Download MetaTrader 5

Discussion of article "How to Develop a Profitable Trading Strategy"

To add comments, please log in or register
MetaQuotes Software Corp.
Moderator
192607
MetaQuotes Software Corp.  

New article How to Develop a Profitable Trading Strategy has been published:

This article provides an answer to the question: "Is it possible to formulate an automated trading strategy based on history data with neural networks?".

The process of developing successful trading strategies with implementation of technical analysis can be divided into several stages:

  1. Attach several technical indicators to a chart window of a financial instrument's price, and identify patterns of market correlations and signal indicators.
  2. Formulate data obtained from the previous correlation step.
  3. Convert strategy to a relevant programming language to create a mechanical trading system.
  4. Run the trading system through a simulator based on history data and try to match its input parameters (optimize).
  5. If the previous step hasn't increased the balance, proceed to step 1.
  6. Run the system obtained through the previous stages on demo accounts for testing.
  7. If the previous step hasn't brought any profit from virtual money, proceed to step 1.
  8. Use the system in real-life trading, occasionally adjusting its input parameters to the changing market conditions.
That's it, really. The system created this way can be used in automated trading, as well as an advisor in manual trading in order to suggest the most important signals sent by technical indicators.
      
Let's see what happens, if we try to computerize the whole process.

This article analyzes the use of a simple single-layer neural network for identifying the future price movements based on the readings of the Acceleration/Deceleration (AC) Oscillator.

Author: Yury Reshetov

Sherif Hasan
7406
Sherif Hasan  

Great article Yuri. i really i learnt here some missed informations regards startegy.

Happy Piping..

Filipe Acerbi
843
Filipe Acerbi  

Hello

Thanks for the article. It is very interesting.

 I has one doubt about the number of points taken from indicator. In the test, you used 4 points with a interval of 7 bars from each other.

 How we calculate the number of points and the interval from each other? This is calculated empirically or mathematically?

 Thanks 

Rasoul Mojtahedzadeh
33319
Rasoul Mojtahedzadeh  

I don't know why, but I cannot reproduce the same results. The MetaQuote's history data has been used for this backtest. Any idea what I'm doing wrong?

 

 

Filipe Acerbi
843
Filipe Acerbi  

Hello

Has any form to train the neural network with other algorithm method? Like Standard backpropagation?

 I read that the genetic algorithm used by the tester is not the best one.

 Thanks 

Jeremy Roach
2380
Jeremy Roach  
Rasoul Mojtahedzadeh:

I don't know why, but I cannot reproduce the same results. The MetaQuote's history data has been used for this backtest. Any idea what I'm doing wrong?

 

 

I get the same poor results?

Any thoughts? 

Komgrit Sungkhaphong
200
Komgrit Sungkhaphong  

great great great article

Yury is pure master

i'm running optmization base on the code.

well, it is profitable. 

 

Jing Wang
302
Jing Wang  

Great article! Thanks for sharing.

People like Yury are the backbones of this community. We newbies need to read more of these and pay our attention to learning the skills rather than trolling through the "Market" tab in MT. 

Junqui
19
Junqui  

Saludos a todos.

Señor Yury Reshetov me puede decir por favor ¿porqué en el indicador del Perceptron (AC) usted escribe "symbol()" en lugar de "NULL"?.

=======================================================================================================

Greetings everyone.

Sir Yury Reshetov please tell me ¿why in the Perceptron indicator (AC) you write "symbol()" instead "NULL"?

El suyo / yours.


El mio / mine.

franzzzz
26
franzzzz  

can anyone confirm the backtest? im getting different results when optimising it. Then when I tried the results that he "got" to the testing period, it is different also.

1. Optimise period is 2005.11.14 to 2006.09.17. M30, EURUSD, openprices, Maximal DD < 35%.

2. results from optimise is. "We will obtain the results for the input parameters: x1 = 146, x2 = 25, x3 = 154, x4 = 121, sl = 45."

3. Testing period is "3 months after optimization period". So 2006.09.18 to 2006.12.18. correct?

4. result is supposedly profitable, but not to me.

really interested in knowing this but the backtest is not consistent.

Thanks!


regards,

Francis 

franzzzz
26
franzzzz  
Komgrit Sungkhaphong:

great great great article

Yury is pure master

i'm running optmization base on the code.

well, it is profitable. 

 

hi can you share what did you do, parameters, etc? im getting different results. Would really appreciate it

Thank you!

To add comments, please log in or register