"There is room in that section, but the blade of the blade has no thickness; with the lack of thickness into the room, there will be room for recovery, so it is nineteen years old and the blade of the blade is new."

Ten billion quantification, some open and close, multi-line narrative. Some of them go deep into the room, study the subtleties, search for every piece of data, and thoroughly study every factor, and the frontier of the century is one of them.  

01 "Follow logic and believe in data"

In 2010, Wu Di was studying for a doctorate at the Chinese University of Hong Kong, and Chen Jiaxin was studying quantitative finance in the second year of his sophomore year. Due to fate, the two life trajectories began to intersect and highly overlap.

Later, Chen Jiaxin joined Wu Di's entrepreneurial project as a partner. Although the project was shut down three or four years later, the two completed the first round of running-in. "The quarrels between me and Wu Di were all over when we first started our business." Chen Jiaxin said eloquently, calmly.

A few years later, after Wu Di received his Ph.D. from CUHK, he and Chen Jiaxin joined a platform company in Hong Kong, where they started the second phase of "working together".

This platform company has different trading teams. Chen Jiaxin and Wu Di are engaged in the cross-border arbitrage strategy of the domestic commodity exchange and the Chicago Mercantile Exchange (CME), and they have quickly done well.

In 2015, Wu Di and Chen Jiaxin left the platform company and formally established "Century Frontier Assets". Wu Di is responsible for the company's team creation, core trading strategy development, and the company's operation and management, while Chen Jiaxin is responsible for guiding the company's strategic direction, leading the investment research team to develop trading models, and comprehensively cover the strategic research work of quantitative investment.

After becoming independent, the team started self-operated trading, stock index futures in 2015, commodity futures in 2016, high-frequency stock lending and securities lending in 2017, and private placement licenses in 2018 to enter the low-frequency field, external fundraising in 2020, and 2021. After development, it has reached a scale of tens of billions.

Of course, there are difficulties and twists and turns in the middle. " Stocks do not have such high requirements for speed, but problems often occur. Data is often interrupted, packets are lost, and data is missed. There are also occasional problems when trading, and sometimes there are some blacklists that cannot be bought. Compared with futures, the operation of stock trading is very cumbersome. In addition, stock trading is different from futures, and many fundraising must be done through brokerages, which means that it is not possible to only be one brokerage and deeply bind one. Multiple brokerages also mean more There is a lot of operational pressure. " Chen Jiaxin recalled that from September to November 2020, he was "fighting fire" almost every day, and things could not be done, and the company was under great pressure.

It took less than a year and a half for Century Frontier to develop from self-operated to tens of billions of dollars. How do you view such rapid growth?

Chen Jiaxin described Century Frontier as a small but sophisticated company with a conservative style. Small and refined is that until the beginning of 2020, Century Frontier had less than 20 employees. Although there are currently 55 people in the team, Wu Di and Chen Jiaxin have always wanted to maintain a close relationship. Collaborative atmosphere. The conservative style is actually to form a stable and steady investment and research atmosphere. The conservative style is reflected in the fact that as early as 2016, when Century Front entered the field of commodity futures, it found that the market capacity was obvious and the industry ceiling was visible to the naked eye. Open up a new front in the stock market.

Small and precise and tens of billions of quantification seem to be a contradiction, but in Chen Jiaxin's view, team size and strategic capacity are not necessarily related. Capacity is more related to the choice of research direction, such as stock selection and low-frequency, which can better determine the size of capacity. Large-capacity requires large-ticket and liquid stocks, and requires signal advantages. The mid-to-low frequency turnover cannot be too high. If the turnover is too high or the timeliness of execution is high, it is difficult to make a special capacity. big strategy.

A small and sophisticated team will also face corresponding advantages and disadvantages. The advantage is that the connection is more closely, forming a decentralized research direction, and it can deepen the field and see more closely. The disadvantage is that the diversity will be worse, especially in Early on, focus more on key, important issues. For example, in 2020, more than 80% of the models were volume and price signals, and there was a lack of long-term fundamentals or event-based judgments. In the second half of 2021, the team began to consciously make up for shortcomings. The company's overall product framework and research system have matured. Especially in terms of low frequency and medium frequency, it was designed to support the operation of large-scale funds from the beginning, so it runs smoothly even after the increase in volume, so in Chen Jiaxin's eyes, such a scale growth is actually within a reasonable speed.

For the future, he believes that the equity market is gradually increasing, and the overall level of capital activity and market capital capacity is under the background of improvement. Institutions will naturally grow, but the company hopes to pursue natural growth, not blindly in a short period of time. Strategic support range. "But there is no limit. It should be a step to stand firm and continue to move forward."

Chen Jiaxin has not thought about it, hoping that the outside world should see what keywords should be used by Century Frontier. In his view, Century Frontier is a technology company in essence, " follow logic and believe in data. We may not have the concept of active investment in the quantitative industry, which is very fancy. When choosing a certain track or value, we want to follow some In a scientific way, the development of the entire system and strategy of investment research is a logical process, which we hope to convey to everyone the impression of the forefront of a century. ”

02 Quantitative 1.0-3.0 detailed training

Quantifying the nature of the industry leads to continuous involution of talents and fierce competition. This requires everyone in the industry to make rapid progress. At the same time, to maintain a competitive advantage, they must run fast and make continuous progress in research and development. In Chen Jiaxin's view, continuous progress is shallow. The level is a research style, and the deep level is a research system.

He doesn't quite believe in digging the sea purely from the data itself, and only making high-quality factors with solid logical support. Now the Century Frontier factor library has only 3 to 400 factors, and only about 100 are actually used. "You must ensure that the phenomena you find can be It is tenable, and the advantage is that for every logic in it, we all hope that it can be done very cleanly, and that things can be long-term and effective in the long run.”

Therefore, for the application of artificial intelligence technology, Century Frontier is very careful, "It is not a black box for us, but at least a gray box, because from the perspective of pure data, it involves a lot of detail control, and there are many places that can go wrong. A link is not well understood, and it is easy to be brought into the ditch. The result of the selection may not be able to be seen in the data in the sample, because there may be no difference in the sample. Saying it is inherently a risk does not generate new benefits.”

At the system level, Chen Jiaxin hopes that the company is a research and development-driven organization, forming the overall direction of the company from top to bottom. Based on his own experience, he summed up the different characteristics of the era from "quantization 1.0" to "quantization 3.0":

In the era of quantification 1.0 (the era of high-frequency futures) , one person can do transactions, remove noise from data, extract information features, and automate operations, all of which can be done by one person, but the market is progressing very fast, and the era of single-handedly has passed very quickly.

In the era of quantification 2.0, the platform system can separate data, signal feature research, execution platform, research platform, operation and maintenance execution, and risk control, and streamline transactions and streamline operations. Each professional is only responsible for a small piece of things. Putting it all together becomes a strategy with high overall performance . "Engineering problems are all independent, and those who do research can focus on core things, look at data and find patterns by themselves. This is the characteristic of industrialization." Dedicated personnel to full-time everyone can play a better level, but there is also a phenomenon that is prone to occur. , we overcome easy problems or do many, many homogeneous things. At the same time, due to lack of communication, progress is relatively slow, which will lead to too many attempts on shallow things. The way to quickly gain profits is the PM-type model, recruiting a large number of people from different backgrounds, and letting PMs find ways to expand their strategies. The organization can increase coverage in a short period of time and form diversity, but in the long run, it is still one person who solves all problems. Problem, see the upper limit, back to the 1.0 era.

Chen Jiaxin hopes to go a little further. To this day, he still takes two meals a day to communicate with colleagues. “Strategies are very sensitive. The ability of researchers to discuss strategies means that there must be a very high level of trust and recognition within the company, including competence and character. This also involves cultural construction. It is easier to establish a high level of trust between a relatively small core. Form cohesion and realize the leap from the era of Quantitative 2.0 to Quantitative 3.0."

From 1.0 to 3.0, from entrepreneurship to tens of billions, Wu Di and Chen Jiaxin are also facing their own transformation. "In the beginning, do it yourself. In the middle stage, you will guide others to do it, and solve other people's problems. Then you will be called strategic thinking. Think about things that everyone will not think about, and guide everyone to build a better overall. A research-based organization, how to improve the overall research efficiency of everyone and be responsible for the entire result.”

Working in a group of people with high IQs, profit is the most essential way of adjudicating after a dispute. " Speak with data and respect the facts. But some things cannot be completely understood from the perspective of data. It is necessary to carefully take out the details and what the specific characteristics are. This is a very objective process. "

Of course, salary is also an important factor in talent development. Wu Di and Chen Jiaxin also have an important task to develop a good benefit-sharing mechanism for the team. The purpose of the mechanism is to ensure that once the team members make good strategies, the personal returns can exceed the amount obtained by job-hopping or starting a business. Once enforced, it is not simple to be accurate and fair. "The evaluation of each Alpha in the strategy is relatively easy to analyze. However, there are no very quantitative indicators for the contribution of some heterogeneous dimensions, such as how much weight data occupies, and how much weight technology and strategy occupy. It is difficult to accurately distinguish. We must try our best. Use scientific rules to calculate how much each person contributes, and avoid slapping the head and saying that this person is 2 million and that person is 5 million. Of course, it saves trouble, but as a party, you will not understand it, and you will feel that you are taking less. Even if you give the formula to If he sees it, he will also feel that he is taking less." But if the partners explain the whole framework clearly, the parties will feel that it is reasonable, and then there will be a reasonable expectation.

More importantly, under such a transparent mechanism, conflicts will not accumulate within the company, and the establishment of trust is indeed a long-term process. "Gradually establishing a research and development system and building a sense of trust requires a long investment in the early stage, and then slowly builds up." Chen Jiaxin said. Under such a mechanism, Century Frontier has always maintained a cautious rhythm in recruiting. "We need to recruit with high level, sense of responsibility, and sense of collective honor, which makes our expansion relatively slow."

03 "Quantification should not do too much style judgment"

"High-frequency ability is a tool. If we want to rely on high-frequency to make high returns, it means that we are very picky about opportunities, so there will be few trading machines, and the capacity limit will be obvious. If the medium and low-frequency ability is strong, in fact, it depends on holding positions. Making money, but having the advantage of high-frequency execution means that I can do better for the same signal, and I can get the positions I want at a lower cost. If the performance standards are the same, it means that my capacity is larger, here It's a little counter-intuitive." Chen Jiaxin shared this in an earlier interview.

Chen Jiaxin has been in the business for 10 years, and it is difficult for Chen Jiaxin to say whether he prefers mid-low frequency or high frequency or other strategies. In his opinion, high frequency responds very quickly, and there are fresh and exciting things every day, but after relatively stable, he does not need to think too much. The mid-low frequency is a process of progress every day for a long time. There is no way to get too much feedback in the market, but from my own research, I can gradually feel that I am gradually accumulating. The high frequency is particularly good. If you do something right, you can see the effect immediately, which is obvious. Of course, there are more thresholds, and every detail needs to be scrutinized, technical delay requirements, and strategy details. If the most extreme configuration is not reached, it is difficult to make a profitable strategy.

It is precisely because it started from high frequency that the research on data of Century Frontier is very meticulous. "Domestic volume and price is still a mainstream part of the strategy. Everyone uses similar data, and the details will be different. This part actually tests everyone's understanding of data. Whether the cognition of the data is sufficient. We spent the most time and did the most research, and this is still the most important one and the one with the largest amount of data. Everything that can be seen in theory should be visible, but In fact, our feeling has been improving, and basically every month or two, we can still find some new things in terms of cognition on the model." Chen Jiaxin believes that data cognition is an area of expertise, but she is still far from it. Exhaustion is still a long way off. At the same time, he believes that artificial intelligence is essentially a tool for data analysis, and the advantage is to use this tool to better identify the features in the data in the context of a large amount of data.

The middle and low frequencies are just the opposite. The impact is more long-term. The impact of the strategy cannot be seen immediately, and it is difficult to judge whether the strategy is reasonable in the market in a short period of time. It needs to be inferred from the most basic logic. From the perspective of data It may not be possible to find the answer, but it is easier to do a "okay" strategy in the medium and low frequency environment.

Too fast style rotation is also a problem faced by many quantitative investors in 2021. Chen Jiaxin is reluctant to make judgments on style and industry. In his view, the objective situation is that fund managers also have suitable styles or industries, especially active investment. It is more inclined to disperse the positions, and some will be relatively loose in style, giving a certain degree of freedom to the signal of the model. However, Century Frontier will not place too much weight on this, because in his opinion, styles and industries are small samples, limited data, and narrow transaction width, which means that it is inherently difficult to stabilize the source of income. Adding a little possibility to the portfolio It's a reasonable thing to improve the whole combination, but it can't be too much. On another level, in fact, all styles must be completely neutral, unless you directly buy the benchmark index, otherwise it is quite difficult.

Looking at style changes objectively, it is necessary to realize that the excess return of quantification itself is also cyclical, and its more source is the activity of the market itself, when the turnover is relatively large, there is no direction, and the style rotation is particularly fast. In essence, it belongs to the cycle. In the long run, the increase in turnover is the trend, and it is normal for there to be some twists and turns in the trend. Of course, it will be particularly significant in 2021. Everyone is under pressure on the style rotation. In essence, competition has changed under pressure. Institutional functions try to catch excess when the cycle is good, and do a good job of defending when the cycle is poor.