Question about walk forward optimization

Discussion in 'Strategy Building' started by travis, May 29, 2003.

  1. travis


    I have Dennis Meyers' software, but not Optilogix, which is another software that does it.

    My problem with Meyers' software is that it calls itself "walk forward", but it doesn't seem to deliver on its promise and the manual's promise, because the two periods - sample data (on which the optimization is done), and out-of-sample data (on which the optimized system is tested) - are sliding only in size, which makes the testing useless, and there isn't a a box where to set start date for the sample data, nor a box where to set end date for the out-of-sample data.

    So, instead of having 2 "fixed sliding periods", one to optimize the system, the other to test the system on, we just have a "sliding" date, dividing the whole data set in two periods of varying sizes (out-of-sample gets smaller as sample increases). Now, since the manual prescribes a testing with two "fixed sliding periods", I don't understand why the software doesn't allow it. And I know I could maybe do it by modifying the signal, but it's not easy, and then I don't understand the purpose of a software that has to be rewritten to be used at all.

    Has anyone used this software or a different software and can you answer my questions, please.