Before encountering flush iFind, Starry Sky Jun learned Python for a while, and then used several free databases to crawl data, realized the automatic storage of financial report data and transaction data, and then formed the desired analysis report.
Later, it was found that a website engaged in quantitative strategy solicitation, so I went to register an account and systematically understood the current quantitative situation.
The so-called quantification is automated trading that relies on computer technology. About 70% of the trading volume of US stocks comes from quantification, A-share data is not very transparent, and it is optimistic that about 40% of the trading volume comes from quantification.
Public information shows that half of the top ten known private placements in China are quantitative.
The other half, hey hey, Starry Sky Jun believes that quantification 3.0 is used (what is quantization 3.0, see breakdown below).
It should be noted that quantification is only a way of trading, and what makes money is the logic of the model, not the way of trading, and whether it is quantified or not has nothing to do with it. The right logic and model, whether quantified or the human brain, will make money; incorrect logic and models, whether quantified or the human brain, will not make money.
In the development of history, quantitative technology has undergone fundamental changes, and the current cutting-edge neural network-based quantification, and the early quantification, are no longer a species at all.
From the perspective of practical operation, quantitative is more suitable for high-frequency trading, and qualitative investment in the human brain is more suitable for low-frequency long-term price investment.
First, quantification 0.1, the age of formulas
Many stock trading software will have some formulas, some are the warning points provided by the software, and some are the buy and sell points defined by the stock speculators.
This is the simplest quantification, and the logic is clear.
In Python's quantitative formula, there are many such buying points and selling points.
This is the quantization of the primary order, but it is not yet quantified, and the Starry Sky King defines it as quantified 0.1.
Second, quantification 1.0, the factor era
It should be said that the vast majority of investors who are biased against quantification are still stuck in the factor era.
What is a factor?
It is clear that it is nothing more than an if statement.
However, there is a problem in China's financial market, in addition to the financial engineering graduates, most of the finance and economics majors are literate and lack of scientific thinking.
The cognition of quantification lacks scientific thinking.
Those who engage in financial engineering are making a lot of money in a muffled voice.
Xingkong Jun secretly paid attention to a quantitative association and found that the top talents who are now studying quantification are basically the most high-end high-tech talents in Qingbei.
Their study of factors is no longer so simple.
For example, here's this: correlation analysis of Weibo sentiment and stock market volatility.
They not only quantify stocks, but even quantify social media such as Weibo and Douyin, and use big data to directly and automatically sort out factors related to stock prices...
Third, quantify 2.0, the era of the cloud
If computing in the 1.0 era mainly depends on servers, it will start to go to the cloud in 2.0.
The advantage of going to the cloud is: unlimited computing power.
As long as Xu Xiang can express his investment logic, he can land on the ground, and then exhaustively lift and go to the clouds.
Fourth, quantification 3.0, the model era
Interestingly, in the era of Quantization 3.0, there may be quantifications that are not quantifications.
How to understand it?
As long as the investor's investment logic is stable and repeatable, then the investment model he expresses is quantitative.
Whether or not he is connected to the computer, even if he just uses the human brain to invest over and over again, it is also a kind of quantification.
Why define this mindset as quantification?
It's because the new era of quantitative 4.0 is here: artificial intelligence.
Fifth, quantitative 4.0, artificial intelligence
There was a rumor that Alpha Dog ran to A shares to do quantification, and then exited with a loss.
In fact, this is a rumor, on the one hand, Alpha Dog has not come to A shares; on the other hand, neural network artificial intelligence based on Similar Logic of Alpha Dog is shining in the US stock market.
If 1.0-3.0 is an investment strategy that simulates and replicates people, then 4.0 is another way to play: beyond the human brain and independent thinking.
Taking alpha dogs playing Go as an example, when it killed all humans, including the go genius Ke Jie, Ke Jie said: I feel that my whole body is trembling, really, cold trembling. I couldn't control my emotions any longer, so I rushed out of the game room and found a corner where no one was crying. Because of the upcoming 3:0, such an ending is too desperate for me.
Why is that?
Because the late alpha dogs no longer use the human idea of playing Go.
This is the wonder of neural network machine learning, you just have to tell it the target, it can learn on its own. Relying on the world's strongest computing power, it can review all the recorded stock speculation logic in human history in the shortest possible time, and then learn independently.
I'll give you another chestnut, there's an emerging profession called artificial intelligence trainer.
For example, the trainer tells the AI that this is a human face, that is also a human face, and this one is not.
Slowly, the artificial intelligence began to learn on its own, and then it began to learn to draw human faces.
For example, the left side of the following figure is the landscape depicted by people, and the right picture is the picture "drawn" by artificial intelligence according to the understanding of "oneself".
In fact, even without input, artificial intelligence can draw the picture that "wants" to express through independent learning.
(Tsinghua Artificial Intelligence Team 2018 works, artificial intelligence independent painting)
What am I trying to express?
Every click, buy, and sell operation of each stock speculator is recorded by artificial intelligence, and then based on this to learn independently, and all shareholders are treated as artificial intelligence trainers, what will be the final quantitative transaction?
Let's take an example:
Flush has hundreds of millions of users.
Flush has made great breakthroughs in the research and development of intelligent speech, natural language processing and other technology applications, and two papers published in the field of speech recognition have been included in INTERSPEECH 2020, the top international conference on speech processing;
The reading comprehension team participated in the global authoritative competition of machine reading comprehension SQuAD2.0 and won the third overall ranking and the first place in the single model;
The natural language processing team participated in DSTC9, the top competition in the field of global dialogue system technology, and won the first place in the cross-language dialogue status tracking task.
Interestingly, SQuAD 2.0 this competition, hKUST iFLYTEK, Ali Damo Academy, Google and Microsoft Asia Research and other well-known institutions have participated, an Internet securities company took the top three and single first, can you imagine the ambition behind it?
Let's take a look at the projects published by Flush in the annual report: