True AI Trading, the divergence is actually not in the model,

And in "whether to use history".

Category One: Instant Judgment AI

Do not train history,
No long-term backtesting.
All data is accessed in real time via the API.
Looking at:
What happened at this moment.
Speed is everything.
Category Two: Short-term Pattern Type AI

can look at history,

But only look at a very short section.
Three minutes, five minutes, fifteen minutes.
It's not the future that is judged,
but whether the trend has taken shape.
Category Three: Mid-term Structural AI
Not caring about a single K line,

Pay attention to a trend.

History is used to judge rhythm, structure, and state.
instead of the point.
More is about whether it goes smoothly or not.
Category Four: Long-term Evolutionary AI
Truly rely on history.
What I'm looking at is:

Is public opinion accumulating?

Is consensus forming?
Is the funds being migrated?
The longer the time,
The signal becomes clearer.