Looking for Yoda

Pedro Severin
1263
Pedro Severin  

Hi all!!

As funny as the topic might sound, its true. I'm looking for a "master Yoda" thats willing to teach me. I'm pretty new to MQL5 and trading, but I'm full dedicated at this (~15 hrs/day).

That huge number of hours a day it's because I needed to learn MQL5 and understand what things do, some basic strategies and concepts. 

So far, I have developed some strategies (not very good ones) and translated EA's from MQL4 to MQL5.

What I'm looking for is to learn more coding and strategies for profitable investment.

I have some EA's that are "profitable" that uses no SL, but I feel that not using any kind of SL is like "click and pray" for the market to go our way (mostly it does with this) and I also feel that the strategy itself isn't that robust.

I'm also aware that its no good to use very tight stops because it's necessary to give the market some room to breath.


So, to sum up I'm really willing (and commited to) learn more coding and strategies for this. In return for the community, I will share my results and any kind of idea or EA that I might come up.

Thank you all and have a great 2019! 

:)

PD: Since I'm new to this, I always wonder if it is possible make a living through trading or it is an expensive hobby with some benefits time to time? 

Icham Aidibe
11958
Icham Aidibe  

Then you're able read and write the MQL language. That's the easiest step.

Building a strategy is the greatest part.

If those strategies you ripped were really profitable, you wouldn't have get the code for free; particularly if most are without SL, that's not the way to do ...

So ...

Then you have to take the time to, to get inspire, to imagine ways to proceed, to test, to reimagine, to retest, to think, to erase, to restart, to adapt ... 

Pedro Severin
1263
Pedro Severin  
Icham Aidibe:

Then you're able read and write the MQL language. That's the easiest step.

Building a strategy is the greatest part.

If those strategies you ripped were really profitable, you wouldn't have get the code for free; particularly if most are without SL, that's not the way to do ...

So ...

Then you have to take the time to, to get inspire, to imagine ways to proceed, to test, to reimagine, to retest, to think, to erase, to restart, to adapt ... 

Hi Icham, thank you for your (funny) advice 😂.

Where should I start looking? Books? What should I pay attention while developing a strategy? Price action? Which indicators should I look at? Price distribution, like normal or beta? Probabilities? Fibonacci? I feel a little bit lost... 

For example, do you have a strategy? How did you created it? What did you looked up? Sorry if this is like a personal question, I'm just looking to learn and become a good trader/coder.

Regards, 
PS
Icham Aidibe
11958
Icham Aidibe  

There's standard strategies you can get inspired from all over the web & books : these are often basics, like how to use a moving average, or a macd, but it's not enough ; for most of these ones, you don't need to code, there's wizard and automatic EA makers.

But you can use a macd or a moving average as a starting point, and perfectionnate it ; there is a real effort to produce. 

Skills in MQL, indicators and some well known techniques described (hedging, trailing, martingale etc ...) aren't alone enough to build an EA, they expand perspective & possibilities at your disposal to reach your goal, they could be assimilated as tools but the definitive "shape" of your EA will depend of your creativity & personal touch : the more vocabulary has a poet, the richer his poetry.

One could light up a joint, get an idea, draft it and retire the following week or month - while another one can spend his all life making MACDSample-like EAs. 

One could ignore everything about MQL's language and have a great strategy - while another one could be a code guru with a MACDSample-like strategy. 

Finally we could conclude there's no Yoda for that :)


Pedro Severin
1263
Pedro Severin  
Icham Aidibe:

There's standard strategies you can get inspired from all over the web & books : these are often basics, like how to use a moving average, or a macd, but it's not enough ; for most of these ones, you don't need to code, there's wizard and automatic EA makers.

But you can use a macd or a moving average as a starting point, and perfectionnate it ; there is a real effort to produce. 

Skills in MQL, indicators and some well known techniques described (hedging, trailing, martingale etc ...) aren't alone enough to build an EA, they expand perspective & possibilities at your disposal to reach your goal, they could be assimilated as tools but the definitive "shape" of your EA will depend of your creativity & personal touch : the more vocabulary has a poet, the richer his poetry.

One could light up a joint, get an idea, draft it and retire the following week or month - while another one can spend his all life making MACDSample-like EAs. 

One could ignore everything about MQL's language and have a great strategy - while another one could be a code guru with a MACDSample-like strategy. 

Finally we could conclude there's no Yoda for that :)


Thank you for your words. Honesty, I don't aim to be a MQL guru, but I want to know enough to code by myself. I think its more important to be able to create a strategy or, at least, the starting point of one and that's what I'm seeking right now.

There are so many topics and so many ideas that I really don't know where to look right now. What I need (or at least I think that I need) is focus or a pointing direction.


For example, I thought of a strategy that goes like this:

- In a 5 min char, put 2 moving averages to see trend (I started with 2 DEMAs, but currently I don't know if it is a good choice).

- Go to 1 min chart and look for good buy/sell opportunities depending on the 5 min trend.


The idea is this: the 5 min candle have the information of 5 1 minute candle, so on average if there is an uptrend in 5 min chart, you could probably see some good pullbacks on a 1 minute chart.

I coded this (one of my first things I tried) and it wasnt thaaat good... In fact, it wasn't at all. But I don't know if it was because my idea is wrong or my coding skills were poor back then.


Please, let me know what you think :)


Regards,

PS

Pedro Severin
1263
Pedro Severin  

Another thing that I do not understand very well is the optimization of an EA. I mean, let's say I optimize an EA for a period of 3 years. Those results will be the best results I would get if I had executed the EA with those parameters, but that does not mean that I will see good results in the future (like a self-fulfilling prophecy). How should I use the optimization function? Should I run an optimization with 3/4 - 1/4 forward and use that configuration for the same 1/4 time? (ie 4 months => 3/4 backtesting and 1/4 forward, so each month I run an optimization and use the best settings for the next month).


Is there a way to make an EA constantly optimize over time?

Icham Aidibe
11958
Icham Aidibe  

There's one way to make an EA optimized ad vitam aeternam - over time & across instruments - self training/self optimizing.

There's tools for that, neural networks is one, there's others like genetic algorithm : for that you'll need to feel comfortable with the coding part as it's complex enough

Forward test is a way to make your test more reliable but it's not a Bible neither.

A strategy you build in MQL5 could be built in any other language (python, perl, c++, mql4), cross platform test, cross language & cross data tests are also a way to proceed, after all ... 1+1=2 whatever the language, right ? 

https://www.mql5.com/en/blogs/post/715169

Effective stress tests I use to avoid curve-fitting and its dangers
Effective stress tests I use to avoid curve-fitting and its dangers
  • www.mql5.com
Since I had never used the blog offered by MetaQuotes and since it's a recurring difficulty for anyone testing an EA (self-made or bought) I thought it was an opportunity to share my tips about. But first, know that English isn't my main language, so be indulgent. What is curve-fitting ?  Over-optimization (curve fitting) is a process of...
Pedro Severin
1263
Pedro Severin  
Icham Aidibe:

There's one way to make an EA optimized ad vitam aeternam - over time & across instruments - self training/self optimizing.

There's tools for that, neural networks is one, there's others like genetic algorithm : for that you'll need to feel comfortable with the coding part as it's complex enough

Forward test is a way to make your test more reliable but it's not a Bible neither.

A strategy you build in MQL5 could be built in any other language (python, perl, c++, mql4), cross platform test, cross language & cross data tests are also a way to proceed, after all ... 1+1=2 whatever the language, right ? 

https://www.mql5.com/en/blogs/post/715169

Neural networks... I know a little about that. Correct me if I'm wrong, but with optimization: let's say I optimize an EA for 3 weeks. Should I expect it to work well around 1 week? So I would need to optimize it every week and the backtesting optimization period would be 3 weeks?

I am getting it right? It would be the same if I use neural networks? I ask because as far as I know, using NN does not auto-optimize the EA over time (but it can be coded as it is mentioned here). I mean, a NN itself it is not enough to let the EA run for ever. It has to be optimized at some point, right?

Nikolai Semko
7452
Nikolai Semko  
ppsev:

Neural networks... I know a little about that. Correct me if I'm wrong, but with optimization: let's say I optimize an EA for 3 weeks. Should I expect it to work well around 1 week? So I would need to optimize it every week and the backtesting optimization period would be 3 weeks?

I am getting it right? It would be the same if I use neural networks? I ask because as far as I know, using NN does not auto-optimize the EA over time (but it can be coded as it is mentioned here). I mean, a NN itself it is not enough to let the EA run for ever. It has to be optimized at some point, right?

If manual auto-tuning is required, it means that there is no AI there. The artificial intellect is therefore called that because it takes decision-making instead of a man. And do not forget that artificial intelligence is not only NN. 
Pedro Severin
1263
Pedro Severin  
Nikolai Semko:
If manual auto-tuning is required, it means that there is no AI there. The artificial intellect is therefore called that because it takes decision-making instead of a man. And do not forget that artificial intelligence is not only NN. 

So it should be a combination of NN with GA (Genetic Algorithm)? Like using GA to optimize the weight used in the NN?

Nikolai Semko
7452
Nikolai Semko  
ppsev:

So it should be a combination of NN with GA (Genetic Algorithm)? Like using GA to optimize the weight used in the NN?

Of course not. In general, for the creative process should not be "should be"