Optimization duration design experience

[Sky All]

Dear experienced traders



Could you share any your experience about optimization duration design with us?

These days a mind is in my head is, if a system is not a martingale system, long duration optimization might not meaningful. 

If a system work with TP and SL, then it have to also fit to the current market. 

How long do you think the optimization duration should be? What is your criteria for duration design? 

Thank you in advance and best wishes,

Sky

[Sergey Golubev]

I made a lot of optimization in past, and I compared the duration, different algorithms and money which I was spending with cloud.
You can find the links to all of that on the post #186 (together with the EAs by the source codes, the settings, and the trading results).

----------------

Could you share any your experience about optimization duration design with us?

If someone will share the optimization results so please - upload your EA by the source code together with the setting and everything.
Because it is free forum and everything should be for free here.

[Sky All]

Sergey Golubev:

I made a lot of optimization in past, and I compared the duration, different algorithms and money which I was spending with cloud.
You can find the links to all of that on the post #186 (together with the EAs by the source codes, the settings, and the trading results).

----------------

If someone will share the optimization results so please - upload your EA by the source code together with the setting and everything.
Because it is free forum and everything should be for free here.

Thank you so much for your continuing support, Sergey! 

I didn't find the optimization duration design information by a quick glance. Let me check it carefully again later. 

Actually, I'd like to make my EA as an open EA, because how many people I can communicate with, how many idea will be created. 

Also, the current EA works simple indeed and I don't think it will be more than a toy to experienced traders here. 

However, I still have to discuss and make a decision with important stockholders. 

Anyway, what I want to discuss in this forum topic is about the philosophy/idea/method of optimization duration design. 

Thanks again! 

[Sergey Golubev]

Sky L:

...

I didn't find the optimization duration design information by a quick glance. Let me check it carefully again later. 

...

If about the duration only (and about "how much")

• post #66 and #67
• post #75 

But please note that I did it 6 or 7 years ago (with cloud but it was long time ago).

--------------

Everything with backtesting (I have many results/statements for optimization and backtesting, it is just an example):

The setting for EAs: optimization and backtesting 

• optimization results for this EA for EURUSD H4
  
• backtesting results for EURUSD H4 with the settings #1
  
• backtesting results for EURUSD H4 with the settings #2
  
• optimization results for EURUSD M15 
  
• backtesting results for EURUSD M15 with the settings #1
  
• optimization results for GBPUSD M15
  
• backtesting results for GBPUSD M15 with the settings #1

[Sky All]

Sergey Golubev:

If about the duration only (and about "how much")

• post #66 and #67
• post #75 

But please note that I did it 6 or 7 years ago (with cloud but it was long time ago).

--------------

Everything with backtesting (I have many results/statements for optimization backtesting, it is just an example):

The setting for EAs: optimization and backtesting 

• optimization results for this EA for EURUSD H4
  
• backtesting results for EURUSD H4 with the settings #1
  
• backtesting results for EURUSD H4 with the settings #2
  
• optimization results for EURUSD M15 
  
• backtesting results for EURUSD M15 with the settings #1
  
• optimization results for GBPUSD M15
  
• backtesting results for GBPUSD M15 with the settings #1

Thank you so much, Sergey

[Sergey Golubev]

Sky L:
...

..

Anyway, what I want to discuss in this forum topic is about the philosophy/idea/method of optimization duration design. 

...

...

I made a conclusion about the method of optimization (it was 7 years ago):

Thus, my conclusion is the following.

• Cloud is helping a lot.
• In case of complex EA with many parameters to optimize - it may be better to start with Slow Complete Algorithm using Cloud. Yes, we will spend some money (more than few dollars exactly) but we will receive possible ranges of the settings for many parameters in more quick way. 
• After we get it - we can make fast generic optimization (using Cloud) spending less than 0.1 credit just to obtain more specific information about some particular parameter.

It is from the post #29

[Sky All]

Sergey Golubev:

I made a conclusion about the method of optimization (it was 7 years ago):

Thus, my conclusion is the following.

• Cloud is helping a lot.
• In case of complex EA with many parameters to optimize - it may be better to start with Slow Complete Algorithm using Cloud. Yes, we will spend some money (more than few dollars exactly) but we will receive possible ranges of the settings for many parameters in more quick way. 
• After we get it - we can make fast generic optimization (using Cloud) spending less than 0.1 credit just to obtain more specific information about some particular parameter.

It is from the post #29

Thank you, Sergey. 

There is a misunderstanding. 

What duration I want to talk about is how long do you set start time to end time in back test and forward test in optimization. 

Please do not delete your valuable comment. 

Let me make a new topic later with a better explain. 

[Sergey Golubev]

Forum on trading, automated trading systems and testing trading strategies

Cost of cloud computing

Sergey Golubev, 2013.01.21 16:47

Hi ToolMaker,

I am using Cloud in almost every day. From what I understood: if I am making fast generic optimization so I am paying something as 0.01 credit or less but optimization is going slow (not fast); if I am making Slow Complete Algorithm optimization so it is going very fast (because all the agents are used) and ... I pay something with range 0.40 to 20 credits for one optimization (for example - I paid 0.40 today, and two days before yesterday I paid 0.73; and one week ago I paid 20.22 credits for one optimization but it was special case with very complex EA with a lot of parameters to optimize). You can see how credits are deducted in real time in Agents window so you can stop optimization with Cloud any time you want.  

You can see some optimization results with log file data incl cost - on this thread How to Start with Metatrader 5 and this thread PriceChannel Parabolic system . And basicly you need just 2 sources: this article Speed Up Calculations with the MQL5 Cloud Network and this page (about costs and how costs calculated) https://cloud.mql5.com/en/faq/payments

[Sergey Golubev]

The last one from me:

----------------

#1 and as a conclusion:

Forum on trading, automated trading systems and testing trading strategies

No profiting from completed jobs (agents)

Sergey Golubev, 2013.07.08 16:07

So, if you see that I pay 0.05 dollar so It means that I made optimization for fast algorithm (which is many hours by duration). If you see that I paid 12 dollars/credits so it means that I made slow optimization which is less than 1 hour (very fast) and in very comfortable way.

This is my experience.

----------------

Please note that it was long time ago, and the duration of the optinmization can be found on my previous posts (look at the links).
I did not for the EAs which were coded for free by the coders and which were placed on my threads by source codes.
Use can use search function to find more, for example - this search results.