"Are we getting enough data now?"

"Where is the core point of alternative data?"

"Is quantification the pursuit of absolute correctness or vague correctness?"

"Are derivatives more dangerous, what do you think, what are you afraid of?"

"Should we apologize for the investment process?"

Looking for the fuzzy correctness in the massive data, and combining the market and probability to find investment opportunities in the fuzzy correctness. Quantitative investment is not a big gamble, but a protracted research and penance.

During this hardship, Han Xue, the quantitative investment research team of Guotai Junan Futures Asset Management Department, was asking questions and answering himself. She felt "deceived" into the industry, but she also chose the most difficult one among all the tracks. She tried to form her own technical barriers from a long-term perspective. Now that the quantitative market is constantly in line with the international market, she has found herself in the investment field. a place.

01 "Are we using data enough now?"

"Have we fully utilized the data now?" Han Xue asked the writing team shortly after the exchange started.

Han Xue, worked successively in domestic high-frequency private equity, strategy director of overseas fund companies, and investment manager of Guotai Junan Futures. She gave up a high salary and a stable life. She chose to challenge involution and dreams. Now she is mainly engaged in options investment, although she is under 30 years old. Although he is old, he has built the company's option investment and research data monitoring screen, characteristic database, and option localization high-frequency backtesting platform.

" Everyone has a process of mining data and information. " The writing team was a little surprised that the respondents were "anti-customer-oriented", but still replied.

From traditional financial data and daily-frequency data of transactions to high-frequency data and alternative data, the industry continues to progress from the shallower to the deeper, and this is a necessary process. Take news public opinion as an example, traditionally, each news is given a sentiment score, optimistic or pessimistic, and then made into a binary variable of 0 and 1, but this kind of processing is rough and swallows the huge information hidden behind it.

Subsequently, the industry adopted a natural language processing model to analyze the emotion of each sentence in it, and then a more refined continuous 0-1 indicator was obtained. Some people would also observe the same analyst and judge the chronological order of the same stock. Dig for information above.

As far as Han Xue is concerned, with the development of natural language processing technology, including the emergence of Attention, Transformer and other technologies, the labels of text analysis can be more specific and accurate, and the information that can be extracted can be more diverse, rather than a single emotion determination. Traditional financial and regular data are no longer incremental data, because there are only so many types of data, and this is one of the reasons for the industry's involution, because regular data has not expanded, and everyone will always be among these types of data. In mining, everyone gets the same data, so the quantification can only be narrower and finer, from minute to tick to tick-by-tick.

Of course, in this process, everyone's understanding of data is also different. "Take implied volatility, it's not a new word, but there are so many ways to calculate implied volatility, what's the reason behind this? Because it's an inferential formula, and the man Black-Scholes really is Very great, but very shameless! Because the formula is full of all kinds of assumptions, stock prices follow a lognormal distribution, and stock prices follow geometric Brownian motion."

She believes that the next explosion of incremental data comes from alternative data, but alternative data will face: 1. Data sources are difficult to determine; 2. Clean and stable data is expensive; 3. Privacy boundaries. 4. The volume of unstructured data is large, and it is difficult to predict how much returns will be generated. 5. The data should follow the local customs, and also involve the problem of language construction.

" So, let me think about the core point of alternative data? " Han Xue asked the writing team again, this question is not in the outline.

"It's about storage. This is a problem in the future of data, and it will definitely be encountered in the next 5 to 10 years."

3 or 4 years ago, the chips she could buy were worth 2-3,000 yuan, and the price in January 2022 is close to 20,000 yuan. The reason is that a large number of institutions need to combine, clean and organize data. Next, with the quantitative development process, after the price of hard disks has also experienced a sharp rise, large institutions will have their own data centers, while other institutions may share data. There are already mature manufacturers of common volume and price rule data, but Alternative heterogeneous data is rarely available. In the process of incremental data development, computing power will continue to expand. From multi-threaded running to GPU, it will become more efficient and flexible, and achieve more convenient isolation. In lighter docker, learn how big Internet companies respond The elastic computing method of kubernetes with data shock can not only optimize the allocation of resources, but also realize cross-border between languages.

02 "Is quantization absolutely correct or vaguely correct?"

"Data, algorithms, computing power, in terms of quantification, are we pursuing absolute correctness or vague correctness?" Han Xue asked the writing team, and was also asking questions about this profession.

She is an idealist, trying to be perfectly right in the midst of ambiguity. She completed a double master's degree overseas, skipped a grade halfway, and returned to China after working in the United States for a year. She chose derivatives with few people to enter the market at a high frequency.

"Why did Black-Scholes make an inferential formula? You will find that there are various possibilities in each fuzzy process, so this is the more interesting point. " Han Xue asked herself to be in every algorithm. You have to work harder and do your best, but later you find that it is useless to do your best, because there may be a vague correctness in every aspect, and then you will run first in the vague correctness.

She chose the track of derivatives because of the higher threshold of the track: "Because many people think that derivatives have a complex structure and make less money than stocks, it requires a lot of accumulation at the beginning, and a lot of manpower and material resources will be invested in the follow-up. , financial and computing power. But it will not crowd with everyone, and it can reduce the influence of some others on me.”

She did not deny that she joined the industry because she wanted a "money printing machine". In the 13th century, the "Hennekau Magic Wheel" was born, which made scientists dream of "perpetual motion". Hennekau Magic Wheel", but they all failed miserably. However, the countless failures did not dampen the enthusiasm of scientists, but they became more and more enthusiastic about the exploration of perpetual motion.

But even money printing machines have to keep improving and iterating. At the same time, she believes that she did not choose to quantify, but that active investment did not choose herself. "I don't understand people's speech very well, and I can't judge whether it is up or down from the changeable "talk" of announcements, but the trend is very visible to the naked eye. I can do probability statistics, and then use the law of large numbers to win it. Okay."

"If quantitative trading is based on the principle of large numbers and probabilistic advantages, under the current trading sample size, is it a placebo or is it really effective?" the research team asked.

"Investing for 10 years is definitely effective." Quantification is always considered to be short-term and short-term, and Han Xue is high-frequency, but she thinks she is looking at 10 years. "There is a big difference between investment at home and abroad. Foreign countries believe that investment is a perception of the future. They hold annual meetings, but we publish weekly, monthly, and quarterly reports. The investment cycle is very short, even as short as every day. Report it. Should you apologize if the yield drops? We don't have a way to apologize, the market is always right, there is no right or wrong, why should we apologize? Because this is a process of investing, a process that is always moving forward, and we shouldn't deal with it A process of apology? Failure is the mother of success, and we should look at the longer term. We are just friends of time and companions of joint investment.”

Of course, Han Xue also encountered a setback, and the first time she encountered a big setback was during her student days. "One day during the winter vacation, I lost hundreds of thousands of dollars in my personal account, and I was not very happy. It reached nearly half of my total assets. I expressed a little panic, so I kept staring at the plate like an ant on a hot pot. At this time, Han Xue's father came to ask what happened.

"I lost money," she thought for a while, then she thought, "Dad, will you scold me?" In Han Xue's opinion, this is a problem that many young investors and traders will encounter: caring about what other people think of me .

The father's answer was not "Parents are rich" or "Investing is dangerous and you shouldn't be involved". Instead, "This is your own account, and you should manage it yourself." This gave her a good start to her investment view: every investor will encounter drawdowns, and they should have their own ways to relieve them. Don't go If you care about other people's opinions and vision, you can cry and vent, just like Ruan Ji's poor road, but as long as you are responsible for your account, reflect on, review, strengthen risk control or logical errors and stop your losses as soon as possible, it is the next move that should be done. .

03 "Derivatives are more dangerous, what do you think?"

Han Xue's job is to build thousands of derivative portfolios and look for opportunities among them. In her opinion, the biggest advantage of options lies in leverage, but the danger may also be leverage, which is both an antidote and a poison. Investors should choose what kind of winning ratio to play at, and make reasonable position control to keep the game going. This is very important. Don’t go offline completely after a certain failure.

And stocks give Han Xue the greatest diversity of feelings. "There are more than 4,000 aspects of more than 4,000 companies, which have their general features and their unique features. There is a word called consensus. I think this word is very interesting. Do you think Let the market agree, or do you want the market to agree?"

She doesn't want to agree with others, she prefers to make unique investments. Clausius proposed the second law of thermodynamics, which states that in a natural process, the total disorder of an isolated system does not decrease. In short, the entropy of an isolated system never decreases automatically. The entropy does not change during the reversible process, but increases during the irreversible process. It can be said that it clearly points out the direction of the irreversible process, and you can independently think about how to create your own closed loop of investment logic. .

" I am willing to dormant when the market is deeply smashed, waiting for a very certain result, the moment when the 'pop' is lifted, it is not only a very beautiful curve, but also the market's response to The resonance of your investment philosophy. ”

Therefore, she believes that she is very suitable for derivatives. Derivatives have their own leverage, which will make investors feel awe. In the process of needing to be more cautious, like value investment, it takes a lot of time to research and ambush.

"What are you afraid of? What you are afraid of is not knowing, not knowing, not understanding, but not whether it will change tomorrow, the answer will definitely change." We are always in awe and looking forward to this market, leverage, technology, and new varieties . From her point of view, it is difficult to say that a "mature" investment manager in this market is a mature investment manager, because there is no perfect formula, so there is no real maturity. Seniors such as Buffett can hardly predict that after 80 years of U.S. stocks, they will encounter consecutive encounters. 4 fuses.

Therefore, this market needs risk control. The common use of the industry is Greek letter control. This is a sign, but if the investment manager has a thorough understanding of the strategy, and the risk is more controlled by the structural mechanism of derivatives, many risk controls are unnecessary. Plus, it will affect the transaction speed, so she believes that the real risk control is after understanding the principle.

Another more common angle is the risk control of market prediction, that is, the risk control of the market, and the second is the risk control of operational errors, that is, the specification of the transaction process. The former requires a rough prediction of the market's return, and a maximum return limit can be set. For the latter, in fact, quantification is to reduce operational mistakes, emotional mistakes, and fatigue mistakes. But operation mistakes are always unavoidable for the machine. There may be a problem with a certain interface. Don't be afraid of mistakes, because there is only one standard answer, that is to test more, test regularly, find bugs, and then debug.

"There are also investors who think options are dangerous, so what do you think?" Han Xue asked.

04 The quantitative industry lacks "fools"

"Quantitative trading still needs to introduce more algorithms. This industry has never lacked smart people, but "idiots", and treat it as a startup." Han Xue believes that institutions should not set strategies through interviews or interviews, everyone is smart people , this move will deplete the trust and reputation of talents in the company. But she is willing to communicate with the industry, including algorithms, heterogeneous data, etc., in order to combat entropy increase, and get timely feedback for common progress. "I will think about what kind of purpose I want to achieve, and then I will deduce what kind of algorithm is better to use. Sometimes when I communicate with you, I will encounter some people in the industry who have always wanted to try, but have never used it, and no one wants to. When the first person to eat crabs, the performance may not be very good, the exploration process cycle will be longer, and there will be some cuts in the middle. But failure is the mother of success, although we don't like failure, just like we don't like it Bugs are the same, but debug is the way to go.”

She firmly believes that the current algorithm still needs innovation. As the requirements for the fineness and granularity of data become higher and higher, the industry can no longer adopt simple and crude algorithms, but we haven't heard about the birth of new algorithms for a long time. I can't help but sigh that the industry has been struggling with new algorithms for a long time. Newly recruited interns or young employees should fully mobilize their curiosity and creativity. More conventional things have been done, and this is one of the reasons for the involution of the industry. For example, various regressions, genetic algorithms, neural networks, etc., no matter whether there is a model or no model, everyone still needs to develop their thinking. For example, recently I am paying attention to the application of blockchain technology to the financial field, through common nodes to authenticate, systematically To solve the trust problem, it is easy to prevent transaction records from being tampered with, but it will also encounter very few tail conflicts.

Overseas quantitative layout in China is also a challenge faced by domestic quantitative practitioners.

"In the future layout of large overseas quantitative institutions such as Renaissance, I think it must include entering China, because capital will go to emerging markets with high alpha, so it will be fully rolled out in the next 3-5 years." She believes that this For domestic capital quantification people, it is an invisible pressure, which must be broken by oneself and mastered by foreigners. "Overseas institutions are currently pursuing ultra-high returns, not scale returns. They have high requirements for capital and are not in a hurry to expand for the time being. Therefore, the opponents of tens of billions of private placements are overseas institutions rather than other tens of billions of institutions. When these overseas institutions want to expand, the market appeal will be very strong, and the domestic tens of billions of quantitative scale advantages will no longer exist

Of course, stress also promotes learning. In Han Xue's view, foreign investment and research are very solid. For example, foreign institutions will measure soil temperature and cooperate with certain research institutions. These seemingly distant information is the increment of heterogeneous data, which is closely related to the market. "She doesn't think that overseas institutions must be called leaders, because domestic practitioners also have the same potential, but overseas uses a more open thinking, such as transferring behavioral economics and psychology into the investment field, and then using different methods to Solving problems, and China often likes to follow hotspots.” Moreover, quantitative technology is not only reflected in strategies, but also in engineering. For example, many hardware devices are not domestically produced, and many artificial intelligence papers are also published overseas. It cannot be said that we do not have classes. Representative, but in terms of average score, I think China's quantification is still in an awkward position, but it needs to be in a state of catching up and daring to be the first. "So Han Xue is very supportive of institutions buying computing power, researching supercomputing, quantum computers, and building their own effective fundamental databases. China must step out in this step, and do things that others have not done, even if it fails. , even if it retraces, it is worth it.