Walk Forward Equity Curves

Posted By: Murmurs

Walk Forward Equity Curves - 06/22/16 15:31

I'm looking for some precise clarity on the equity curves produced when doing walk forward training. After we "TRAIN", and then press the "RESULT" button, we view the results, including an equity curve.

It appears this equity curve includes the equity produced during the initial optimization period and perhaps only the final test period. Am I correct? Can we get a plot of only the test periods, segmented together into one continuous equity curve?
Posted By: Murmurs

Re: Walk Forward Equity Curves - 06/23/16 19:10

I'm a little surprised nobody else finds this to be an issue or of interest. Perhaps I'm just dense and not seeing the obvious, so if someone could enlighten me, I would appreciate it.
Posted By: boatman

Re: Walk Forward Equity Curves - 06/24/16 04:33

Originally Posted By: Murmurs
It appears this equity curve includes the equity produced during the initial optimization period and perhaps only the final test period. Am I correct?


This is not correct. After training a walk-forward script, clicking TEST begins the simulation period at the end of the first optimization cycle. The equity curve is the out of sample periods of each optimization cycle stitched together.

You can run into trouble if you change the StartDate and/or EndDate parameters between training a WFO and testing it. These parameters set up the beginning and end of the simulation period, and are used to calculate the break points between your WFO cycles. If you change those parameters, you need to re-train your walk forward.
Posted By: Murmurs

Re: Walk Forward Equity Curves - 06/24/16 17:50

Originally Posted By: boatman
This is not correct. After training a walk-forward script, clicking TEST begins the simulation period at the end of the first optimization cycle. The equity curve is the out of sample periods of each optimization cycle stitched together.


Okay. Thank you. For clarity, please verify the following for me:

True or false: When we click TEST, each out of sample period, representing one walk forward cycle, is produced from the optimized parameters for that cycle. In other words, if we're optimizing the variable X, and in the first cycle, it's optimized value is 0.2, and in the second cycle, it's optimized value is 0.3, then clicking TEST changes the value X (0.2 and then 0.3) to its respective values over those two cycles of the equity curve.

Are these optimized values and equity values with their associated date periods reported in some type of report or log file in text format?
Posted By: Murmurs

Re: Walk Forward Equity Curves - 06/24/16 18:12

I just TRAINED Workshop6_1, then TESTED it. Clicking RESULT after both operations yielded the exact same equity curve. Is the reported equity curve after TRAINING also reflective of only out of sample results?

Problem is, I seem to recall forcing a single walk forward cycle on a script, in which case the out of sample result would have only been a fraction of the total time period, but the reported equity curve spanned the entire period. That's at least what I recall. I'm not certain.
Posted By: MatPed

Re: Walk Forward Equity Curves - 06/25/16 09:39

Originally Posted By: Murmurs

Are these optimized values and equity values with their associated date periods reported in some type of report or log file in text format?


yes, .par file(s). Details are all in the "Train" section of the manual. Ciao
© 2024 lite-C Forums