#Training My AI Trader Day 9

So far, the role of rag has become quite clear. The model no longer needs to remember every branch situation; it will first judge the market with a naked K-line, looking for whether it's in consolidation, trending, and how long the trend has lasted. Then it will find relevant experience from Brother Jie’s knowledge base to match the current market conditions. Personally, I think it’s pretty good. Yesterday, I asked it to analyze the market, and it told me not to short during the middle of the 4h Vegas to catch a rebound, but to short after confirming a breakdown. Both points can be found in Brother Jie’s knowledge base, so it’s clear that it’s starting to imitate Brother Jie’s way of thinking, which is very good!
However, the big issue is that its thinking is still not deep enough. Out of 400 pieces of knowledge, it only hit about 14, indicating that my indexing is still not good enough. This needs further optimization.
Currently, I think a very good aspect is that AI has the ability to read images and truly think about the market. What’s not yet done is further refinement of Brother Jie’s database. I’ve now extracted about 400 pieces of experience from Brother Jie’s 38 videos and created a small vector database. But the problem is that the extraction is still not enough.
Next, I need to abstract the database content into a real mathematical model, such as adding visualized charts for concepts like breakouts, false breakouts, and FRVP. I believe this will further enhance AI’s understanding of these concepts.
This is also my strength—modeling and analysis. I will continue practicing in this direction. I believe that as the vector database improves, its capabilities will become stronger and stronger.
Below is AI’s trend judgment on ETH from last night. It’s clear that it has started to imitate Brother Jie’s thinking process, but the reasoning is still not enough.
• ETH workflow has already started. I will wait for the local snapshot to return, then simultaneously check the 4h/1d search results and the recent K-lines to directly determine whether it’s in a trend leg, a false breakout, or just noise.
• The two cycles of ETH are now of the same structure: 4h inside a tunnel, 1d near the top of a box. The next step is to confirm whether this has already formed a “completed daily false breakout.”
Action plan:
- Do not chase shorts now; wait for a rebound to short.
- Prioritize treating 2116-2150 as a rebound short zone: 2116 is near the upper edge of the 4h tunnel, and 2149.95 is the daily box top.
- The first target is 2063, then 2053; if it continues to lose support, then 1973.
Finally, I want to thank Brother @kdbskak49557 for his strong support.