In October 2024, a notable live trading test took place at the Alpha Arena AI Trading Competition hosted by the US financial AI lab Nof1. The organizers selected the six most advanced large language models on the market and entrusted them with real capital to manage. The results demonstrated that even the world's most intelligent models cannot guarantee success in "doing investment well"—the investment returns generated by different models varied dramatically. This raises a critical question: is this due to inherent differences in model capabilities or other factors? According to Yang Liu, Head of the Systematic Investment Group in the Quantitative Investment Department at Zhong Ou Fund, this experiment represents an extreme attempt to directly apply large models to final investment decisions, often referred to as an "end-to-end" approach. However, large models are not a panacea; they cannot yet solve all our problems. Nevertheless, there is no doubt that the integration of artificial intelligence and financial investment is deepening. Yang Liu believes that before achieving the ultimate "end-to-end" goal, artificial intelligence has a series of applicable scenarios in investment, which can be divided into three parts: operational tasks, assisting Alpha generation, and Alpha generation (i.e., the generation of final investment decision signals). This is also the direction that the Zhong Ou quantitative team has been exploring in recent years. Judging by the results, they appear to have already unlocked the code for quantitative investing in the AI era. As of the end of 2025, Zhong Ou Quantitative managed 17 products with assets under management exceeding RMB 10 billion (Fund Regular Reports, as of 2025/12/31), primarily focused on index enhancement and active quantitative strategies, with products demonstrating significant outperformance in multiple specialized segments. Even in the widely recognized "red ocean" arena of the CSI 300, the Zhong Ou CSI 300 Index Quantitative Enhanced Fund A achieved 9.13% of excess return over its benchmark for the full year. Extending the timeframe to three years, the Zhong Ou CSI 500 Index Enhanced Fund A delivered 25.04% in excess returns, while the Zhong Ou Small-Cap Growth Fund, benchmarked primarily against the Guozheng 2000 Index, achieved an even higher excess return of 56.06%, both ranking in the top 5% of their respective peer groups. These achievements are the result of the Zhong Ou quantitative team's systematic evolution from version 1.0 to 3.0, and also serve as a microcosm of the successful implementation of Zhong Ou Fund's "industrialized" investment research system. I. System Evolution: The Quantitative Revolution from 1.0 to 3.0 Since launching its first quantitative product—the Zhong Ou Data Mining Mixed Fund—in 2016, Zhong Ou Fund entered the 1.0 stage of "fundamental quantitative" investing. Its core involved translating deep industry research and logic into quantitative models, striving to proactively capture industry inflection points. At that time, the team dedicated 80% of its effort to expert interviews and梳理ing core industry logic, using breadth of research to compensate for depth, akin to "radar scanning" for overlooked opportunities across different sectors. During this phase, Zhong Ou Quantitative already displayed distinctive characteristics. Between 2019 and 2021, when market capital heavily concentrated on the baijiu and new energy sectors, Zhong Ou's quantitative models focused attention on the traditional passenger vehicle segment; they also proactively identified risks in the agriculture, forestry, animal husbandry, and fishery sectors as hog prices approached cycle peaks. In 2024, with the addition of veteran active investment manager Wang Jian, Zhong Ou Quantitative progressed to its 2.0 stage, dedicated to achieving deep integration between active management and quantitative investing. This was not merely a strategy overlay, but a restructuring of organizational architecture and processes. The GARP strategy brought by Wang Jian and its underlying valuation pricing system highly complemented the quantitative stock selection logic, forming a closed loop of mutual empowerment between "logic and data": quantitative models broadened the scope and efficiency for active investing, while the experience from active research infused the models with foresight and logical validation, significantly enhancing their barriers and robustness. This evolution marked a qualitative shift from "instrumental borrowing" to "systemic integration." Subsequently, to address challenges like intensifying industry competition and accelerating factor decay, Zhong Ou Quantitative continued iterating and entered the 3.0 stage—the "Triple Low-Correlation" strategy system, comprising fundamental factors, price-volume factors, and deep learning end-to-end strategies. Among these, fundamental factors benefit from the empowerment of Zhong Ou's active equity investment research and years of alternative data accumulation. Essentially trend factors built around sell-side analyst recommendations and earnings forecasts, the team optimizes them using alternative data and internal/external earnings forecast adjustments to form a consensus expectation, focusing on medium-to-long-term Alpha realization. Price-volume factors are built upon a strong quantitative research team and systematic data processing capabilities. They focus on relatively short cycles,挖掘ing price-volume relationships and market inefficiencies, exhibiting negative correlation with fundamental factors in terms of prediction horizon and Alpha capture. The deep learning end-to-end model predicts stock prices from a higher-dimensional perspective, showing low correlation with both aforementioned factor types. The three complement each other, further expanding the potential for capturing excess returns and carving a new path in the highly homogenized public fund arena. II. Artificial Intelligence: Igniting a New Era of Asset Management Transformation In fact, applications of so-called "end-to-end" models are not uncommon in our daily lives. Take new energy vehicles as an example: vehicles use onboard cameras, GPS signals, and various sensors to collect real-time environmental information, location data, and vehicle status. This data is fed into trained AI models, which then make driving decisions: steering, braking, accelerating, or obstacle avoidance. Although the investment field cannot achieve truly autonomous "end-to-end" investing overnight, the advantages of end-to-end models cannot be overlooked. Yang Liu illustrated that, for instance, they can process vast amounts of non-linear data, including images and time-series data, while adapting more quickly to market changes. More importantly, they can discover higher-dimensional features from data, effectively complementing traditional modeling logic that relies on human intelligence. Of course, one must also confront the challenges posed by end-to-end models. Take the contentious "black box" problem as an example: the specific reasoning paths and logical connections within deep learning models are often difficult to trace and explain clearly. In the financial sector, which emphasizes transparency, attribution analysis, and risk control, this lack of explainability can introduce certain risks. In response, the Zhong Ou quantitative team has strengthened infrastructure development at both the data and modeling levels. Data undergoes extensive processing before being fed to models, and risk management constraints are incorporated during model establishment. Simply put, this involves combining traditional investment experience with advanced technical methods to fully leverage the advantages of the "Triple Low-Correlation" strategy system. In Yang Liu's view, having AI make investment decisions directly is the ultimate goal of quantitative investment evolution, the "stars and oceans." But before reaching that goal, investment research teams need to make numerous incremental improvements, applying AI throughout the entire investment process—including data and logic organization, model construction, and backtesting—while continuously refining details. It is these daily, step-by-step advances, the "grains of sand beneath the soles," that pave the way to the distant stars and oceans. The evolution of Zhong Ou Quantitative from 1.0 to 3.0 is not merely an upgrade of strategies and factors, but a vivid demonstration of its "digital-intelligent" investment research system. On one hand, through the standardized processing and long-term accumulation of traditional financial data, the team transforms non-standard information into model-able, reusable digital assets. On the other hand, leveraging deep learning technology, the team achieves intelligent identification and signal generation for high-dimensional, non-linear market features, enhancing model adaptability while making Alpha output capabilities more controllable and stable. Ultimately, through the empowerment of technology and data, it improves the decision-making efficiency and quality of quantitative investment. III. Industry Trends: The New Quantitative Landscape Under Index Enhancement The progression path of Zhong Ou Quantitative coincides with a critical juncture in the development of China's public fund quantitative industry. With the popularization of passive investment理念 and regulatory新规 strengthening "benchmark anchoring," index enhancement products are attracting increasing attention. These products retain the transparency and style stability of index funds while offering the potential for excess returns, making them ideal tools for balanced investors in a slow-bull market. Currently, the target indices for enhancement products are also diversifying. From traditional broad-based indices like the CSI 300 and CSI 500 to products focusing on sector-specific enhancement or thematic stock selection, investors now have more diversified choices. However, this also presents greater challenges for fund managers. On one hand, in increasingly crowded broad-based index tracks like the CSI 300 and CSI 500, managers must挖掘 factors with low correlation to peers and construct distinctive models to avoid their products becoming merely "high-cost index replicas." On the other hand, for enhancement products targeting specialized sectors or themes, a单一 quantitative model is insufficient. Managers need to genuinely delve into industrial logic, combining the depth of fundamental research with the breadth of quantitative tools, screening factors and models based on different stock universes to achieve differentiated sources of excess returns. In the view of the Zhong Ou quantitative team, this is precisely where their years of accumulated strength lies. The system built by Zhong Ou Fund—"fundamental quantitative foundation + active depth empowerment + triple low-correlation iteration"—provides a differentiated solution for the market: its excess returns do not rely on betting on a单一 industry or style exposure, but stem from a continuous, systematic, and evolving understanding and encoding of the industrial logic behind thousands of companies. Simultaneously, risk management is deeply embedded throughout the entire process, from strategy development and portfolio construction to trading management, in a procedural, node-based manner. It aims to continuously seek excess returns while proactively identifying, measuring, and constraining various risk exposures, striving for the long-term optimization of risk-adjusted returns. For investors pursuing "transparent Beta + robust Alpha," this signifies an investment solution that is analyzable, traceable, and committed to reducing market homogenization risks. This iterative capability is inseparable from the systematic support within Zhong Ou Fund. The practices of the Zhong Ou quantitative team also represent, to some extent, a direction for the evolution of domestic public fund quantitative investment: through the specialized division of labor within a high-quality investment research team, meticulously refined industrialized processes, and the digital-intelligent efficiency gains from advanced technology, continuously deepening insights are transformed into executable, verifiable, and sustainable sources of excess returns, thereby providing investors with long-term competitive solutions in a rapidly changing market.
Comments