THANK YOU SOOO MUCH for all your efforts!!! I am getting closer to the point where I need to optimize my strategy (a strategy with many parameter combinations). After reading up on all of this genetic optimization technology, I'm convinced I'll need your GO.
I do want to commend your persistence in getting this done and 'dealing' with the brick-wall that is NT support. I saw many of your posts on NT's forum and I was viscerally upset with how they treated you -- someone looking to expand their product in such a meaningful way for FREE! unbelievable.
In any case, I think I have some more studying ahead of me regarding GAs and getting your GO to work with my strategy. I believe I'll need to modify my strategy in the ways you described on this thread to get it done. I'll post here if I run into any issues or have any questions. If you need any assistance, let me know; I'm willing and able.
a 'MinimumPerformance' cutoff parameter. once the best found performance reaches this value the optimizer will automatically stop.
PHStochastic - a new optimizer. it works by randomly choosing sets of strategy parameters giving higher preference to values of each individual strategy parameter that have previously yielded better performance. (breath). i can't really tell if it's better than the GO since there's so much randomness involved. but it does have the advantage that it never gets stuck, so it doesn't have to keep resetting like the GO. for longer runs it may be worth a try.
hope that helps!
The following 4 users say Thank You to piersh for this post:
thanks a lot for your work. I tried to import the last version 1.12 into NT7 public beta.
I always get the error message:
"Import failed. The NinjaScript Archive may contain duplicate method names that already exist on your PC or there is a required file missing on either your PC or in the import archive."
I even installed a fresh and empty NT7 public beta on a new blank machine and tried to import the optimizer-zip - but got the same error message. There are NO other custom indicators or strategies which makes this behaviour really strange.
One of my parameters I am testing has a min value of 0.0001 and a max of 0.0009 with an increment of 0.00001. After running the optimizer the best value was given as 0.0089 - which is of course increment away from the max value.
I then change the max value to 0.0010 and ran the optimizer again - the new optimal value was at 0.0009 - so it seems that the boundary values are not included in the test. I am not saying it is a bug - just thought that it would be good to know.
Also - it would be great if the values written to the output window has more than two digits after the the decimal point - as my example shows I would like at least 5 digits after the decimal.
The following user says Thank You to verge for this post: