+1 888 340 6572
MultiCharts Project Management
previous_open_issue.png
Go to the previous open issue
previous_issue.png
Go to the previous issue (open or closed)
star_stroke.svg
Please log in to bookmark issues
feature_request_small.png
Open Feature request MC-1479

Continuous Walk Forward Optimization

minus.svg
Please log in to vote
18
Votes
pluse.svg
Please log in to vote
next_issue.png
Go to the next issue (open or closed)
next_open_issue.png
Go to the next open issue
Description

Walk forward optimization currently breaks up data into different IS/OSS segments. The trading strategy is only allowed access to the data in each segment. The approach causes many drawbacks and problems for strategy development.
When a strategy transits from one data segment to another, it must wait until it has enough data to calculate indicators before it can start trading. This results from not being able to access any data from the previous segment. Long look back periods, frequent recalibration, and small out of sample data window will all lower performance. As a result, many strategies will fail the walk forward test due to insufficient data rather than over fitting.
Ex. You're running a simple moving average strategy with a look back period of 50 days. Your trading strategy must wait at least 50 days in each OSS segment before it can start trading.
Therefore, I would like to have the option to break IS/OSS into trading segments rather than data segments. This mean the trading strategy will have access to the all the historical data, but only allowed to trade on the dates of IS set being optimized on. Therefore, there is no longer a need to programmatically load different data segments. This version of WFO is much easier to program and implement.

Steps to reproduce this issue

Ex. Walk Forward Analysis with 2 year IS and 1 year OSS
IS Set #1/OSS Set #1
IS Set #1: 1/1/2000 to 12/31/2001
OSS Set #1: 1/1/2002 to 12/31/2002
IS Set #2/OSS Set #2
IS #2: 1/1/2001 to 12/31/2002
OSS #2: 1/1/2002 to 12/31/2003

WFO Logic for IS Sets
If Date >= 1/1/2000 and Date = 1/1/2001 and Date = 1/1/2002 and Date = 1/1/2002 and Date

Comments (4)
#0
user-offline.png  momentum
Jan 02, 2014 - 05:19

The WFO data selection is fixed in that the user specifies the IS and OOS data days or bars. What is needed is a rolling optimization as above but with the addition of the WFO finding the best ratio of IS and OOS data. The user should speify a date range and then be able to optimize whether the ratio of IS to OOS should be, for example, 80% to 20%, 70% to 30% etc

#0
user-offline.png  Guest
Jan 03, 2014 - 03:54

Regardless of what the user selects for the IS and OOS days, the data issue still exists.  Try doing a WFO with an OOS of 1 day, and you will see what I mean.  What you are talking about is trying to optimize the IS and OOS lengths.  If CWFA were implemented then you would only need to find the optimal IS length since the OOS would always be set to 1 day iwhich reduce the complexity of the problem you're trying to solve.    

#1
user-offline.png  MultiCharts Support
Mar 30, 2017 - 09:43

It is not planned for the future versions.

Changes:
  •   icon_milestone.pngMilestone changed: Not determined => Declined/Not a Bug/Will not Fix/Cannot Reproduce/Duplicated
  •   icon_status.svgStatus changed: Under Review => Declined
History
Issue basics
  • Type of issue
    Feature request
  • Category
    Usability
  • Targeted for
    Declined/Not a Bug/Will not Fix/Cannot Reproduce/Duplicated
  • Status
    Declined
User pain
  • Type of bug
    Not triaged
  • Likelihood
    Not triaged
  • Effect
    Not triaged
Affected by this issue (2)
People involved
  • Posted by
    user-offline.png  QuantArb
  • Owned by
    Not owned by anyone
  • Assigned to
    Not assigned to anyone
  • Subscribers
    3 subscriber(s)
    Click here to show the list of subscribers
Times and dates
  • Posted at
  • Last updated
Issue details
  • Resolution
    Not determined
  • Severity
    Normal
Commits (0)
There are no code checkins for this issue
×

Sign up and get MultiCharts free

Use its powerful simulation mode with data included out of the box. Just fill out the form and start honing your trading skills with a few clicks.