I have tried excluding certain time periods, like the open, or news, etc and found results were not improved. It takes fewer trades and makes less money, that is expected. But what I mean is the win percentage or draw down were not any lower. There was not any perceived reduced risk which was surprising but nonetheless I run them from early morning (5am or so) to middle of afternoon. I have some that stop at lunch but there was no major reason backtesting wise, it was a personal preference of not wanting to trade the whole day.
The following user says Thank You to caprica for this post:
Thank you for your question. It is not complicated anymore after I wrote the initial function. The purpose was to produce something similar to a trailing stop, but SetTrailStop() has many problems in Ninja like many things do and I wanted greater control of when to adjust the stops a fixed stop increment is too restrictive.
You might want to check out the sample on NT forums of how to manually do a trailing stop rather than use SetTrailStop(). You can modify it slightly to use TrailThreshold, TrailStep and TrailFrequency like the NT ATM system.
The benefit is that you only have 3 parameters to optimize instead of 9, and you can also implement techniques like chasing target which is where price can move through the target and the target chases it upwards until there is a downtick (or vice versa).
I make quite a lot of pips chasing volatile thrusts, compensates for the slippage incurred when you are on the wrong side of the thrust
Actually reducing parameter count is hugely beneficial for GO because of their tendency to get stuck in a local minimum. In other words, GO algorithms often trade accuracy for speed and in general more parameters means less accuracy, less chance of finding the optimal solution.
That said, it sounds like your current solution is working well for you so no sense in fixing what isn't broken
Intrigued by the weighted stops idea - never thought of that before. Will have to investigate further. My "problem" with scaling out is the huge drop off in expectancy at the 'cost' of an improved win percentage. I have never quite come to terms with the mix on this - would be interested in your views.