How do you view robustness test pass conditions?
In several articles I find that people are using relative conditions to filter results of robustness test.
For example, if the Ret/DD at 95% confidence level is less than half of original strategy, it then fails the test.
But here I doubt whether the absolute values are also important. Say original strategy’s Ret/DD is 20, and is 9 at 95% level. Another strategy has an 5 original Ret/DD value and 3 at the 95% confidence level. Is the latter one necessary better than the first?
If you are using the 20-9 strategy, even if you are in the most unlucky 5% scenario you still beat the best circumstances in the 5-3 strategy at all time!!!
Any one agree with me?
the strategy 20-9 has bigger chance that the max historical DD will be bigger, so the real results will in most cases be not as good as in backtests. this can cause your portfolio DD will be bigger than you expect. so earnings are not as important as DDs.
also when the results are 20-9 with 50 runs, with 200 runs it can decrease to 20-6. I recommend you to use the best of what you find. If best is 20-9, use it. (generate as much str as you can and then keep 5% of them…if that 5% are around 20-9, its ok, if 5% are around 20-6, generate more strategies.)
Viewing 2 posts - 1 through 2 (of 2 total)