I'm doing some tests on my system through the walk forward optimizer and I have been noticing some interesting behavior that I cannot find the answer to.
While doing some low parameter range WF's (such as a single period parameter), the performance results were drastically different. Although unlikely, I thought that my system may have been overtly overfit to the data being used.
Still not convinced, I decided to run the WF with no parameter variables which would match my backtest exactly. The WF results however, were very different.
The average trade was drastically reduced and the number of trades for the sample was increased by about 50%.
The theory that I have is that the WF test will close the trade at the end of the test period. This would explain why there are more trades and a lower average trade.
Is that correct? Does the WF close the trades if it is being held over the test period and does the trade re-open or wait until the next signal?
Thank you!
Comment