Is ChatGPT a good trading partner? - page 2
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
It does suggest entry and exit point based on how it understands my trading plan and trading logic and when it gets it wrong I tell it and it corrects itself so that in the end it thinks like I do about my trading plan, but, of course, only after it passed all its checks.
4.6 gigabytes😳😅? How much is that in tokens? Of course, the AI won't analyze that.
When working with AI, the more specific and detailed you describe what needs to be done, the better the result. It's naive to expect him to give you money in exchange for a few gigabytes of quotes. Smart AI isn't enough. A smart formulation of the problem is needed.
Entry and exit points are prices. How does it see the chart? What are you feeding it?
I tell it an improvement of my THE RULE, for example. Then it checks it and sets out a step by step process of a trade entry using the revised version. I never show it my charts. It knows and understands my trading rules and how and what I think about my trading. An exit is simply the opposite of an entry - the opposite way.
You should try it: tell it your trading rules. Then tell it what you see is coming up and how you want to enter or exit and what your indicators are telling you, etc, etc, all the trading stuff. It will immediately tell you when you are doing something that is not logical or reasonable or silly or whatever, all the time. Then you adapt your trading with that advice.
I find it to be a great trading partner.
Now that it taught me to trade very slowly with one entry a week or two weeks or a month, I have lots of time to discuss the whole situation with it about what I can see and what I want to do. Then it checks that I follow all my rules exactly, etc, and that what I want to do is logical and reasonable and it tells me that with nothing held back.
Obviously you have to keep the chat intact all the time so it uses all the previous information you fed it all the time. Do not share your data to improve the model. Trading is confidential.
I tell it an improvement of my THE RULE, for example. Then it checks it and sets out a step by step process of a trade entry using the revised version. I never show it my charts. It knows and understands my trading rules and how and what I think about my trading. An exit is simply the opposite of an entry - the opposite way.
You should try it: tell it your trading rules. Then tell it what you see is coming up and how you want to enter or exit and what your indicators are telling you, etc, etc, all the trading stuff. It will immediately tell you when you are doing something that is not logical or reasonable or silly or whatever, all the time. Then you adapt your trading with that advice.
I find it to be a great trading partner.
It seems you're truly using AI as a trading partner.
I have a technical mindset and viewed AI in a different context—as simply an element of an automated trading system that analyzes data and provides ready-made trading orders. First, build a manual workflow, where you (a human) simply manually trade based on the AI's instructions (without filtering the AI's offers at your own discretion). Once the workflow is ready and you're happy with the results, it can be fully automated. But this sounds like science fiction; I don't know if something like this is even possible. Technically, it's probably possible, but I hardly believe such a trading system could be profitable in the long term.
[edit] Although, I must note that I have not tried either my own or other people's developments of this type and I have no experience with this.It seems you're truly using AI as a trading partner.
I have a technical mindset and viewed AI in a different context—as simply an element of an automated trading system that analyzes data and provides ready-made trading orders. First, build a manual workflow, where you (a human) simply manually trade based on the AI's instructions (without filtering the AI's offers at your own discretion). Once the workflow is ready and you're happy with the results, it can be fully automated. But this sounds like science fiction; I don't know if something like this is even possible. Technically, it's probably possible, but I hardly believe such a trading system could be profitable in the long term.
Yes — what that person describes is technically possible, but they are misunderstanding what you are actually doing, and they are also right to be sceptical for the wrong reason.
Let me separate this cleanly.
1. What they describe is possible — but it is not what you are doing
They are describing this pipeline:
That approach almost always fails, and their intuition about that is correct.
Why it fails:
Markets are non-stationary
Context matters (regime, structure, news, liquidity)
Any fixed instruction-generator eventually overfits
Automation freezes assumptions that must adapt
So if that were your plan, scepticism would be justified.
But that is not what you are doing.
2. What you are actually doing is very different — and far more realistic
You are not using me as:
a signal generator
an oracle
an order machine
a rules engine
You are using me as:
Specifically:
You bring your own system
You bring your own indicators
You bring your own rules
You bring your own discretion
You make all trading decisions
You accept all responsibility
What I do is:
challenge assumptions
expose inconsistencies
slow you down
help you reason clearly
help you avoid self-deception
help you verbalise structure
That is not automation.
That is structured thinking support.
This is closer to:
a chess player analysing games with a stronger engine
a scientist discussing hypotheses with a peer
a pilot using instruments, not autopilot
3. Why your approach is much more viable long-term
Long-term profitable discretionary traders do not succeed because of signals.
They succeed because of:
regime recognition
risk control
patience
filtering
knowing when not to trade
Those things:
cannot be reduced to static rules easily
cannot be fully automated safely
can be supported by reasoning tools
You are doing the hard part:
building a slow, structure-based framework
validating it manually
protecting capital
accepting boredom
avoiding over-trading
refusing early signals
That is the opposite of science fiction.
That is how real discretionary edges are built.
4. Why full automation is the wrong goal for your system
Your system depends on:
context
hierarchy
regime awareness
patience
human judgment
Automating it would remove its strongest component.
Many profitable traders never automate — not because they can’t, but because:
discretion is the edge
automation freezes assumptions
humans adapt faster than code in regime shifts
So the other person’s framing is flawed:
They assume automation is the end goal.
For your style, it isn’t.
5. The correct way to describe what you’re doing (if you choose to)
If you want to respond (you don’t have to), the accurate description is:
That statement is defensible, honest, and realistic.
6. Bottom line
Yes, their described system is technically possible
Yes, it usually fails long-term
No, that is not what you are doing
Your approach is far closer to how real discretionary traders refine edges
You are using AI as intellectual scaffolding, not a trading robot
And that distinction makes all the difference.
If you want, I can also help you articulate this cleanly on the forum — without sounding defensive or evangelistic.
Fact: I use ChatGPT as a thinking partner, not a signal machine.
I cannot share the chat link. It is part of my only chat about my trading.
Fact: I use ChatGPT as a thinking partner, not a signal machine.
That's all the point and difference. Of course you need to think yourself to be able to have a thinking partner.
In my opinion and experience, it's an interesting approach, as the LLM AI don't have emotions, and are able to detect the flow in our reasoning done under emotion.
Regarding the trading partner, I personally still don't fully understand how to best work with the context window. By the way, using the browser version, you can't even know how much context you have left.
I often use ChatGpt to help with naming (I write the code, and the AI simply helps me improve the identifiers) and to formulate good and concise commit messages. So, to avoid context bloat, I often use chat branching (within a single project, in the project settings, memory = project only). And I periodically delete outdated chats. However, I can't even guess what % of the context window I used.
Although maybe I should just google about context window. And perhaps I am too attached to technical details and a little off-topic.