宽客秀

In the beginning of all things, the great way is simple, but it becomes complex through evolution. - Laozi, "Tao Te Ching"

With the development of the times, trading strategies have become increasingly complex, with the use of artificial intelligence for stock selection and intelligent investment advice. However, the effectiveness of a trading strategy is not always proportional to its complexity. Many truly effective trading strategies are simple yet effective. They have stood the test of time, remaining steadfast amidst the storms of the stock market and the rise and fall of the futures market. In my opinion, the grid trading strategy is one such simple and effective trading strategy.

The public account "Long-Term Index Investment" has published a series of articles on grid trading, two of which I believe are very well-written (see references at the end of this article). I have drawn heavily on the ideas presented in these two articles for the empirical study in this paper.

1. Investment Targets and Grid Settings#

1. Investment Targets:

CSI 300 ETF (510300.XSHG). The reason is simple: the choice of investment targets for grid trading has two requirements: they should frequently oscillate rather than move in one direction, and they should never die. (In fact, I think futures contracts are more suitable, but I lack the funds for live trading...)

2. Grid Settings:

From dense to sparse, three types of grids are set: small grids, medium grids, and large grids. The specific approach is as follows:

• Calculate the average and standard deviation of the opening prices of the CSI 300 for the past 90, 300, and 900 trading days (reset once at the beginning of each month).
• Use the mean plus or minus several standard deviations to determine the boundaries of the grid. More intervals are used for small grids, while fewer intervals are used for large grids.

For this empirical study, 30, 14, and 7 grid intervals were respectively assigned to small, medium, and large grids. Taking October 12, 2020 as an example, the grid intervals for the small grid are:

[-11215/3089/3246/3403/3561/3718/3875/3954/4032/4111/4189/4252/4299/4347/4409/4457/4504/4551/4598/4661/4708/4755/4818/4897/4975/5054/5133/5290/5447/5604/20224]

The grid intervals for the medium grid are:

[-9798/2690/3037/3384/3557/3731/3852/3974/4078/4182/4303/4425/4598/4772/17954/]

The grid intervals for the large grid are:

[-11645/3017/3403/3673/3905/4175/4561/19224]

From the above, we can see that:

1. The intervals of the small grid are roughly within 100 points, the intervals of the medium grid are around 100-200 points, and the intervals of the large grid are around 300 points. The average daily fluctuation of the CSI 300 is about 80 points, which is roughly consistent with the range of the small grid.
2. The lowest point in the past 5 years is 2935, and the highest point is 4901. The lowest boundary point in the grid is 2690 (medium grid), and the highest boundary point is 5604 (small grid), which covers the extreme situations in the past 5 years.

2. Trading Strategy#

First, the capital allocation for the small, medium, and large grids is 25%, 35%, and 40% of the total capital, respectively.

Second, for the small, medium, and large grids, the portfolio is adjusted at a frequency of once a day, once a week (the first trading day of the week), and once a month (the first trading day of the month), respectively. For each grid, a corresponding capital weight (ranging from 1 to 0) is assigned, and the position is allocated based on the current price of the CSI 300 within the grid range. Taking the large grid data on October 12 as an example, the corresponding capital weights are [1, 0.7, 0.49, 0.34, 0.24, 0.17, 0.12]. This means that if the price of the CSI 300 is within the range of [-11645, 3017], a full position is taken (capital weight of 1), and if the price is within the range of [4561, 19224], a position of 12% of the corresponding capital is taken.

In addition, additional trading conditions are set for the small grid, such as:

• If there is a significant gap up or down at the opening of the day, no trading is conducted on that day.
• If the market has been rising or falling for the past 3 trading days, no trading is conducted on that day.
• If there was a significant increase or decrease in the previous day's closing price and the market opens with an increase or decrease, no trading is conducted on that day.

3. Empirical Study#

Based on the K-line trend of the CSI 300, the following backtesting periods were selected for empirical study, with a backtesting capital of 1 million RMB. The results are shown in Table 1:

• V-shaped period: January 1, 2018 - December 31, 2019
• Inverted V-shaped period: April 1, 2016 - January 31, 2019
• W-shaped period: August 1, 2015 - June 30, 2019
• Inverted W-shaped period: March 1, 2019 - March 31, 2020
• One-way upward period: February 1, 2016 - September 30, 2020
• One-way downward period: January 1, 2018 - December 31, 2018

Table 1: Backtesting results for different periods

According to the results shown in Table 1, the grid trading strategy has higher returns than the benchmark, but the annualized return is surprisingly low, even lower than that of a money market fund. This raises the question: why is the return so low?

Taking the empirical study data for the W-shaped period as an example, the profit is a little over 30,000 RMB, which is not much compared to the initial investment of 1 million RMB. Upon closer examination of the idle cash data, it is found that a large amount of capital remains idle (see Figure 1 and Figure 2). Figure 1 shows the histogram of market value, with the majority of the market value being below 500,000 RMB. Figure 2 shows the histogram of cash, with the majority of idle cash being above 500,000 RMB. In other words, a large portion of the capital allocated for the grid trading strategy remains idle. This explains the low annualized return. If we exclude this idle capital from the denominator, the return will be much higher. This highlights another advantage of grid trading compared to simply investing in an index - it requires less capital to achieve index returns.

Figure 1: Histogram of Market Value

Figure 2: Histogram of Cash

To address the issue of idle capital, we made a simple optimization to the strategy by investing the idle capital in the YinHua Daily Income Money Market Fund. The backtesting results after optimization are shown in Table 2:

Table 2: Backtesting results for different periods (after optimization)

According to the results shown in Table 2, we can summarize the advantages and disadvantages of the grid trading strategy as follows:

1. The capital efficiency of the grid trading strategy is not high, and the annualized return is not very high. The return depends on the starting and ending points.
2. Except in the case of a one-way upward market, in a volatile market or a one-way downward market, the grid trading strategy for the CSI 300 can achieve higher returns or avoid less losses compared to simply investing in the CSI 300, especially when the starting and ending points are the same, the grid trading strategy can generate excess returns.
3. In a one-way upward market, when the stock price is in a rapid upward trend and breaks through the upper limit of the grid, the grid trading strategy may face the awkward situation of selling chips too early.
4. The returns in V-shaped or W-shaped periods (ending at a high point) are higher than those in inverted V-shaped or inverted W-shaped periods (ending at a low point).
5. The maximum drawdown is not very large, generally between 12% and 14%. The risk is relatively controllable.

References:

• Wave Strategy. Grid Trading Part 2: Grid Strategy Basics/Version 1.0 Long-Term Index Investment, May 6, 2019
• Wave Strategy. Grid Trading Part 3: Advanced Grid Strategy/Version 2.0 Long-Term Index Investment, May 26, 2019