Trading System Development Process: I'm interested in your experiences - page 2

New comment
 

Things I have discovered before I "got it":

1.) I encountered information overload trying to read forexfactory.com a year ago and so I stopped reading - too many systems, too many lines. Now, I only read this forum - a good chunk of programmers and a few mathematicians here - a much better voice and a much smarter 'crowd'.

2.) I relied heavily on the backtester and tried to optimize and optimize. Problem is that with FOREX you're trying to hit a moving target - a non-deterministic system.

3.) Testing using the backtester takes a bit of time and can be maddening hitting 'Start' over and over and over. Using this tool is much faster: 'Visual Testing of the Profitability of Indicators and Alerts'

4.) I was trying to set limits for parameters that were of my design and not of the market's. I was trying to trade with X period, when the right period was X*4.5: http://arachnode.net/blogs/arachnode_net/archive/2009/06/13/automatic-optimization-of-forex-indicators.aspx

5.) I didn't pay attention to risk vs. reward : Know what all of the figures on the backtester mean.

6.) Know when your pair is active and trade during that time.

7.) If you are a good programmer, learn to write genetic algorithms. Specifically, feedforward genetic algorithms: http://www.ai-junkie.com/ann/evolved/nnt1.html I have a script that generates 10-15 profitable expert advisors per day. Run the EA's for week and replace them with the newest set. If you don't want to code a GA in MQL4, let the backtester do the GA work for you.

8.) Employ iStdDevOnArray to help keep indicators flat when the market is going sideways.

9.) Take a break from coding and go searching for others' work: http://arachnode.net/blogs/arachnode_net/archive/2009/06/11/top-performing-expert-advisors.aspx There is no shame in letting someone else make some money for you while you learn.

10.) Find a mentor or FOREX friend in your area - the main benefit of this is that a partner in code will test your code more thoroughly.

11.) Trade longer timeframes and trade on the OPEN.

Finally, what is your profitable system?

 
arachnode.net:

Things I have discovered before I "got it":

1.) I encountered information overload trying to read forexfactory.com a year ago and so I stopped reading - too many systems, too many lines. Now, I only read this forum - a good chunk of programmers and a few mathematicians here - a much better voice and a much smarter 'crowd'.

2.) I relied heavily on the backtester and tried to optimize and optimize. Problem is that with FOREX you're trying to hit a moving target - a non-deterministic system.

3.) Testing using the backtester takes a bit of time and can be maddening hitting 'Start' over and over and over. Using this tool is much faster: 'Visual Testing of the Profitability of Indicators and Alerts'

4.) I was trying to set limits for parameters that were of my design and not of the market's. I was trying to trade with X period, when the right period was X*4.5: http://arachnode.net/blogs/arachnode_net/archive/2009/06/13/automatic-optimization-of-forex-indicators.aspx

5.) I didn't pay attention to risk vs. reward : Know what all of the figures on the backtester mean.

6.) Know when your pair is active and trade during that time.

7.) If you are a good programmer, learn to write genetic algorithms. Specifically, feedforward genetic algorithms: http://www.ai-junkie.com/ann/evolved/nnt1.html I have a script that generates 10-15 profitable expert advisors per day. Run the EA's for week and replace them with the newest set. If you don't want to code a GA in MQL4, let the backtester do the GA work for you.

8.) Employ iStdDevOnArray to help keep indicators flat when the market is going sideways.

9.) Take a break from coding and go searching for others' work: http://arachnode.net/blogs/arachnode_net/archive/2009/06/11/top-performing-expert-advisors.aspx There is no shame in letting someone else make some money for you while you learn.

10.) Find a mentor or FOREX friend in your area - the main benefit of this is that a partner in code will test your code more thoroughly.

11.) Trade longer timeframes and trade on the OPEN.

Finally, what is your profitable system?

Arachnode

You set my head racing with this post - I've been a while digesting it. I've experienced points 1, 2, 3, 4, 5 & 10 directly and the articles you refer to on genetic algorithms and neural nets are brilliant.

I have a question on your use of the indicator tester, but it's not really consistent with this thread. I fell into a bit of trap with it, which is what my question/comment is about. What's the best way?

Cheers

Jellybean

 

Thanks for sharing your experience.

There are two type of EAs, Statistical and mathmatical
1) Statistical. EA that made on statistical calculation, like rank correlation between pairs.
this site gives very useful information
https://www.mql5.com/go?link=http://forexautomaton.com/research/34-correlations20022007
and this site https://www.mql5.com/go?link=http://www.rba.gov.au/publications/bulletin/2007/dec/pdf/bu-1207-1.pdf
This means we study the pair's behaviour and their correlation at each GMT hours and get statistical on your winning side.
then I apply the mathmatical strategies accordingly.
2) mathmatical EA that made on mathematical calculations rules, like MACD + MA + RSI etc, and filter them with Stdv + ATR. it is very hard to distinguish the very earlier trend that is in its baby phase from ranging period. overall, I experience the alternative 1, gives much better hitrate on EAs that possibly making money than alternative 2.
 
arachnode.net:11.) Trade longer timeframes and trade on the OPEN.

I don't always trade at open price. I trade when a HIGH or LOW is breached, speacilly when trading resitances and supports.
12
New comment