Thanks to everyone for your input and thoughts over these last 24 hours. There have been some questions asked.
MarkGavalda, how often to optimize? how large of a sample to optimize over? My beginning benchmark has been once a week over the last 6 months. As I make the EA better, we should be to optimize it over longer and longer periods. I believe though that the market wave's height and length will be substantially different from the waves 4-5 from now. So until I have a expert that can basically optimize itself continuously, I wouldn't optimize it over 8 months. But I don't want curve-fitting. When I optimize, I make the majority of the passes gave still gave me significant returns, and that there many many passes with similar parameters that got similar returns.
When the next version comes out, I'll make sure to add all the additions to it. I spent today moving the declaring and defining above the start(), and then getting rid of all the ranging code. Before I had crudely just doubled all the trend code for the ranging side of the EA, doubling the size of the EA, and making the EA define 80 parameters with every tick. It was extremely crude. So I got ride of all the range code, made all the code standard, (so now the EA has 40 parameters) and I move the defining above the start() function so it's only being called(and define) once for every bar. The end result is optimizations are much much quicker and the EA is much easier to work with since it's only half as long. Thanks a million to raytracy, I'd have no idea any of that was even possible without him.
Tonight I'm going to be loosening up on the entry conditions in some areas. With the range and trend trades I'm looking for at least 6-10 trends a month and we're not there yet. This trade that was opened this morning looks like its about to be closed with a loss. I don't know whether the market is going up or down tomorrow, but the EA should, so obviously its still not as smart as it should be.
MarkGavalda, how often to optimize? how large of a sample to optimize over? My beginning benchmark has been once a week over the last 6 months. As I make the EA better, we should be to optimize it over longer and longer periods. I believe though that the market wave's height and length will be substantially different from the waves 4-5 from now. So until I have a expert that can basically optimize itself continuously, I wouldn't optimize it over 8 months. But I don't want curve-fitting. When I optimize, I make the majority of the passes gave still gave me significant returns, and that there many many passes with similar parameters that got similar returns.
When the next version comes out, I'll make sure to add all the additions to it. I spent today moving the declaring and defining above the start(), and then getting rid of all the ranging code. Before I had crudely just doubled all the trend code for the ranging side of the EA, doubling the size of the EA, and making the EA define 80 parameters with every tick. It was extremely crude. So I got ride of all the range code, made all the code standard, (so now the EA has 40 parameters) and I move the defining above the start() function so it's only being called(and define) once for every bar. The end result is optimizations are much much quicker and the EA is much easier to work with since it's only half as long. Thanks a million to raytracy, I'd have no idea any of that was even possible without him.
Tonight I'm going to be loosening up on the entry conditions in some areas. With the range and trend trades I'm looking for at least 6-10 trends a month and we're not there yet. This trade that was opened this morning looks like its about to be closed with a loss. I don't know whether the market is going up or down tomorrow, but the EA should, so obviously its still not as smart as it should be.