Something I was working on yesterday, and I figured may interest people.
I wanted to test that what happens if I tried having my system look for trend continuation instead of reversal. While it works on most currency pairs, a few (EURGBP and NZDJPY spring to mind) are incredibly resistant to it. Now, my system is symmetric; it enters and exits long/short on the same criteria, but upside down. So, simple enough to change short and long entry/exit conditions around, and test it.
Basically, it rapidly became obvious that minimal profits could be made on currency pairs that tend to seriously trend. These never matched what my system was making on watching for reversal, and the optimised parameters were significantly less reliable. Optimising on a subset of the data then expanding out to cover the entire data period also basically wrecked the profits.
Now, this isn't proof my system wasn't overfitted, but I feel strongly shows that the underlying design is sound, and profits are not merely an artefact of fitted data.
Hope that was interesting to people, feedback welcomed.
I wanted to test that what happens if I tried having my system look for trend continuation instead of reversal. While it works on most currency pairs, a few (EURGBP and NZDJPY spring to mind) are incredibly resistant to it. Now, my system is symmetric; it enters and exits long/short on the same criteria, but upside down. So, simple enough to change short and long entry/exit conditions around, and test it.
Basically, it rapidly became obvious that minimal profits could be made on currency pairs that tend to seriously trend. These never matched what my system was making on watching for reversal, and the optimised parameters were significantly less reliable. Optimising on a subset of the data then expanding out to cover the entire data period also basically wrecked the profits.
Now, this isn't proof my system wasn't overfitted, but I feel strongly shows that the underlying design is sound, and profits are not merely an artefact of fitted data.
Hope that was interesting to people, feedback welcomed.
