Hi Howard, You did correctly interpret that my logic and parameters are all hard coded in this system. What escaped me in this case was the role of the objective function in the Walkforwad process. Thanks for setting me straight on this.
Regards, Ray --- In amibroker@yahoogroups.com, Howard B <howardba...@...> wrote: > > Hi Ray -- > > I would interpret your statement "has no parameters to optimize" to mean > that all of the choices of logic and parameters have been hard-coded into > the algorithm. > > "To optimize" means, to me, "to generate a large number of alternatives, > then choose the best based on a metric that describes my preferences". The > emphasis is on "to choose the best based on a metric that describes my > preferences". "To optimize" is not fully defined without specification of > the objective function that will be used to rank the alternatives. > > If, indeed, the algorithm you refer to in your posting in the only one that > is being considered, then the generate and choose tasks have already been > done. And the choosing was done using some metric, even if it was > subjective. > > The walk forward process is important in validation of trading systems. It > incorporates optimization. Among the benefits of walk forward testing are > the many examples you get to see of the transition from testing to trading, > and gathering a set of performance results that can be used as an estimate > of future trading results and as a baseline for comparison of real-time > trading. If you go directly from development to trading, you will be > performing the out-of-sample tests with real money, you will be observing > the transition from testing to trading with real money, and you will have no > baseline by which to compare results. If you are comfortable doing that, > then you can skip the validation. > > You might consider an alternative. > > Review the system you described. Select some logic and / or parameters that > might be varied, with each variation being an alternative trading system. > Write optimize statements to control those and run walk forward tests. If > your first idea was, in fact, the best, then it will appear consistently at > the top of the list in each walk forward step. You will have spent some > amount of time but no money, you will have increased your confidence in your > system, and you will have some estimates of expected performance. > > Thanks for listening, > Howard > > > On Sun, Aug 15, 2010 at 8:21 PM, raymondpconnolly < > raymondpconno...@...> wrote: > > > > > > > Hello all, > > > > I've designed a system that has no parameters to optimize. For this special > > case would a backtest be materially equivalent to a walkforward or perhaps > > even superior since there are no boundary distortions? > > > > If necessary I've added a dummy variable to allow an optimization for the > > walkforward but I would be interested in opinions on this special case > > situation in case the backtest is sufficient for validation. > > > > Regards, > > Ray > > > > > > >
