I made this topic to discuss about randomize trades test with resampling method. As I learnt in the SQ course, this test is made by “eye”, don´t has a objective criteria to filter the strategies and was done looking the robustness graphs. But, when I did this test, I observed that each test could show a different graphical result for the same strategy. I attached four simulation of the same strategy using and 200 random simulations of resampling method. As we can see, the first and the last simulation passed the test and the second and third failed due some curves had worst results from the majority of curves. So, due the random nature of the test, a good strategy could be excluded depending of the random simulation. I think in this case, is needed to follow objective criteria to filter the strategies. One criteria, could be RET-DD (RT)/ RET-DD (IS) > 0.33 (1/3). We usually uses RET-DD (RT)/ RET-DD > 0.5 to filter the strategies, but for this specific test, I think that is better use a lower factor to filter the strategies. Somebody has a objective criteria to filter the strategies in this case?
Attachments:You must be logged in to view attached files.
I found the ratio RET-DD<0.33 using two strategies that passed all other robustness tests. I ran 1000 resampling tests for these two strategies and saw that the results are more stable with 1000 tests than 200. The two strategies showed RET-DD ratio greater than 0.33. I will test more strategies and see the behavior. Thanks
You must be logged in to reply to this topic.