十余年深耕破局！深圳量道投资以AI驱动筑量化壁垒，产品创新拓宽量化边界｜独家专访

Core Viewpoint - Quantitative investment is transitioning from a marginal role to a mainstream strategy in the volatile A-share market, with Shenzhen Liangdao Investment leveraging its strong quantitative technology foundation and systematic strategy framework to advance in the competitive asset management industry [2]. Company Overview - Shenzhen Liangdao Investment was established in April 2015, with a founding team that has been dedicated to quantitative investment and algorithmic trading since 2010. The company has research centers in Xiamen, Beijing, and Shanghai, and its core personnel have over 10 years of experience in quantitative strategy research and risk control [2]. - The name "Liangdao" signifies the "path of quantitative investment," reflecting a deep understanding of the essence of investment—finding the optimal solution between returns and risks [2]. Performance Highlights - Liangdao Investment's quantitative stock and CTA (Commodity Trading Advisor) products have shown outstanding performance this year. As of September 30, 2025, its "Liangdao Shunfeng No. 1 A-class share" and "Liangdao Chengfeng No. 1 A-class share" ranked in the top 10 for excess returns among similar-sized institutions in their respective strategies [3]. - The "Liangdao CTA Selected No. 17 A-class" product achieved a return of ***%, ranking 4th among private equity firms with over 5 billion in assets [3]. Differentiation Strategies - The company attributes its performance to four key differentiating features: 1. Low correlation with the market, achieved through an independent research system and various machine learning models, resulting in an average correlation of 0.40 with excess returns [8]. 2. Utilization of GNN (Graph Neural Network) technology to uncover hidden style rotation opportunities through dynamic relationships between stocks [8]. 3. A unique alternative database that enhances the robustness and stability of price-volume factors in event-driven markets [8]. 4. A specialized risk control system that combines the Barra system with A-share characteristics, such as additional controls for dividends and micro-cap stocks [8]. AI Model Optimization - Liangdao Investment has made significant improvements to its AI models, focusing on input layer optimization and dynamic graph structure construction, which sets it apart from traditional models [10]. - The company employs AI and machine learning algorithms to explore non-linear relationships between factors and returns, enhancing the ability to capture hidden signals from unstructured data [10]. Risk Control Mechanisms - The company's risk control strategy encompasses pre-trade, in-trade, and post-trade measures, ensuring comprehensive oversight throughout the investment process [12]. - Pre-trade controls involve constraints on major style factors and market characteristics, while in-trade controls utilize third-party systems to monitor trading activities [13][14]. - Post-trade controls are implemented through a self-built system that analyzes account holdings and ensures consistency with investment strategies [15]. Market Adaptation Strategies - To adapt to the rapidly changing styles and themes in the A-share market, Liangdao Investment has developed a "multi-dimensional, multi-model" investment system that allows for quick responses to market changes [16]. - The company regularly updates its factor versions and model training to incorporate the latest market characteristics, ensuring its strategies remain relevant [16]. Future AI Strategy - Liangdao Investment's future AI strategy focuses on four main areas: enhancing AI applications in strategy development, strengthening data mining and analysis capabilities, optimizing risk control systems, and cultivating AI talent [21].