Walk forward does WHAT?

[raycam]

I cant seem to get my head around walk forward. I have checked out the links and end up just shaking my head.
I have been perfecting two strategies one for longs and one for shorts on one market. I first back~tested on about 9 weeks data and used the strategy analyzer to optimize till I was real happy with a nice profitable bundle.
Then I bought 5 years worth of data and have back~tested (at this stage to the beginning of the year) using the same optimization to get to a figure considerably less per month- but still a decent income.
Now I read forum posts that talk about walk forward and how that gives different results ( poorer).
What is it all about? How does it work?
What parameters do I use. In the end I want a stable automated trading strategy that will come close to replicating ( within reason) the back and forward testing results.
raycam

[NinjaTrader_Bertrand]

raycam, I'm not sure I follow where the confusion stems from...walk forward optimizatation is a rolling process, you optimize on backdata and determine which paramters are best on this set (like you would do in realtrading, for example optimizing on last 2 weeks). Then you move on to applying those best parameters (according to your 'Optimize On' criterion) to a unknown (out of sample) data set to test how they hold up...this dataset becomes then the next in sample set to ge the 'next best set' of paramters for the following out of sample set...and so on until you're testing period is completely covered. You can then examine how good the strategy holds up on unseen data to judge stability / robustness if parameters vary.

[raycam]

Thanks. So what would be an adequate sample period- 2 weeks, a month?
And do I have my variables in play in the live market- eg- a stop~loss set to between 5 and 15 ticks? At the end of- or during the live walk forward I would be able to see which ones need tweaking to move forward more profitably?
Is that it?
Id love a video on it!
Also is it only accessible through the strategy analyzer? If so do I have to set it up each day?
Sorry about all the questions.
raycam

[NinjaTrader_Bertrand]

raycam, correct this is for the Strategy Analyzer only - as it is a backtesting tool.

Thanks for the video suggestion, I forward it to our educational department.

There are no correct / false settings, you will need to test what is adequate for your market and timeframe, for a shortterm intraday system running on < 5 min data, I'd say 2-3 weeks should do it.

For the stops / targets you want to check the MAE / MFE graphs in the report.