[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[amibroker] Re: Margin of Error



PureBytes Links

Trading Reference Links

Isn't that what the psychology of trading is about ? ... Having as 
many comfort blankets as possible ? ... Without them one in this case 
will typically suffer from one of two things, inability to pull the 
trigger or unexpected losses after which one will suffer from the 
former ... IMHO it is best to have a pretty good idea what to expect 
OOS before one trys it using real money.  Anyone who doesn't want to 
take the trouble can look at their IS performance metrics and take 
half the CAR and double the DD's and HOPE their OOS experience is 
that good ...

--- In amibroker@xxxxxxxxxxxxxxx, "brian.z123" <brian.z123@xxx> wrote:
>
> O.K
> 
> I am not disagreeing on the value of OOS/WF testing in practise.
> Theoretically though, if we push the boudaries, it is only a 
comfort 
> blanket.
> 
> I will attempt to defend that position, as an interesting point, in 
> my next two posts.
> 
> I would remind our younger viewers not to attempt this at home.
> 
> BrianB2.
> 
> --- In amibroker@xxxxxxxxxxxxxxx, "quanttrader714" 
> <quanttrader714@> wrote:
> >
> > I'd rephrase the question to, what method or methods will tell us 
> how
> > the system is likely to perform *when traded in real time*?  Short
> > answer IMO is OOS testing and analysis.
> > 
> > P.S. MCS not a crystal ball or magic.  It's just a tool that uses
> > brute force to estimate what is difficult or impossible to 
> calculate
> > otherwise and garbage in, garbage out definitely applies. 
> >  
> > --- In amibroker@xxxxxxxxxxxxxxx, "brian.z123" <brian.z123@> 
wrote:
> > >
> > > Hello Fred,
> > > 
> > > Precisely.
> > > 
> > > I'm not going crazy after-all!
> > > 
> > > What method or methods will tell us how the system is likely to 
> > > perform out of sample; since in the end system trading is 
> nothing 
> > > but a perpetual walk forward test?
> > > 
> > > BrianB2.
> > > 
> > > --- In amibroker@xxxxxxxxxxxxxxx, "Fred" <ftonetti@> wrote:
> > > >
> > > > While MCS is a good tool for validating some things it is not 
> a 
> > > > substitute for out of sample and/or walk forward testing ... 
> If 
> > > for 
> > > > example I:
> > > > 
> > > > - Write a system
> > > > - Test it to make sure the rules are working as intended 
> > > > - Optimize the variables to have the system produce the best 
> > > results 
> > > > it can based on some metric or metrics within the confines of 
> the 
> > > > rules 
> > > > - Use an MCS on the trades that are generated  
> > > > 
> > > > This tells me nothing about how the system is likely to 
> perform 
> > > out 
> > > > of sample.  It only tells me about the statistics related to 
> the 
> > > > optimized rules of the system which are the result of 
> scrambling 
> > > the 
> > > > order of the trades that resulted from using the same in 
> sample 
> > > data 
> > > > that system was optimized on.
> > > >
> >
>




Content-Description: "AVG certification"
No virus found in this incoming message.
Checked by AVG Free Edition.
Version: 7.1.409 / Virus Database: 268.14.11/542 - Release Date: 11/20/2006