Walkforward Question

[Buttermilk]

When an optimization is done, a number of results are generated.
"Keep Best # Result"
The Walkforward Optimizer takes, I asume, the "best" and use it to trade the out of sample the out of sample data and then move forward. My question is how do the WalkForward systems select the "best" results for Walking forward.

[Baruch]

In WF as in Optimizer you optimize on something, like "Net Profit"...
It calculates the performance on this method and the best performance you trade.

Baruch

[NinjaTrader_Bertrand]

Thanks Baruch - the 'Optimize On' would be the performance measure used then - http://www.ninjatrader-support.com/H...AStrategy.html

[Nomar]

More WalkForward Questions

This looks like the right place to ask my questions, since searching the forum and Help Documentation didn't yield much, so here goes:

I am Optimizing/WF using 200days/100days for splitting the data, and holding the top 100 best results.

In the WF results, I see that the Days allocated for WF Optimization are divided into 3 or more segments with unique date ranges. I assume that this means that some form of Rolling WF Method is being applied. That being the case, my questions are:

1. In the WF process, are the same 100 optimization results applied to each WF data range to determine the best fitness ?

-or-

2. In the WF process are the Best 100 Optimization results applied to the first data range, the results of the first applied to the second data range, and finally the results of the second applied to the third?

Depending on how this process is implemented, then:

3. Do I use the best result of the 3 data range segements (profitable or not) to test in realtime ?

-or -

4. Do I use the results from the last segment (profitable or not) since it has gone sequentially through multiple data range segement WF tests ?

My apologies for being so detailed, but I wanted to be as clear as possible.

Thanks in Advance

[NinjaTrader_Bertrand]

Nomar, it's a rolling process - the best result ('optimize on') is used to trade the next slice of data until the full data range specified is traded. The WalkForward tab then displays the used parameters for each slice and their performance, there's no general conclusion given which parameter set is the best to use then in realtime for the next trading day, but you get a sense how your strategy is holding up for different parameter sets.

The Instruments tab lists the result of the full data range with the default paramters used.

[Nomar]

Bertrand,
Thank you for your prompt reply. The information you provided helps clear up the mystery.

The only question I have left is:

I can see in the WF Tab that the 1st slice uses the Best Performance from the Optimization Tab (the parameters track).

But the 2nd & 3rd slice use parameter sets that are further down the list of Performance (worse) and not the same Best Performance as used in the 1st slice.

So my question is:

How are the parameter sets selected for WF testing of slices 2 thru n ?

Thank You for your patience.

[NinjaTrader_Bertrand]

The same procedure is used throughout the walk forward process - if you compare the listed optimization results for the different 'slice' periods, you notice the best performing parameter set is used to trade the next period for evaluation (depending on your chosen performance criteria).