(Friends who are interested in the "Probability Trend Trading System" can contact me on Weibo and share with you a complete trading system theoretical framework mind map.)
The previous four articles on how to build a trading system explained
1. Trading system framework
2. Develop 3 key points of the system and understand the technology
3. Using a short-term strategy to determine the logic of profitability and the direction of thinking about subsequent issues
4. How to perform backtesting
This article still uses a short-term trading strategy to specifically explain how to backtest and optimize the trading system.
Figure 1 shows the backtest data of the trading system from January 15, 2023 to February 24, 2023. The backtest content includes the following according to the characteristics of the trading system: signal appearance time, entry time, trading direction, profit and loss, single risk amount, profit and loss ratio, and total amount. The capital curve is also drawn.
In a month, there were 34 trading signals, 18 of which were selected through judgment conditions. In the end, there were 14 profit-taking transactions and 4 stop-loss transactions, with a return rate of 115%. This is equivalent to doubling the amount in a month. The capital curve is shown in Figure 2.
Of course, this is backtest data. As mentioned in the previous article, the backtest performance of the system is its upper limit, that is, the result without considering the influence of factors such as mentality, environment, and execution. And the backtest data should preferably span two rounds of bull and bear markets and cover most of the market conditions to illustrate its adaptability and feasibility. Only then can there be sufficient data support for system optimization.
Next, we need to optimize the system based on the backtest data. There are many optimization directions, such as:
1. Entry logic
2. Exit logic
3. Stop loss amount
4. Trading frequency
etc.
Every time an optimization plan is proposed, the historical market conditions must be backtested again, and then the backtest performance must be compared with the performance before optimization to draw a conclusion on whether to modify it. This is a long and time-consuming task.
Taking the optimization of stop loss as an example, we can observe from the backtest data in Figure 2 that among the 14 profitable transactions, 10 have a profit-loss ratio of no more than 1. So if you want to increase the profit-loss ratio, one way is to reduce the stop loss space. Then you can try to adjust the previous stop loss logic, such as reducing it from 2ATR to 1.5ATR. Then backtest the same market again. It is very likely that the winning rate has become lower, but the average profit-loss ratio has become higher. Whether the profit can grow in the long run depends on the actual backtest results.
We will continue to demonstrate optimization cases of this short-term trading system in the future.
