Difference between high-frequency trading companies and quantitative investment companies
In general, high-frequency trading firms and quantitative investment firms are both related and distinct. In the United States, it is commonly said that high-frequency trading firms are generally self-trading firms, which mainly include Getco, Tower Research, Hudson River Trading, SIG, Virtu Financial, Jump Trading, RGM Advisor, Chopper Trading, Jane Street, etc.; while quantitative investment firms are generally hedge funds, including RenTec, DE Shaw, Two Sigma, WorldQuant, AQR, Winton, BlueCrest, Citadel, etc. In addition, Citadel, Two Sigma, and others, both with high-frequency trading and quantitative investment, and companies such as DE Shaw, both with quantitative investments and non-quantitative investments, have many companies aimed at a more integrated development.
Historically, many high-frequency trading companies were founded by traders who were originally engaged in derivatives, trading, and so on. These jobs did not require much knowledge to begin with. As computer technology developed, the degree and frequency of automation of trading gradually improved, and these companies gradually hired some people with a strong background in mathematics, statistics, and computer to adapt to the situation. Of course, this process also led to some differentiation, with some companies still retaining the dominant position of traders in the company, and never giving up on manual trading, eventually forming semi-automated trading with a combination of humans; while other companies have a higher degree of acceptance of new technologies, often adopting fully automated trading models.
The biggest disadvantage of artificial trading is that the place of manual ordering is far from the exchange, and the order is often not picked up at the time of market turmoil. At this point, companies with fully automated trading can minimize the time of signal transmission through the host machine room, but automated trading is often due to the complexity of the process, coupled with a large number of company personnel flow, there will be some errors in the maintenance of the program, and the end of the program is a disaster, such as the famous knight capital.
As for the over-fitness to black swan events, it is an unavoidable problem for both manual trading and automated trading. Generally, Getco, Jane Street, SIG, Virtu Financial and others are semi-automated trading, while Tower Research, Hudson River Trading and Jump Trading are fully automated trading.
Quantitative investment firms are very different from high-frequency trading firms. First, U.S. quantitative investment firms are generally founded by people with a strong background in quantity, such as Renaissance founder Simmons, a mathematician, D.E. Shaw founder David Shaw, a computer professor, AQR founder Cliff Asness, a financier, and high-frequency trading firms are founded more by traditional traders; second, quantitative investments generally rely on complex models, while high-frequency trading generally relies on efficient code.
The holding time of quantitative investment companies often reaches 1-2 weeks, the information needed to process to predict such a long-term price trend is naturally very large, the model is therefore more complex, and the speed of the process is less sensitive; the time of processing high-frequency trading information is very short (microseconds or milliseconds), it is impossible to analyze a lot of information, so the model also tends to be simple, the competitive advantage is more dependent on the efficiency of running the code, many people even write directly on the hardware program; and finally, the volume of capital invested can reach hundreds of billions of dollars, while high-frequency trading companies often have only tens of billions to billions of dollars, but because the strategy of high-frequency trading is more stable than long-term quantitative investments, such as Virtu Financial trading 1238 days of loss per day, therefore, they are generally transactions, and automated funds are generally helpful to investors.
Models for quantifying transactions
Below is a model of quantitative trading, from simple to complex:
The simplest is represented by John Murphy's algorithm for analyzing the futures market, which uses high-level mathematical knowledge such as indices, logarithms, etc. It is easy to understand and is better suited to subjective trading, or semi-automated trading where a computer calculates and sends trading signals manually ordered by a person.
A little higher in level, represented by Dennis's shark-in-the-sea trading rules, mathematics after all, using the content of university junior mathematics such as evenness, difference, normal distribution, the testing of strategies is also more scientific, and offers a reliable money management method, but the disadvantage is still not to get rid of the traditional idea of trading based on a combination of arrays of trading rules. However, if the strategy is well designed and the trend is great, it can still have a good effect.
The higher level is mainly reflected in the integration of trading signals, such as the organic integration of traditional technical indicators using more modern statistical methods such as regression analysis, neural networks, support vector machines, and the screening and testing of variables using more rigorous statistical methods. Given the time characteristics of financial data, it is often necessary to use rolling optimization to obtain test results outside the sample, so the resulting models are also more robust.
However, it is difficult to implement these functions in general programmatic trading systems, which require a more general programming language.
Simmons, founder of the Renaissance Fund, explains quantitative investing
If it is a quantitative investment, then in addition to the market information, other basic information must be collected, the corresponding time sequence is organized and integrated into the prediction model. Generally, the successful model is not how deep the mathematical theory is used, but how much it integrates the information from different sources. Even the simplest linear regression, if the parameters are strongly predictive and the correlation is low, the prediction effect of the model is good; on the contrary, even with the use of complex deep learning theory, if the parameters are meaningless, the resulting model is not useful.
Modeling is one thing, and solving models are equally important. For example, there are many models in physics that accurately describe reality, but they are often difficult to solve due to the lack of efficient scientific computational methods. Quantitative transactions are the same. The calculation, filtering, optimization, retesting, etc. of parameters are often accompanied by huge computations.
Common misconceptions in the field of high frequency and quantification
Quantitative models can't beat the black swan event
In fact, any investment method that relies on historical predictions of the future, is afraid of a black swan event, and there will be a reversal. The benefit of quantification is that after encountering a reversal, you can quickly incorporate the latest situation into the model, timely adjustments, re-review, optimize, simulate, and strive to reverse losses in the shortest time. For example, after Renaissance suffered a historically rare 9% reversal in August 2007, Simmons took the decision to rebuild the model, in a letter to investors, he declared that we had found three strong trading signals for the new model, which returned very quickly in the following days, reversing losses and reaching 80% returns for the year.
In fact, LTCM is a multi-strategy fund whose purely quantitative trading strategy ended up earning $100 million in 1998, and its biggest losses were from counter derivatives with extremely poor liquidity, many of which were even designed by itself to compete with the banks' counterparts, and failed to clear positions in time for the Black Swan event. These products were generally only used to help with quantitative models at the time of pricing, and the specific trading execution, product design, and sales were not tracked by quantitative tracking, and it is generally believed that LTCM's bankruptcy was more due to liquidity risk, not to the model.
High-frequency trading harms investors
The views of books like The Flash Boys are in fact very controversial, but the author's excellent writing and highly provocative narrative style have attracted a lot of attention. In addition to the media, it should be said that the United States is currently the strongest in demand to ban high-frequency trading, basically traditional traders of the time.
In the domestic market, now options are ready to be listed, stocks are also likely to open T + 0;; for these two pieces of fat chicken soup, foreign high-frequency traders have long been craving. If in the field of futures high-frequency, we can also rely on the rich experience in programmatic trading to compete with abroad, then in the field of options and high-frequency stocks, our practical experience is zero, the gap is greater than abroad. To this, he believes, on the one hand, we can not be arrogant, feel that foreign investment is too strong, simply do not do it; on the other hand, we can not rush to achieve, delusion of a year and a half to achieve great results.策略研究要慢工出细活，急于求成，频繁改变研究方向，最终很可能一事无成。