Is changing TF an effective robustness filter?
2 replies
FX Success
4 years ago #278191
I was recommended to use higher and lower timeframes when filtering strategies for robustness pourposes. However, results are terrible. I want to create an 1H strategy and I tried 15M and 4H as higher and lower TF. No matter how many potential strategies I finf, when testing that higher/lower filter not a single one is passing it! The most frustrating case was one in which the builder found more than 14500 strategies (yes, almost 15000) and not a single one passed the TF robustness.
What do you think about that? Is that filter so fundamental or could I skip it? I find difficult to believe that more than 14500 strategis can be all bad! 🙁
0
Tomas Matejka
4 years ago #278215
Hello,
in our opinion it is highly dependent on what type of strategy is used for the test. Some strategies will fail naturally in such test because they use very short term edge (advantage) and higher timeframes simply cannot provide such edge. Generally speaking though I always check on other timeframes. Strategy does not have to remain profitable necessarily but when it fails hard I look for a reason why.
Remember there are no strict rules in data-mining. More strict you are less strategies are left when the build process is finished but still you have no certainty whether the passed multi-timeframe test will guarantee future profitability
0
FX Success
4 years ago #278256
Interesting. Thank you very much for your reply.
0
Viewing 2 replies - 1 through 2 (of 2 total)