After the article on ideas, we have gradually explored some tips for using GPT in quantitative trading through in-depth research with various experts.
So, as before, let's use an example to illustrate.
We also hope that friends with ideas and thoughts can join us in exploring the "Holy Grail" of quantification.
As usual, the latest research results script is attached at the end.
100 days, 3x leverage, 355% return.
Off-topic discussion:
During this period, I have really met many experts... real experts.
① Someone summarized the content about trading ideas in a community by a certain KOL over the past two years, and input it into GPT to form a quantitative strategy. The KOL instantly became a tool.
② Someone used the formed harmonic and PA theories to make GPT create a quantitative script and adjust the parameters, skipping the learning process and participating in graduation exams directly.
③ Someone used GPT to expose various pseudo-trading bloggers.
Adjusting the programming language:
Received a request for assistance from a new friend, but it is obvious that this script is not in a language that can be applied in TV's Pine editor. In this case, we can use the command:
[Please help me modify this script to a version suitable for backtesting on Tradingview]
Let GPT help you with the script language modification.
Regarding the TV loop error after GPT modification:
Usually, GPT does not output 100% correct code. In general, we paste the error feedback to GPT and use the command [I received an error message, please check the code for me, the error is...], and it can correct the code. However, sometimes solving problem A extends to problem B, and then problem B reproduces the situation of problem A. In this case, we can use the command [Please try another solution].
The importance of verification!
After outputting the trading strategy, we need to let GPT analyze the trading script to see if it matches our expectations. At this time, we can use the command [Please analyze the above trading script and analyze its trading logic].
About the intelligent use of AI:
A complete set of trading logic cannot be fully reflected through TV. In other words, a mature trader will not use TV backtesting for trading. PS: TV is a better display tool, which is why many people who provide signals love to use TV.
Sometimes the strategy performance is not satisfactory, but don't give up. Give it to GPT! We let GPT analyze the data!
How did we get the information to complete the above item?
Command: [The profit of this strategy on the 15-minute candlestick is not good, but I think its ability to capture trends is still good. Maybe the closing strategy is not very good. Can you help me adjust the trading strategy based on the feedback?]
After verification, GPT can recognize the content copied from the table without further adjustment.
The Holy Grail is achieved by chance.
There are many examples that tell us, including many invention experiments, that sometimes it's just helping a newcomer who doesn't know how to use TV backtesting. I didn't expect to come up with such an excellent strategy...
Public strategy, hoping for everyone's discussion and optimization.
Please teach me how to do it, GPT-ETH-30Min/1H Strategy - #ETHUSDT TradingView https://cn.tradingview.com/script/faqyIUkc/
Finally, people say I am very polite and even respectful to AI... I have indeed become very respectful since seeing this picture for no reason.