Quantitative strategy development, high-quality community, trading idea sharing and other related content
Hello everyone, today we will share the gp08 strategy. The long-awaited one has just come out. Due to xxx reasons (I can’t say it, if you are curious, you can chat with me privately), we started in September. The strategies we will share later, whether backtesting or real trading, are based on the Bigquant platform. Factor research will be done later. We will provide you with a self-developed system.
As usual, let’s share the actual performance first, as shown in the figure below:
I won’t show you my recent positions because there are many customers who have made firm offers, so as not to reveal their position plans to the firm bosses.
1. Strategic background and logic
In the past few issues, we have shared technical multi-factors, reversal factors, financial factors, etc. In this issue, we will cleverly integrate them with small market capitalization to construct a small market capitalization enhanced and stable version of fundamental factors that can be directly implemented. We will publish 2 versions of this strategy, one aggressive version and one low drawdown version.
This version of the strategy also uses two stock selection logics, screening factors and sorting factors. Please refer to the previous article for detailed explanation. Let's first give a brief introduction to the visual quantification tool, its specific usage and logic. I will explain it during the live broadcast, as shown in the figure below:
The picture above is the simplest form of process based on the "visual blank" strategy. Roughly divided into 3 points:
-
code list
This refers to the stock pool
-
Input feature list
The characteristics here are "factors".
-
Basic (derived) feature extraction
Factor data extract merged list according to cross section
-
backtest
The overall process uses the developed logic modules to conduct strategy development, backtesting and research in the form of visual building blocks.
In the traditional process, we need to clean and organize our stock data, and for the accuracy of backtesting, we often need to pre-weight the data. Then we have to calculate various factor data, and then cross-section the factor data and stock data. All groupby is good. Just adding bugs to the content of the above three sentences will cost you at least a month. I am speaking based on the standards of an expert. This does not include subsequent performance calculations, performance indicators, etc.
Therefore, this platform has perfectly solved the above problems and has reserved hundreds of factors for everyone, as well as pre-calculated factors of financial valuation indicators. As shown below:
In the input feature list, if we enter a keyword at will, the factor data expression that has been calculated for you will pop up. When we run it through the node module, we can view the data operation results, as shown in the figure below:
In the filtering module, we only need to use the module visualization function to remove "ST", "*ST", Kechuang and Beijing Stock Exchange stocks, etc. As shown below:
The most critical and important thing, what I like most is that the AI module is also integrated into the visualization. I won’t go into details here, I will explain it further during the live broadcast.
Then someone will ask, overall, do you need to write code? In fact, it is still needed because it involves some logic. For example: in the input feature list, you need to write some code to reflect the factor data. In backtesting, initialization and k-line logic processing, etc., you need to do some processing, as shown in the figure below. :
2. Strategic performance
Low drawdown version
radical version
Strategy backtesting cycle: January 1, 2019 - September 8, 2023
Handling fee 30,000
Due to differences between platforms, the backtest performance is subject to the BIGQUANT version! ! !
This strategy is only used for learning and communication, and investors are personally responsible for profits and losses in real trading.