Need a personal advice about machine learning ...

 

I'm 50 years old ...

Programming tools, polishing my math a little, statistics, machine learning and optimization algorithms, etc etc ...

My estimation is ... I need to spend the nights of my weekdays of my life for 4 years at least to get a robust and working knowledge in machine learning.

Experienced traders who use machine learning in their trading operations ... what would you think ... is it really worth it to spend so much time (when you are 50, yes, it is so much) to study?

At the end of everything, can I hope a significant increase in success in my trading operations?

Documentation on MQL5: Constants, Enumerations and Structures / Trade Constants / Trade Operation Types
Documentation on MQL5: Constants, Enumerations and Structures / Trade Constants / Trade Operation Types
  • www.mql5.com
Trade Operation Types - Trade Constants - Constants, Enumerations and Structures - MQL5 Reference - Reference on algorithmic/automated trading language for MetaTrader 5
 
Stop thinking a buzzword will help you. Learn to code, find a strategy, code it.
 

This is my personal opinion (I'm 55 by the way) ...

If you wish to study machine learning, simply for the fact to broaden your knowledge and challenge yourself, because you find it interesting or fascinating, then by all means, go ahead and do it simply because it is fun, or rewarding at some personal level. If you enjoy programming, logic, maths, or science, then study it for the simple pleasure of being more knowledgable about it.

If however, your objective is to improve your trading profitability, then don't. It will not achieve any meaningful success on its own. Instead dedicate your time to improving your trading skill and knowledge based on the more well known, tried and tested, methods. Keep it simple. Use your own natural "neural network" (your brain), to study the markets and observe them. Focus more on the risk and money management of trading and less on the actual strategies.


Forum on trading, automated trading systems and testing trading strategies

Automating EA Optimization

Fernando Carreiro, 2023.06.06 00:41

To be honest, I don't have much faith in AI driven EAs and I have yet to see one that can outperform a conventional EA (non A.I.).

I'm a follower of the K.I.S.S. method, and prefer simple EAs, especially when they are self-adapting and require no optimisation.

To me, I consider A.I. EAs as equivalent to Rube Goldberg machines, trying to achieve the simplest of tasks.

So in essence, I have never faced your requirements.

Forum on trading, automated trading systems and testing trading strategies

Writing a profitable trading bot

Fernando Carreiro, 2022.08.16 15:54

Based on my own experience, I find that the most important aspect of having a consistent and profitable strategy, is how you handle your risk and your money. I would rate the importance of different aspects as follows:

  1. Risk & Money Management (maximum importance and most essential part of all the strategy rules).
  2. Exit rules.
  3. Entry rules (least important).

Read the following too ...


Forum on trading, automated trading systems and testing trading strategies

EA printing values that shouldn't be there

Fernando Carreiro, 2022.06.10 22:50

As I stated, both at the same time may not be possible!

When I started with MQL, I was already an established software developer with 30 years of experience. But even then, I approached it the way I described. Even with my solid background in C and C++, MQL was not just about coding. It also had to do with the MetaTrader platform and trading concepts, and those were totally new to me.

Many of the strategies I had in the beginning eventually got bashed away by the reality that trading can differ totally from what one thinks it is in the beginning. I came up with very elaborate stuff, only for them to be ultimately thrown in the bin because I was naïve.

As I progressed, I realised that my trusted friend in coding, namely K.I.S.S. (Keep it simple, stupid!), is just as relevant in trading as it is in much of anything else in life. Complex strategies are useless and just distract us from what is really important.

As time went by, my strategies became simpler and my EAs become more refined. However, to my surprise, over time, money and risk management became the central theme of it all. It gradually became clear to me that entry rules were not all that important, but exit rules, risk and money management can make or break a strategy and became paramount for me.

In essences, almost any indicator gives approximately the same signals, given that they all derive those signals from the price action itself. So you might as well stick to the simplest and cleanest of indicators and concentrate on the exits, risk and management of the trades.

With that in mind, I decided that my code should be simple, clean and efficient, and focus on the risk and trade management and very little on the indicators and signals they generate.

This is my advice to you. Learn to code MQL and understand the platform and trading, methodically and properly, because as you do, you might find that your current strategy is going to end up in the trash bin.

NB! My journey with trading and MQL took me years (not months), to get to a level I could be happy with.

 
Alp Duman:

I'm 50 years old ...

Programming tools, polishing my math a little, statistics, machine learning and optimization algorithms, etc etc ...

My estimation is ... I need to spend the nights of my weekdays of my life for 4 years at least to get a robust and working knowledge in machine learning.

Experienced traders who use machine learning in their trading operations ... what would you think ... is it really worth it to spend so much time (when you are 50, yes, it is so much) to study?

At the end of everything, can I hope a significant increase in success in my trading operations?

Yes . But if you want to skip steps you can learn the higher level tools available.

PyTorch for example , its openCL + ML in one package (with an nVidia GPU)

Start there , take a simple strategy ,a silly one even ,export the data build the network in pytorch see what it does ,pull the network back in mt5 let it trade .

Take small steps -since from your words i get that there is no intent to understand why and how (which is very difficult indeed)- and these small steps will help you build an intuition.

Go straight to the goal ,to be able to "squeeze" everything out of an idea with ML .

Fun fact : You can use the custom criterion mode of the tester to see how the fitness function of a genetic algorithm adapts to what you are asking it to evolve.


 
Alp Duman:

I'm 50 years old ...

Programming tools, polishing my math a little, statistics, machine learning and optimization algorithms, etc etc ...

My estimation is ... I need to spend the nights of my weekdays of my life for 4 years at least to get a robust and working knowledge in machine learning.

Experienced traders who use machine learning in their trading operations ... what would you think ... is it really worth it to spend so much time (when you are 50, yes, it is so much) to study?

At the end of everything, can I hope a significant increase in success in my trading operations?

I don't think there is much traders using machine learning in their trading operations. I mean real traders, not gamblers, not marketers, not programmers. I could be wrong, but I have never seen anything like that up to now.

 

Lately I've decided that ... after many years ... I don't want to turn on my computer every morning and see those charts anymore,

But it was a question that keeps my mind busy.

Cleared now,

Thank you all so much ..

After your replies, I am really so happy to be able to spend the rest of my life with music and only reading what I want to read 😊



 

Waste of time for trading price quotes is not predictable, some other data might be predicable though.

But for learing about ai in general "generative deep learning teaching machines to paint write compose and play" is good start.

Everything from variational auto encoders ,energy based models, gan models,normalizing flows models, convolutional neural network, LSTM, Transformers and of course diffusion models haha.

But only with small toy data sets to learn, but  generative models are awesome in general wow.

 
Alain Verleyen # :

I don't think there is much traders using machine learning in their trading operations. I mean real traders, not gamblers, not marketers, not programmers. I could be wrong, but I have never seen anything like that up to now.

Is there any objective data on the share of investment managed by AI? In my opinion, this share is growing, but I do not know how this can be confirmed. And if this growth exists, then any trader needs to understand AI at least to understand the market.

 

I don't think investment entities are required to disclose info about the "how"  so that amount may be well under-reported .

Furthermore if one finds a good set of "weights" for trading they won't be inclined to share it in order to get views or sell it and risk revealing it.

So if we are talking about an individual who cannot trade it due to being in an unfortunate position temporarily , then their best route would be a signal for which they'd need funds anyway.

Were that signal account underfunded it would be more likely for the algorithm to fail at first even if it were optimal due to margin issues.

 

If this source is credible, then 90% of funds use AI.

"A new survey by Market Makers has found that 9 out of 10 hedge fund traders will use artificial intelligence (AI) in 2023 to achieve alpha."

 

I was not talking about big players, but about retail traders which is the audience using MT5 and this site.

AI (machine learning and the like) is also a big buzz word. Seems all what is not done manually is now catalogued as "AI" lool.