One day in December 2013, a monkey walks into the Pacific Cafe at the Hong Kong Science Park. He wants to meet a 14-year-old student from Qingdao, who is the legendary master of quantum trading, Liu W.
The monkey is brown-eyed, wearing a sweatshirt and jeans, wearing black glasses, and has a student's taste.
Dressed in plain clothes, he carried two mobile phones, one of which looked like a very old Nokia handset.
After obtaining his PhD in electronic engineering from Hong Kong, he joined a technology startup. He and his core team of more than a dozen people made the world's first 4G mobile chip based on the TD-LTE protocol, which was successfully demonstrated at the Shanghai World Expo.
At the time, the team's partners were eager to promote 4G mobile in China. Mobile's senior management had said that the Chinese 4G era would be two years later if not for the team's push.
The monkeys are proud of what they have done.
Companies in the region used to be the target of capital. At one time, everyone was hot-blooded, thinking that the top startup board was a nail on the board. But the management was in too much of a hurry, wanting to eat the entire industrial chain, the team quickly had many problems in the details of execution, plus the competitors quickly caught up, and the company's capital chain was quickly tight.
The startup team, which was so close to success, began to struggle. The company fell into a slump. He was working on transformation. He wanted to go back to teaching at the university and study the then-hot wearable industry.
At this juncture, a friend came to me and said that the Qinhua master, who does quantitative trading, wanted to consult on the system issues.
The monkey had a great chat with W. Hearing that the monkey master, Dr. did signal processing, time sequence analysis, and now works as a chipmaker at a startup, after gaining experience with low-latency systems, W said:
“我们公司也需要Ph.D，不如我给你一些数据，你看能做出什么模型来？”
Quantitative trading, especially programmatic trading, requires a person to walk a two-legged path: on the one hand, to understand technology, on the other hand, also to understand algorithms. The monkey is accumulated in both technology and algorithms. He weighed the opportunity ahead of him, secretly telling himself not to miss it.
It was the Christmas holidays in Hong Kong. The monkey took the data from W, wrote two programs all night, experimented, analyzed the time series using various methods, including signal processing, machine learning, and made several models. The monkey spent another day writing a dozen pages of reports.
Monkey wants to give the results to W as soon as possible, because financial data is sometimes efficient. Also, he has always been more of a perfectionist. If people say I'm reliable, that's the highest evaluation of me.
Looking back now, the monkey feels that the model he wrote at the time was not a monkey, the pediatrician must have been a monkey, but how much of it reflects his research attitude.
In 2014, the monkey returned to the mainland from Hong Kong to join the quantitative trading team founded by W, becoming the 22nd employee.
Most of the colleagues on the research team have PhDs, some have done cancer research, some have worked on rocket engines, some have worked on Nature/Science, and some are sea-bound professors, but they are all very low-key and modest.
In the early 1980s, 14-year-old W with excellent results in the examination of Qinghua, after graduation, the public sent to the UK to study doctorate. Later, he worked as a consultant, founded several companies, companies listed or acquired by listed companies. After financial freedom, W played quantitative trading with intelligence and insight without any financial background.
Many of the strategies he wrote are still profitable today.
However, the company also has a place for original branding.
When he first joined the team, Monkey was surprised to find that the company was still running W, a strategy code written in Fortran many years ago.
This is a legacy problem, the boss only knows Fortran, his coding style is so bad, at first glance it is not from writing programs.
Monkey has been programming in C for more than a decade, learning Fortran and standardizing and organizing all the company's strategies.
When he first joined the company, he didn't even know what a hedge fund manager was, he didn't even know what a hedge fund manager was, he didn't even understand some of the names discussed by others in the financial quantitative groups. So he immediately went to Google or Baidu. Sometimes when he Googled a name, he would come across a new unfamiliar name, and he would dig a layer down until he understood the problem completely.
During that time, Monkey was in the library every weekend. The company was about to start trading options, and Monkey needed to join. He bought John Hull's Monkey Options, Futures and other derivatives to learn.
Based on the mathematical results, the comparison between the Chinese and English languages shows that it took two weeks for the monkey to read the Optional Bible.
During his time in the options group, the company's performance suffered a significant decline. In a hurry, Monk began to study futures models in his spare time. He was then transferred to the futures group, where he slowly became responsible for the company's futures team and related strategies.
During the 2015 stock market crash, monkeys traded mainly within the day. During those months, the microstructure of the market changed significantly every day, monkeys were busy as if in a battle, adjusting the model at any time to quickly solve the problems encountered by the model in the real world.
One morning in early July. The futures of the S&P 500 stock index became very liquid due to the sharp increase in collateral, and when I opened the deal, the price stopped and then fell.
If the monkey doesn't see it, stop the strategy immediately. If you do the opposite, 20% will be gone.
In the midst of the stock market turmoil, the company has done well and despite being tired, it has a sense of accomplishment and pride.
But monkeys soon lose interest in intraday trading. You can't really do anything about those things, based on the microstructure of the market, and so on. You generate different models, at best, just scattered signals.
Monkeys looking for breakthroughs began to study low-frequency strategies. He soon found that Overfitting was ubiquitous. A ridiculous model was only complex enough to explain the data in the sample perfectly, but such a model had a very poor interpretation of the data outside the sample.
Monkeys use time-series analysis to find factors that appear to be highly correlated, combine them, and find curves that are beautiful, with a Sharpe ratio of 5 or 6.
I was so happy when I saw it, and then I realized that it was a random fluctuation.
The monkey started to reflect, and he found that many times the factors that the statistics found were related to the monkey, not the monkey's cause and effect.
Well, if you find that yesterday you were stunned, and today you are likely to be stunned, then make a model, and the historical data is pretty good. But you're actually just finding a correlation, not a cause. You don't know when that factor fails, because that kind of factor is inherently illogical.
In addition, even using non-sample data is not true non-sample; historical data to distinguish between in-sample and out-sample would have implied over-fit; only really submitting this model, taking the data from the market, is the true shell of the sample.
A friend of mine who works in the foreign exchange business told Monkey that he had done a review using three years of data and that the machine learning method was better than the traditional method. Monkey suggested that you should look back three years.
How do you guarantee that the next real record is the last three years or the last six years?
Monkey feels that machine learning is useful in some areas, such as go-go; but financial data is limited, there are not enough sample points, and market information is completely asymmetrical. Machine learning is therefore difficult to replace subjective trading.
During the heated black market boom of 2016 (referring to black minerals and related products, including screws, thermosets, iron ore, coke, coke coal and power coal, etc.), Monk saw firsthand how many people made a lot of money in the first 11 months with deep learning and machine learning strategies.
As a result, on the night of November 11, domestic commodity futures overtook the mountain bike market, with several varieties going from a standstill to a slump in just a few minutes.
You feel like the trend has come out, you just started doing more or nothing, the trend has come back again. All the fancy ways of doing models are in retracement, basically losing money.
The price is just the result, not the reason. The monkey has gone through many twists and turns, groped slowly, realized deeply that there must be a strategy, and there must be logic.
The result of the monkey's thinking is: basic.
He said: "Hey, we make futures, but many people don't know anything about coke, coke coal. I've never seen what a copper cork looks like.
The monkeys are determined to get into the circle of the various types of monkeys in the futures industry and learn from them.
Many of these dams have a spot-on background, some of them are made of screw steel, maybe from steel mills, coal, maybe from wells. They understand the logic of the industrial chain very well, which is what the monkeys need to supplement.
When he first joined our WeChat group, no one knew him. He asked us questions, using the red bag model. When he didn't understand what we were saying, he asked privately.
You must ask valuable questions, not stupid ones, otherwise you are wasting your family's time.
Sometimes, the monkey says that recent events are like a historical moment. But the monkey does not do a retest, does not analyze historical anomalies.
The monkey learns basic logic from the monkey; these logics become the starting point for his modeling.
Take last year's coke coke market. The monkey learns that since September, the entire coke stockpile has been zero. The coke plant produces coke, the cars line up at the door, produce it and pull it away.
In terms of supply and demand, production is low, demand is high, spot prices are high, and futures are still in liquidation.
Since joining the team, he has been working daily, daily, options, futures, from quantitative to basic.
Monk does not like to quantify fundamentalist believers, many do quantitative teams, and talk about various theories, models, and so on. He is a pragmatist, he believes that making money is not divided into 369 and so on.
You can make money by multiplying, I definitely don't use multiplication, let alone arithmetic.
Over the past year or so, monkeys have become increasingly aware of the fundamentals of the various futures varieties and the logic between them.
The industry's inspiration for monkeys is to do relative value instead of doing monkeys as two homogeneous monkeys (generic trend-tracking models for simple technical indicators). It makes no difference whether you have two homogeneous monkeys at a low level, or two homogeneous monkeys at a high level, or two homogeneous monkeys in Deep Learning.
For example, since the raw material of screw steel is coke and iron ore, then the absolute value of its production is better than the relative value. The profit of the steel mill can be calculated by coke and iron ore. The profit is high enough, surely many steel mills will resume production, more supply, and the profit will naturally come down.
However, these industry leaders also have a disadvantage: they do not quantify ideas, which can often be very costly to implement.
One of his friends, a big name in the industry, told him that on the night of the twenty-first, he felt that the big market was not right, so he quickly called the trader and asked him to flatten the position. The result was that the trader dropped three prices and did not catch up.
And I said, "That's not a big deal to me, I'll settle it for you soon".
The monkey has a habit of figuring out the reason behind the phenomenon. If the market is pulling hard, he must figure out why. One day, the plastic tailpipe crashed at the last minute.
But financial capital doesn't understand this, it's looking at two straight lines to make trends. The two straight lines say buy I buy, buy I get out of the basics. The industrial giant has a stock, your futures are so much higher than my costs.
For Monkey, quantification is always a means, not an end. The wider the trade path, the more interesting it is for Monkey to do. He feels that the market is full of opportunities.
In February of this year, I had lunch and chatted with a monkey in Shenzhen.
He was wearing jeans and a T-shirt with the "Tsinghua 8" embossed on it. He said he understood Jobs' style and that the simpler his clothes were, the better, and it was better not to make me think about what to wear.
The monkey asked me to eat monkey soup in Shenzhen. As a result, he was busy sharing his trading thoughts with me, and after several hours, he ate only a few bites.
As a child, he dreamed of becoming an archaeologist and was interested in the history of astronomy, geography and international politics. He once felt that he had wasted three years of his precious time studying for his doctorate. Sometimes he would think that if he got out three years earlier, he would be able to buy a house a little earlier.
However, now the monkey has realized the meaning of reading the doctorate. The doctorate gave him rigorous academic training, developing his ability to discover problems, solve problems, which allowed him to be steady on the path of quantitative research and trading.
However, when I was studying for my doctorate, I did my research to write an article, but now it's a lie.
The criteria for a brick dealer's evaluation is very simple: if they are honest, they will get their money's worth, and if they don't make money, they will be trash.
I read a lot of reports written by macroeconomists, and I feel like I'm not grounded, I write reports to write reports.
For example, before the 2017 Spring Festival, the central bank raised the MTF (medium-term credit facility) by 10 basis points, and the day after the Spring Festival, the central bank raised the repo and SLF interest rates (conventional reserve facility).
He believes that these statements are illogical: you did not increase the yuan at the end of last year when it was 7, and now it is stable at 6.8, and the Hong Kong off-shore yuan is more expensive than on the coast, and you run to raise the stable exchange rate, is that not funny?
The monkey also noticed that the IRR (water retention index) at the time was negative 10%. So he chose to strangle the entrance monkey at the highest point of the safety margin.
Before making a decision, the monkey will consider the right logic, the opposite logic, see if he can be convinced, and then come to his own logic.
Monkeys emphasize the ability to analyze logic in this way, which is crucial for strategizing, trading, and even finding bugs in programs.
This ability to analyze logic was also the biggest exercise given to him by his previous startup.
At that time, a chip often ran for days and nights and hung up. The monkey faced a black box and had to quickly figure out why it hung up.
You first need to understand the system very well, to judge from the symptoms and the situation at the time, what are the most probable, secondary and thirdary causes.
After countless training sessions, he was finally able to basically determine the most probable cause, which is the actual cause.
It's the same with trading now.
Sometimes the program gets lost, and the monkey has to immediately analyze it, and find out through the spider's web whether it's a problem with the strategy, the trading system, or the market or the exchange.
In the way of trade and research and development, over the years, the monkey has been like a thin ice. He tries to do his best without joy, without sadness. When he makes money, he is not very excited. When he loses money, he must find a reason and solve the problem.
Monkey is never afraid to communicate with people. The market is constantly changing, there is no one-time strategy to make money. Monkey feels that there is a strategy and those who stick to it have no future.
In our line, R&D is always on the way.
Translated from the Chinese: 春晓交易门