Greetings !
Today I was once again asked about the AI principles that many people use in their MQL software.
At the request of my users, I decided to write a short article about this.
There are many reasons why more and more MQL systems have started to acquire the mark of being made with AI, or supposedly having AI built in.
The question is, are you ready to entrust the management of your car to artificial intelligence?
Of course not !
Let's be honest, some quite living intellects (living people) still can't cope with this!
Well, almost some... Just a little bit
The main problem with any AI is training and the resources for it.
In real life, you would need a supercomputer to train such a network of artificial neurons.
It takes a person from 2 to 5 years to understand this, just to go poop on their own =D
During this time, any market will have changed long ago and we will have to undergo training AGAIN.
Of course you say we can use ready-made AI, ChatGPT / OpenAI / DeekPik DeepSeek
But how to train him correctly? Just make him look at the market and invent the behavior of transactions?
This is complete nonsense and it will cost you very, very dearly.
Therefore, usually inside any system there is a primitive algorithm based on indicators.
Next, we add the word AI (OpenAI) to the system description and create the ability to write messages with ChatGPT, for example
These are all just candy wrappers and pretty packaging, nothing more.
Such systems exist, but they are very expensive and are constantly being trained, which takes up a huge amount of resources.
But even they make huge mistakes and blunders.
Therefore, do not be fooled in any case, it is extremely difficult to create, especially on MQL.
There is also a second reason for such systems. This is excessive over- optimization.
When you get too perfect results during AI testing.
Typically, such systems simply remember the best entry and exit points into the market.
This is a regular black box where you will not be able to manage the system and optimize other currency pairs.
I have always been against the black box, it is very dangerous to buy such systems.
Tomorrow they may simply stop working and you won't be able to do anything about it.
There are many such cases in MQL5.
This is why we NEVER use AI or Black Box.
No AI + No black box + No Martingale
Only open parameters, only clear algorithm of work and behavior of the code.
We also do not allow OVER optimization in the market.
All our Systems are Optimized until the beginning of 2024, everything that happens after this date
It is a continuation of trading, so our settings are not perfect.
These settings can always be better for the current year, but this will already be OVER optimization.
It's like walking along the very edge of a cliff, yes, the wind blows fastest there, but it's always so easy and simple to fall.
Therefore, AI always has a high market over-optimization.
So it might work for a while, then the market will change slightly and the whole AI black box has to be retrained.
Thank you for reading this article to the end, I hope it wasn’t boring!
