self-adaptive systems

Post your new strategies, discoveries or just ideas for development

self-adaptive systems

Postby okx » Sat Apr 07, 2012 7:25 am

Hi all,

this is for automated robot trading, as with MetaTrader Expert Advisors:
The MetaTrader platform has a built-in strategy tester and optimizer which can be used for optimizing parameters of the strategy on historical data (backtesting).
The problem with such optimization is that it more often than not leads to overfitting, not leading to successful trading because the parameters were overfitted for a certain period of time.
But sometimes markets seem to have longer time frame characteristics.
Sometimes a strategy works well for two years in a certain time frame, say H1, and the next two years H1 doesn't work, but D1 does, but D1 didn't work the two years before.
(example: post35229.html#p35229)
Similarly some parameters in strategies might work well over a long period of time, but then need to be re-adjusted because long-term market conditions had changed.
This is only an idea, I can't confirm yet that this is so, it needs thorough testing, and the development of a self-adaptive system.
The parameters that I found that seem to follow long-term patterns sometimes are the RSI levels and the timeframe.
Has anyone made similar observations? Experiences? Suggestions? Comments?

Thanks and regards,
Oliver
okx
 
Posts: 10
Joined: Mon Apr 02, 2012 3:24 am

Return to Forex trading strategies and systems



cron