I am not sure how the average drawdown and the number of drawdowns are calculated. Average dd of 0.43% and number of dd's of 290 seem to be very small values with a 22-year backtest on the Dow Jones portfolio ...
290 times is somehow understandable. but...
if% new highs is at 11.32% the system is in a dd for the other time, right? Or is it 0.43% of the average dd per trade?
can you please explain a little bit how these key figures are to be understood?
290 times is somehow understandable. but...
if% new highs is at 11.32% the system is in a dd for the other time, right? Or is it 0.43% of the average dd per trade?
can you please explain a little bit how these key figures are to be understood?
Rename
The difference comes from counting all the 290 small and big drawdowns by the Average Drawdown logic. Maybe the metric itself isn't very representative since it counts every small dip. :(
P.S. Please use PNG when posting screenshots to improve the way they look. JPG makes them blurry and with lots of artifacts.
http://www.labnol.org/software/tutorials/jpeg-vs-png-image-quality-or-bandwidth/5385/
http://www.labnol.org/software/tutorials/jpeg-vs-png-image-quality-or-bandwidth/5385/
Thanks Eugene, but conversely, if the system makes 11.32% new highs, it is the other 88.68% in a drawdown, or isn't it?
With 250 trading days per year over 22 years, the system would have to be several thousand days in a (no matter how small) dd.
The average DD is also incomprehensible to me. even if every small dip is calculated, the value should still be significantly higher.
or do I have a thought-mistake here?
With 250 trading days per year over 22 years, the system would have to be several thousand days in a (no matter how small) dd.
The average DD is also incomprehensible to me. even if every small dip is calculated, the value should still be significantly higher.
or do I have a thought-mistake here?
Thomas, I just realized that the Average Drawdown code has been changed in a recent refactoring effort. Unfortunately, this had a breaking impact on it. So for example where the old formula returned a -15% average drawdown the new one shows an unrealistic sub -1% figure.
I've restored the classic logic, please look forward to a fix in B18 and thanks for the heads-up!
I've restored the classic logic, please look forward to a fix in B18 and thanks for the heads-up!
Perfect. Thanks!
@Glicht: the problem could also be solved with the DD Times (5%, 10% ...). I once had talked to Dr. Koch about this metric and he liked it.
But now he has implemented my idea in his "Advanced Scorecard".
But it is hidden there and not accessible to everyone. Perhaps you can save these key figures suggested by me in the extended scorecard.
Then the problem would be solved transparently, because the DD Times are implemented correctly (in my opinion).
@Glicht: the problem could also be solved with the DD Times (5%, 10% ...). I once had talked to Dr. Koch about this metric and he liked it.
But now he has implemented my idea in his "Advanced Scorecard".
But it is hidden there and not accessible to everyone. Perhaps you can save these key figures suggested by me in the extended scorecard.
Then the problem would be solved transparently, because the DD Times are implemented correctly (in my opinion).
You should submit a feature request for this. We have so many requests that we use this mechanism to keep track of them all. Create a new post and click the #FeatureRequest tag before submitting. You can see all the requests in the WishList page, link in the Discussions page.
It seems that the calculation of the average drawdown is still (or again?) not working properly. I ran a backtest from 1/1/2007 to 12/31/2013 with the SPY as benchmark. In the extended scorecard the Average Drawdown % for the benchmark is -3.50%. A visual check with the drawdown curve shows that this is obviously not the correct value. Downloading the drawdown curve data to EXCEL and calculation the average there gives a value of -18.8735%. The same comparison for the strategy shows -1.95% in the extended score card and -3.6573% in EXCEL.
I'll take a look.
You're right, I fixed the calculation for Build 105.
Thank you Glitch for the immediate response and for solving the problem.
Your Response
Post
Edit Post
Login is required