Can ChatGPT develop EA? PART.3

I don’t intend to series-ify this, but…

This is the third article about whether EA development is possible with ChatGPT.

It’s impossible to develop an EA with an expected value greater than 1 using ChatGPT.

To ChatGPT

I want you to “develop an EA that has a positive profit”

and I still receive the following kind of response.

Well, obviously.

If an approach like this could develop an EW >1 method, EA programmers would cease to exist.

However…

There may be cases where it can be used as an auxiliary role to assist with programming.

In other words, you can’t develop the entire EA, but you can ask for just a portion of it.

For example,

“a function that returns TRUE when the last two candlesticks form a bullish engulfing pattern”

“a function that returns TRUE when two moving averages have a golden cross”

“a function that returns TRUE when the entry price goes down by 10 PIPS”

and similar things.

For this kind of partial development,

it’s relatively simple, so it might be okay to rely on ChatGPT.

For reference

Below is a function to check whether an overbought/overhigh resistance line is formed

and had ChatGPT create it.

The instruction to ChatGPT is"Prompt Engineering"is also called

This itself is becoming a skill, a specialized field,

so it’s also difficult…

The response varies depending on what instructions you give,

but this time I instructed it as follows:

Then…

Aha, a fairly decent function was created and returned.

Of course, using this as-is carries high risk.

You need to verify whether it behaves as intended

assuming you’ll determine its behavior yourself,

so there is value in using it as an initial framework or template.

For a beginner in EA, using it like this is acceptable.

By the way, in my case,

If I’m going to issue instructions like this to ChatGPT,

it’s faster and more reliable to program directly toward the MetaEditor from the start.

Going through ChatGPT just adds effort…

Changing the subject…

From late April, I consecutively released four new EAs, and

so far all of them have generated positive profits.

What’s especially performing well is

these two.

Recently, due to expectations of an interest rate hike in the UK, the pound has been rising,

the EA above applied to GBPJPY seems to have ridden that current well.

By the way, of course these two EAs

have not used ChatGPT at all.

They are EAs I developed years ago, completed, and actually run.