Systematic investment is becoming the mainstream of active investment in the AI era. One of the most popular topics on Wall Street is "How to empower active investment with machine learning/big models?".
As a global asset management giant with a management scale of $10.47 trillion (as of the end of March this year), BlackRock has already given a mature answer - as its flagship strategy, the Systemic Active Equity (SAE) strategy uses big data analysis methods such as machine learning and natural language processing to capture investment signals that have guiding value for investment, with over 35 years of research experience and investment practice. As of the end of the first quarter of this year, SAE Strategy's global assets under management exceeded $150 billion (approximately RMB 1090 billion).
In the Chinese market where foreign giants are increasing their allocations, SAE strategy has also shown impressive performance in related products.
Capturing investment signals from complex and diverse data
The core operational logic of the SAE strategy can be summarized as follows: based on big data analysis, it filters and extracts valuable investment signals faster, wider, and more accurately, and adopts strict risk management to strive to create sustainable excess returns for customers. This can be said to be a landmark strategy for the implementation of the "asset management technology" concept in actual investment research.
The big data and technological innovation attributes of SAE strategy are also reflected in personnel composition and institutional mechanisms - it is a rare strategy that has about 100 investment research experts, big data scientists, and technology personnel, and independently develops and iterates investment signals, as well as tools used in investment processes such as big data analysis models and investment portfolio optimization engines.
Zhao Rui, Managing Director of BlackRock Group and Senior Investment Manager of SAE Team, introduced the investment methodology of SAE strategy from the aspects of investment goals and concepts, risk management, and transaction cost management
Firstly, the investment objective is the SAE strategy, which pursues long-term and sustained excess returns. In terms of investment philosophy, the strategy model uses multidimensional data analysis to comprehensively analyze and predict the fundamentals and investment sentiment of individual stocks, in order to find the investment target with the most excess potential.
Secondly, there is risk management. The SAE strategy always invests in all positions, does not concentrate positions, practices diversified investment, and has strict risk management in individual stocks, industries, styles, and other aspects.
In addition, transaction cost management is also an important aspect of strategy implementation. The SAE strategy considers transaction costs when optimizing investment portfolios and selects targets with similar expected excess returns and risk contributions that have more cost-effective transaction costs.
The data supports the implementation of this investment methodology.
Zhao Rui told reporters, "Our group (SAE team) has been working on big data and machine learning since 2008. The data science team has been continuously developing and innovating, forming a rigorous and efficient data analysis process, which enables the transmission of data and the interaction between investment and research personnel to be continuously optimized. For example, our natural language analysis engine has now iterated to the sixth generation and can effectively analyze research reports, performance conferences, news, forum comments, and other information to help us understand the emotions of various investors and form valuable investment signals.".
In terms of model construction, the SAE strategy consists of hundreds of investment signals: unlike traditional factor investments, signals are independently developed and iterated based on current market phenomena; In addition, the main source of signal data is non-linear alternative data, which is also the main source of differential excess returns for this strategy.
Zhao Rui further explained what "alternative data" means: "For example, looking at job advertisements to see if the relevant company is still hiring, what kind of positions are being recruited, and what skills are needed for the positions. These pieces of information can help us understand the company's business situation, whether it is in the expansion stage, and the future development direction of the company.".
Zhao Rui believes that the degree of digitalization of the Chinese market is far higher than that of overseas markets, which is related to the convenience of transactions in the Chinese market, the numerous Internet business models and other factors. Based on the massive data in the Chinese market, the SAE strategy model can provide real-time insights into thousands of factors such as industry development trends, regional characteristics of economic activities, business operations, consumer preferences, etc., in order to predict the future trend of stock prices and construct new investment portfolios.
The overseas fund products managed by the BlackRock SAE team for investing in China were officially launched in 2012 and have been integrated into big data and machine learning investment signals since 2013. Starting from 2021, BlackRock's joint venture wealth management company, BlackRock Jianxin Wealth Management, has issued wealth management products using the SAE strategy, achieving localized investment operations of this strategy.
According to the reporter's understanding, the equity wealth management product "Beiying" series issued by BlackRock Jianxin Wealth Management Company adopts the organic combination of BlackRock SAE strategy concept and local investment research operation.
Machine learning signals account for 30% of the strategy model
Systematic investment is becoming the mainstream of active investment in the AI era. One of the most popular topics on Wall Street is "How to empower proactive investment with machine learning?".
"In fact, BlackRock has also used Transformer technology similar to ChatGPT in our self-developed Large Language Model (LLM), which maximizes the accuracy of natural language processing technology and empowers systematic active investment," Zhao Rui told reporters.
Compared to ChatGPT, which focuses more on human-computer interaction applications, BlackRock's LLM is specifically designed for tasks closely related to investment, including predicting the stock price performance of listed companies after financial reporting meetings. Therefore, BlackRock's LLM model relies on more targeted databases for training, which can demonstrate higher accuracy in specific tasks in investment research.
"The proportion of investment signals related to machine learning in our entire strategy model has reached 30%, and in 2019 it only accounted for 15%. Due to technological progress and improved accuracy, their weights are gradually increasing. We were the first overseas fund to use a systematic active stock strategy to invest in the Chinese market, which was issued as early as November 2012," said Zhao Rui.
Relying on machine learning, the systematic active stock investment strategy highlights the word "active", actively mining stocks to strive for excess returns. Its stock selection logic is more similar to fundamental investment, conducting comprehensive analysis of individual stocks, which is also different from traditional quantitative investment. In terms of strategic capacity, the frequency of systematic active stock investment strategy adjustment is not high, and the highly diversified holdings make its requirements for individual stock liquidity lower than traditional quantitative investments, so the limit on strategic capacity is relatively small.
According to the reporter's understanding, the systematic active stock investment strategy mainly explores and invests in three major categories of stocks:
One is stocks with good fundamentals and high valuation attractiveness, which require the company to have stable profits, sustained growth, and room for unexpected increases. At the same time, the stock price valuation should be cost-effective.
The second is stocks with positive market sentiment. The model tracks and analyzes the views of market participants (such as selling analysts), the positions of other investors and their expectations for future fund flows, trading opportunities brought by short-term liquidity, and the correlation between stocks and other assets.
The third is industries and individual stocks that align with macro themes, and the model will focus on analyzing the domestic and foreign economic environment and policy guidance. For example, export industries driven by the rebound in external demand, and related targets encouraged by high-quality development policies in the capital market.