/11/01 to the present), the annualized return, Sharpe ratio, and maximum drawdown of the portfolio under single - leverage are 2.62%, 1.61, and 0.67% respectively [5]. - The unilateral strategy is constructed based on factors such as basis, intraday technical indicators, intraday volume - price, high - frequency capital flow, member positions, and risk assets. The signals are generated by equal - weighting within each factor category and then averaging, with the sign of the average as the long - short signal. The strategy uses the VWAP of the first ten minutes of the next - day's opening as the trading price and buys with single - leverage [9]. 3.2 Commodity CTA Factor and Strategy Performance 3.2.1 Commodity Factor Performance - Due to the expected non - short - term end of the Iranian regional conflict, the prices of upstream and downstream varieties in the energy and chemical industry chain continued to rise last week. The sudden increase in inflation expectations reduced the market's expectation of the Fed's interest rate cut this year, and precious metals were among the varieties with relatively large declines. Going long on volatility still has a certain winning rate, and in the commodity bull market, going long on volatility can be an optimal strategy after unexpected events. The returns of spot basis - related factors and trading volume and position ranking factors have recovered, with the former having an average increase of over 1%, and the warehouse receipt factors have a slight increase. However, the long - term returns of fundamental - logic - driven factors are generally mediocre, and continuous observation is needed [11][13]. 3.2.2 Tracking Strategy Performance - CWFT strategy: Annualized return is 9.5%, Sharpe ratio is 1.63, Calmar ratio is 1.07, maximum drawdown is - 8.81%, recent one - week return is 0.28%, and year - to - date return is 3.06% [12]. - C_frontnext & Short Trend strategy: Annualized return is 11.4%, Sharpe ratio is 1.74, Calmar ratio is 1.70, maximum drawdown is - 6.72%, recent one - week return is - 1.17%, and year - to - date return is 2.64% [12]. - Long CWFT & Short CWFT strategy: Annualized return is 12.8%, Sharpe ratio is 1.43, Calmar ratio is 0.98, maximum drawdown is - 13.07%, recent one - week return is - 0.89%, and year - to - date return is 6.45% [12]. - CS XGBoost strategy: Annualized return is 4.9%, Sharpe ratio is 0.79, Calmar ratio is 0.23, maximum drawdown is - 21.40%, recent one - week return is - 0.43%, and year - to - date return is - 5.44% [12]. - RuleBased TS Sharp - combine strategy: Annualized return is 11.6%, Sharpe ratio is 1.51, Calmar ratio is 1.40, maximum drawdown is - 8.26%, recent one - week return is 0.56%, and year - to - date return is 0.70% [12]. - RuleBased TS XGB - combine strategy: Annualized return is 11.0%, Sharpe ratio is 1.92, Calmar ratio is 2.23, maximum drawdown is - 4.95%, recent one - week return is 0.88%, and year - to - date return is - 2.24% [12]. - CS strategies, EW combine strategy: Annualized return is 12.8%, Sharpe ratio is 1.82, Calmar ratio is 1.73, maximum drawdown is - 7.38%, recent one - week return is - 0.92%, and year - to - date return is 3.74% [12]. - Among the above six strategies, the CWFT strategy performed best last week with a return of 0.28%, and the Long CWFT & Short CWFT strategy performed best year - to - date with a return of 6.45%. The equal - weighted composite strategy of the above cross - sectional strategies has an annualized return of 12.8%, a Sharpe ratio of 1.82, a Calmar ratio of 1.73, a maximum drawdown of - 7.38%, a recent one - week return of - 0.92%, and a year - to - date return of 3.74% [32].

Quantitative Models and Construction Methods 1. Model Name: Momentum + Crowdedness Industry Rotation Strategy - **Model Construction Idea**: The strategy combines momentum and crowdedness factors to optimize industry rotation, aiming to enhance returns by avoiding over-crowded industries and focusing on high-momentum sectors [46][47] - **Model Construction Process**: 1. Exclude the most crowded industries from the previous month 2. Rank the remaining industries based on their momentum performance over the past month 3. Select the top N industries with the highest momentum and allocate them equally in the portfolio 4. Benchmark: Equal-weighted industry portfolio 5. Parameter optimization: OpenClaw was used to test multiple parameter combinations, iterating through different crowdedness exclusion ratios and the number of selected industries. The optimal parameters were determined as excluding the top 30% most crowded industries and selecting the top 5 industries by momentum for equal-weight allocation [47][48] - **Model Evaluation**: The model provides a systematic approach to industry rotation but requires manual debugging and validation due to potential logical errors in AI-generated scripts [46] --- Model Backtesting Results 1. Momentum + Crowdedness Industry Rotation Strategy - **Annualized Return**: Not explicitly stated in the report but included in the backtesting results [49] - **Maximum Drawdown**: Not explicitly stated in the report but included in the backtesting results [49] - **Net Value Growth**: Shown in the strategy's net value chart [50] - **Performance Metrics**: Detailed in the strategy performance table [51] --- Quantitative Factors and Construction Methods 1. Factor Name: Crowdedness - **Factor Construction Idea**: Measures the level of investor concentration in specific industries to identify over-crowded sectors and avoid potential risks associated with them [47] - **Factor Construction Process**: 1. Calculate the crowdedness level for each industry based on investor positioning data 2. Rank industries by crowdedness and exclude the top X% most crowded industries from the portfolio [47] - **Factor Evaluation**: Helps mitigate risks associated with over-crowded industries but requires accurate data and robust implementation [46] 2. Factor Name: Momentum - **Factor Construction Idea**: Identifies industries with strong recent performance to capitalize on momentum effects [47] - **Factor Construction Process**: 1. Calculate the momentum score for each industry based on its past performance over a specific period (e.g., one month) 2. Rank industries by momentum and select the top-performing ones for portfolio inclusion [47] - **Factor Evaluation**: A well-established factor in quantitative finance, but its effectiveness depends on market conditions and parameter settings [46] --- Factor Backtesting Results 1. Crowdedness Factor - **Annualized Return**: Not explicitly stated in the report but included in the backtesting results [49] - **Maximum Drawdown**: Not explicitly stated in the report but included in the backtesting results [49] - **Net Value Growth**: Shown in the strategy's net value chart [50] - **Performance Metrics**: Detailed in the strategy performance table [51] 2. Momentum Factor - **Annualized Return**: Not explicitly stated in the report but included in the backtesting results [49] - **Maximum Drawdown**: Not explicitly stated in the report but included in the backtesting results [49] - **Net Value Growth**: Shown in the strategy's net value chart [50] - **Performance Metrics**: Detailed in the strategy performance table [51] --- Advanced Factor Discovery 1. Factor Discovery Agent - **Construction Idea**: Automates the discovery of potential factors by analyzing the latest academic papers and extracting factor ideas [51][52] - **Construction Process**: 1. Automatically fetches the latest papers from arXiv in quantitative finance categories (e.g., q-fin.PM, q-fin.ST, q-fin.CP) 2. Filters papers related to factor discovery 3. Extracts factor ideas and structures them into "Factor Logic—Construction Method—Application in A-shares" dimensions 4. Summarizes factor classification, usage, data requirements, and feasibility into a daily report 5. Pushes the report to users via Feishu, allowing interaction for further details like PDF access and construction methods [52][53][54] - **Evaluation**: The agent accelerates factor discovery but still relies on human validation due to limitations in AI's logical reasoning and code generation capabilities [51] --- Factor Discovery Results 1. Factor Discovery Agent - **Daily Reports**: Includes structured summaries of potential factors, their logic, construction methods, and feasibility [54][56] - **User Interaction**: Allows users to access original papers and detailed factor construction methods through interactive dialogues [55]