AI - BTC (1)

After discussing gold, let's continue to look at some of the research AI has done on BTC over the past period...
The first of course is similar to the gold approach. If multi-agent trading can work with gold, can it also work with BTC? After all, there is more data on BTC...
For this part, I really searched for a ton of data... Almost everything from the MMT API, which costs $499 a month, was used... to develop this part...
It's also based on the multi-agent idea, including ICT / Price action / Liquidity / Scalping, and various other strategies...
The final result is somewhat hard to evaluate...
Although after a month, the PnL barely breaks even... the process was very bumpy, basically a rollercoaster... If you want to make a big profit, you can also lose a big one...
When things are good, the total PnL exceeds $20k (demo account), but when bad, it can retreat from over $20k to just over $6k...
Overall, this BTC setup is still not usable... it needs to be reset and rebuilt... Currently paused, and in the next 1-2 weeks, a new plan will be developed...
But among this, if you look at Figure 1, you'll find that the only agent with a decent win rate and profit factor (PF) is actually two scalping agents...
At that moment, I had a sudden idea: if the frequency of this agent's trades increases from three times a day—Asian, European, and US sessions—to once every 4 hours or even every 2 hours... would the actual results improve?
So I researched another BTC scalping system...
After intensive testing for a few days, it didn't seem to be very effective... (Figure 2)
Burning tokens every 2 hours was a bit too harsh... so I paused it for now.
Therefore, the BTC trading agent is also currently paused... plus, the MMT API paid service seems to be broken these days, and I can't renew the subscription... once it's restored, I will explore other options...
Next, I will talk about other ideas I have for BTC.