Hi Keith, this is exactly the direction I was searching for. As I understood ranking of the result is not possible, but I could create a new ratio (as you explained) and use this factor for my tests.
Many thanks and kind regards, dubi --- In amibroker@yahoogroups.com, Keith McCombs <kmcco...@...> wrote: > > Dubi -- > You can make up your own metric. See addcustommetric() and custom back > test, CBT, in the users guide and on amibroker.com. > http://www.amibroker.com/docs/ab401.html > > For example, say you want maximum UPI, but you also feel that you would > prefer number of trades to be near 5. > OptTrades = 5; > MultFactor = 2 * OptTrades; > Ntrades = (available via GetValue(string MetricName) > MyMetric = UPI * (MultFactor * Ntrades - Ntrades^2); // just one way of > many > > -- Keith > > dubi1974 wrote: > > > > > > Hi Howard and all, > > > > yes SPSO is the inbuild particle optimizer.... When I optimize for > > Ulcer Index I do not take everytime the best choice, as I do not do it > > when I optimize for %Net Perf or CAR/MDD. Eaxmple: If I have let's say > > 3 parameters 100x100x100 I would get 1mn options. If I optimize for > > %Net Perf with SPSO the list would be about 5000-7000 results. But as > > I do not want to choose the best %Net Perf but the highest %Net Perf > > with a high UPI, a high CAR/MDD, high RRR, low StDev I would decide to > > use the parameters for maybe the 2nd or 3rd etc. rated result. But > > this is not possible if I directly use the inbuild WalkForward Test of > > Amibroker, as the system uses always the first choice and not the best > > choice. > > > > If I would optimize directly for RRR, I got sometimes as first choice > > just 1-2 trades as the StDev was the lowest there, but I would never > > use this parameters. That is why I was searching for an alternative. > > > > Many thanks and kind regards, > > > > dubi > > > > --- In amibroker@yahoogroups.com <mailto:amibroker%40yahoogroups.com>, > > Howard B <howardbandy@> wrote: > > > > > > Hi Bisto -- > > > > > > Thanks. From the original posting, I was thinking that SPSO was a > > metric. > > > > > > Dubi -- > > > > > > The choice of optimization method (exhaustive, cmae, ...) and metric > > (Ulcer > > > Index, CAR/MDD, k-ratio, ...) are independent. The optimization > > portion of > > > the walk forward will be carried out using the optimization method you > > > request in you afl code with the OptimizerSetEngine statement. The > > > selection of which of the alternatives tested to use for the > > out-of-sample > > > run will be determined by the metric you set in the Optimization > > Target box > > > on the Walk Forward tab of the AA Settings. > > > > > > Or did I misinterpret your question? > > > > > > Thanks, > > > Howard > > > > > > > > > On Fri, Nov 13, 2009 at 3:39 AM, Bisto <bistoman73@> wrote: > > > > > > > > > > > > > > > I suppose: Standard Particle Swarm Optimization > > > > > > > > Bisto > > > > > > > > > > > > --- In amibroker@yahoogroups.com > > <mailto:amibroker%40yahoogroups.com> <amibroker%40yahoogroups.com>, > > Howard B > > > > <howardbandy@> wrote: > > > > > > > > > > Hi Dubi -- > > > > > > > > > > What is SPSO? > > > > > > > > > > Thanks, > > > > > Howard > > > > > > > > > > On Wed, Nov 11, 2009 at 9:07 AM, dubi1974 <gonzales74@> wrote: > > > > > > > > > > > > > > > > > > > > > > > Hi! > > > > > > > > > > > > Is it possible to optimize a system (for e.g. best net > > performance %) > > > > with > > > > > > SPSO but then use the highest Ulcer Performance Index or > > CAR/MDD and > > > > use > > > > > > then this parameters for the out of sample test for the next > > period in > > > > the > > > > > > Walk Forward Test of Amibroker? I know I could do this in separate > > > > steps, > > > > > > but I would like to automize that. Or does someone know an other > > > > solution > > > > > > for this? > > > > > > > > > > > > Many thanks and kind regards, > > > > > > > > > > > > dubi > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > >
