The Walk Forward Suite and corresponding Global Suite System, can be used with any system.
The concept is to use historical data for system optimization and then use the optimal set of parameters, based on a maximum goodness statistic, to then test going forward with an out of sample test. In the following chart, the bottom yellow arrow represents an optimization test from 1998 to 2002. When the system is set to step parameters, the test could be hundreds of stepped tests over this time frame to find the optimal parameters for the system. Then the optimal system parameters are used in an out of sample test for 2002 to 2003. The process begins again with an optimization run from 1999 to 2003, and an out of sample test for 2003 to 2004.
Select the Walk Forward Suite, and then select the system you want to test. Set the important system parameters to step as necessary.
On the Global Parameters page, are the walk forward parameters:
| • | Set the Run (Index ) parameter to step from 1 to the number of iterations you want. Increment by 1. |
| • | Set the Optimization Run parameter to step from true to false. |
| • | Set other parameters as desired, but not stepping. |
| • | Time Frame sets the time frame of the testing to weekly or yearly. The optimization period and out of sample period use this as the basis. |
| • | Optimization (week/year) sets the number of weeks or years to use for the optimization runs. |
| • | Out of sample (week/year) sets the number of weeks or years to use for the out of sample runs. |
| • | Retain profits can save the equity from one out of sample test to the next. Setting to false will reset the equity for each out of sample run. In all cases, trades are closed at the end of each out of sample run. |
| • | Show All Runs will show the summary results for all optimization runs, and out of sample runs that are not the optimal. When set to false (default) only the optimal out of sample runs will be displayed. |
| • | Show Stitched Equity Curve will present a chart with the combined equity curves from each of the OOS runs. |
The optimal run from the optimization runs is determined by the maximum goodness value. The goodness statistic is set in preferences, and can be any of the built in statistics, or custom statistics.