WFO Efficiency is Zero
Author: fairone99
Creation Date: 4/20/2016 7:16 PM
profile picture

fairone99

#1
Hi!

I have a simple strategy and I have used both the normal Optimizer and WFO on it.

However, the WFO Efficiency result of the WFO is almost always Zero...

I know that WFO Efficiency over the value of 50 is better, but mind is almost always at the value of Zero.

Here are the steps that I do to perform the WFO:

1. I run my strategy with daily scale, about 12 years data range, and position size of 25%.
2. After I run my strategy, I bring up the normal Optimizer. And, within the normal Optimizer, I take out few parameters that I do not need to optimize on.
3. Than I bring up the WFO, normally with 25% Out-Of-Sample data and in 10 Intervals. With Sliding window type and Monte Carlo with number of runs and number of tests in each run both set to 20. With the Highest APR selected.

I ran WFO for more than ten times with some minor changes to the above configuration. However, I am almost always getting Zero at WFO Efficiency.

Have I done something wrong or missing some needed steps?


Alex
profile picture

Eugene

#2
Hi Alex,

Note how many trades were taken. Perhaps there are too many intervals for given trade duration or a high position size is dropping too many trades. In this case a picture is worth a thousand words.

P.S. Here's some food for thought: WFO Efficiency error 'NaN'
profile picture

fairone99

#3
Hi! Eugene,

Thank you for the suggestions and I did check out the post - WFO Efficiency error 'NaN'


Normally, there is around 1000 to 3000 trades taken during this 10 years period (not 12 years period as mentioned before). Position size is set to 25% and I have calculated before that about 40% trades are NOT taken.

I know the higher the WFO efficiency, the more robust the strategy is... But, I am almost always getting zero on WFO efficiency.

However, I figure out my strategy is somewhat stable against new data.

For example: I always use 3/30/2005 to 3/30/2015 data for any optimization. I run this strategy on the data from 7/1/2003 to 3/29/2005, the result is somewhat within my expectation. Than, the strategy has been run on the really "newer" data which is from 3/31/2015 to 4/15/2016, the result is somewhat expected and stable as well. Both the "older" and "newer" data have NOT been "seen" before by the strategy, so if the result is somewhat expected and stable, I am assuming regardless about the Zero WFO efficiency, the strategy should be considered somewhat robust and stable, right?


Alex
profile picture

Eugene

#4
How high is the Profit Per Bar for your trades Alex? This also affects WFO Efficiency %.
profile picture

fairone99

#5
If using the $25000 (RP) then the Profit Per Bar is about $99

If using 25% portfolio simulation then the Average Profit % Per Trade is about 2%

Just want to make sure that my strategy is stable and somewhat robust (tested against about three years worth of data that this strategy has never "seen" before) even it always has Zero WFO Efficiency...



profile picture

Eugene

#6
Could you copy/paste the WFO results table here for me to have a look?
profile picture

fairone99

#7
Sure, Eugene...

The problem is that I do not have the result table of WFO screen captured before I close it.

I will run the WFO again and post the result back later.


Alex
profile picture

fairone99

#8
Hi! Eugene,

Finally, I re-run the WFO and here I attach the screen shots as you can see I am still getting Zero on all intervals within the WFO result.

However, I do feel my strategy is somewhat stable as it has been run against about 2-3 years "unseen" data and produces expected results.


Alex
profile picture

Eugene

#9
Thanks for the update Alex. Given these numbers, I don't have an explanation at hand. Perhaps this is specific to your strategy? Is there a dependency on strategy, bar scale, DataSet, WFO settings? How can this be reproduced?
profile picture

fairone99

#10
Eugene,

I think that I will check out how WFO efficiency been calculated in more details. And try to find any clue that might be related to my strategy. I will post back, if I figure out something, thanks!