Analysis on Risk Control and Overseas Regulation of Quantitative Trading
At first, let’s probe into the historical development of quantitative trading, algorithmic trading and electronic trading. Quantitative trading itself is a specific way of quantitative investing. As a methodology used in asset management, quantitative investment started to develop in the U.S. market from 1970s under the main driving force of the gradual popularization of index funds and the increasing use of ETFs and other low-cost investment vehicles. Users of these investment strategies need to trade a basket of stocks. The wide application of these investment strategies promoted the development of quantitative stock selection, intra-day algorithmic trading and other quantification-based trading. With the maturity of internet technology and the improved security, many European and American financial institutions started to use FIX as their main electronic trading standards, which greatly reduced communication costs. As a result, electronic trading gradually replaced traditional brokerage to become a major business unit in overseas securities companies. As of present, quantitative trading, algorithmic trading and electronic trading have become a major trend and are used globally. High-frequency trading, which stands at the forefront of the trend, accounts for about 60% of the total trading volume of U.S. stock market. In some Asian markets (such as Australia and Japan), high-frequency trading contributes about 15-30% of the total trading volume.
As such, a deep analysis of algorithmic trading is necessary. Algorithmic trading has a very wide definition and is easy to be confused. In general, algorithmic trading is classified into trading in a broad sense and those in a narrow sense. Algorithmic trading in a broad sense mainly refers to an investment strategy and is one of the quantitative trading strategies. Such investment strategy features many trades in short period of time and simultaneous trading across different markets, and, as a result, is highly sensitive to trading costs. Therefore, algorithmic trading in broad sense is made up of two components: generation of trading orders (focusing on investment return maximization) and execution of trading orders (focusing on trading cost minimizing). From risk management perspective, maximizing investment return doesn’t conflict against minimizing trading cost: minimizing trading cost is very likely to maximize investment return. Therefore, the narrow definition of algorithmic trading refers to special trading strategies that study and improve the execution of quantitative investment strategies in a quantitative manner. So, algorithmic trading in a narrow sense is also called execution strategy. The effectiveness of execution strategy will directly impact whether the algorithmic trading in a broad sense (namely quantitative investment strategy) can maximize investment returns. In addition, as execution strategy covers order handling, position management, market connectivity and quotation interface, there can be a lot of risk points, posing great challenges to management. The aforementioned trading errors by several brokers and market participants are all connected to trading order execution.
The development of overseas algorithmic trading system is a gradual process. In the beginning of 21st century, when algorithmic trading and electronic trading just emerged, the system architecture was very simple. Usually, the system was a large and inclusive computer software program, always developed by quantitative programmers (not controlled by IT departments). In 2006, trading systems became increasingly complex. The original model failed to fulfill the needs of system expandability. Therefore, securities companies started to divide systems into several modules, and IT departments started to be involved in system construction. During the past 2-3 years, trading systems have been gradually decoupled from quantitative investment departments in many companies. IT departments become increasingly responsible for development, deployment, maintenance and upgrading of the trading system because of the commoditization of the algorithmic trading technologies.
After the emergence and popularization of electronic trading and algorithmic trading, procedures of risk management (especially trading procedures) are different from before. Specifically speaking, risk management metrics and inspection points have been spread out in the entire trading system. Risk management metrics for electronic trading, algorithmic trading and quantitative trading began to appear and overlap operation risk management procedures. Risk control measures using information technology and quantitative methods emerged and became real-time. As Exhibit 1 shows, the quantitative strategy module (the Strategies module in the figure, often developed by quantitative trading team) stays at the center of the entire design, but it’s still one of the many modules. There are a large number of risk points spread in the entire system. In algorithmic trading, processing speed, handling capacities, memory and CPU use, FIX message handling, order status and its transition process, position management, account management, pre-trade analysis and post-trade analysis will be monitored and recorded on a real-time basis by specific modules, including Pipe (risk management module for sub order), Manager (risk management module for main order), MDS (quotation module) and SDS (data module). This means risk management and regulation for innovative trading models like algorithmic trading and high-frequency trading have to be conducted via real-time sampling and monitoring across many points in the system, and multi-level and multi-tier oversight of the entire order placing process, in an attempt to prevent accidents. Of course, the setting of above risk management points will reduce order placement speed for some algorithmic trading and high-frequency trading strategies that pursue high-speed trading. Instead of reducing risk points, the solution to this problem is to adopt more advanced software and hardware technologies to improve risk inspection speed. FPGA technology, which is extensively used in overseas markets, applies hardware with software embedded and is widely adopted to speed up risk inspection.
At last, we will discuss the global regulation dynamics for algorithmic trading and high-frequency trading. In fact, trading deregulation is one of the driving forces for the development of algorithmic trading and high-frequency trading in overseas market, at least before 2010. In 2000, American stock market started to allow one-cent tick size, which spurred the popularization of computer-driven trading and algorithmic trading. In 2006, the enforcement of Reg NMS enabled traders to trade one stock in different places. Such regulatory changes lowered trading costs and brought rapid development of trading technologies. However, the lagging behind of risk inspection points caused trading errors, such as the “flash crash” in May 2010 and the Knight Capital incident in August 2012. Global regulators are mulling rules to prevent such incidents. For instance, Hong Kong Securities and Futures Commission launched a consultation statement on regulation of electronic trading in 2012, in order to solicit participants’ opinions on regulation of electronic trading and algorithmic trading. The specific measures proposed by Hong Kong Securities and Futures Commission mainly include: first, related practitioners are required to hold corresponding qualifications, have deep understanding for related models and technologies and receive regular training courses; second, the importance of quantitative tools and methods, especially pre-trade and post-trade tools, in electronic trading monitoring and management process shall be highlighted; third, it’s required that securities companies who provide electronic trading services to their customers and/or have in-house proprietary trading units to ensure their electronic trading systems meet the standards with respect to technology stability and integrity.