Ok, I was talking abour another problem. What does mean in the chart the lines Buy&Hold IS and Buy&HoldOOS? For I have the same amount of dollars in OOS equity and Buy&Hold equity.. And the usual behaviour for just Buy and hold is to be less than a good system.. there's something I misunderstand..
bye --- In [email protected], Thomas Ludwig <[EMAIL PROTECTED]> wrote: > > > > Thanks Thomas. Very good collection of parameters. > > I see my system falls from CAR 64 in sample, to CAR 30 out-of-sample. > > But I think something must be wrong: The result of the buy and hold > > is exactly the same that the OS result. > > Do you have this problem? > > That OOS results are worse than IS results is normal. You should also > bear in mind that the concatenated IS period is not identical to the OOS > period. That's why IS and OOS CARs are not directly comparable. Example: > If you start your walk-forward optimization on 01/01/2000 with a, say, > 2-year IS period it's obvious that both CARs are very different. That's > why in my code Walk-forward Efficiency (WFE) is NOT calculated as OOS- > CAR/IS-CAR (as Robert Pardo suggests) but rather as RELATIVE IS- > CAR/RELATIVE OOS-CAR (where Relative CAR is IS-CAR or OOS-CAR divided by > the benchmark CAR for the respective period). This makes more sense, > IMHO. > > > And, I had already seen the gfx code in the Amibroker page. It's > > fine, but it does not show DD, and total CAR.. although I didn't test > > the trick of changing the code to see OS equity line..very good.. > > I think you should use both my and TJ's code - they save different > purposes. MaxDD as a percentage is shown for IS and OOS in my code. And > if you right-click the pane and open the parameters window you can also > plot DD and MaxDD as absolute figures both for IS and OOS. But I strongly > suggest that you plot it only for, e.g., OOS (and disable all the IS > plots - or vice versa!) - otherwise the pane becomes too cluttered. > Particularly if you also plot IS and OOS Relative Performance ... ;-) > > Greetings, >
