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

Re: [amibroker] Re: Fitness Criteria that incorporates Walk Forward Result



PureBytes Links

Trading Reference Links

Can't find the post where Howards writes that IS metrics are worthless...


I think this statement may be true for trading systems with few trades. However when your system makes 1000+ trades and produces a nice equity curve without many, if any, optimizations, I doubt that many developers would throw it out. Especially if such a system would exhibit broad performance, i.e., perform well on a fair number of securities.


imo, A few additional factors to consider are the number of trades, testing period, number of optimizations/iterations, look of equity curve, and whether the system works on more than just a few securities.


best regards,

herman


For tips on developing Real-Time Auto-Trading systems visit:

http://www.amibroker.org/userkb/


Thursday, May 8, 2008, 3:51:20 AM, you wrote:


> You might be interested in another idea I have played with:


> I agree with Howard that the IS metrics are of no value whatsoever 

> BUT what if the OOS validates that top model (or top X no. of models).


> Doesn't that then also validate the IS metrics.

> If so, can we learn anymore, or get more out of the exercise by 

> adding them and treating them as one sample (N has now increased and 

> we have the variance between two valid samples and/or one larger 

> valid sample to play with).


> (sorry I can't post anymore on it at this stage)


> brian_z



> --- In amibroker@xxxxxxxxxxxxxxx, "brian_z111" <brian_z111@xxx> wrote:


>> Paul,


>> I would love to talk over your proposition in more depth but as you 

>> can guess, I am busy at the moment.


>> Keep in mind that I assume I understand what you are talking about 

>> which is not always the case (without some concrete examples)/


>> Also I am not certain what you mean re the geometricAve of Ln IS & 

>> OOS? 


>> A quick observation:


>> I like that you are pushing yourself in this area - IMO while the 

>> commentators books (Pardo, Aronson, Bandy etc) and Fred's work give 

>> us an excellent base to work from there is still more to learn in 

>> theory, and application, and we should keep picking away at it.


>> Yes, I am specialising in this area and I am fine with your idea 

> e.g. 

>> say we 'optimise' 100 combinations and 5 pass a minimum value for 

>> our 'goodness' test.


>> There is no reason we can't OOS those five and then select one of 

>> them based on say Walk Forward Efficiency (refer to Pardo for WFE 

>> definitions ).


>> Also there is no reason the range of WFE for the five OOS tests 

> can't 

>> be treated as another kind of sensitivity analysis (note I haven't 

>> had time to study Fred's work yet so if I am merely repeating 

>> something he has already said then apologies).


>> WFE is only one example - if we put our thinking caps on we could 

>> find more (can WFE be applied to other metrics beside netprofit as 

>> used by Pardo?).


>> brian_z





>> --- In amibroker@xxxxxxxxxxxxxxx, "Paul Ho" <paultsho@> wrote:

>> >

>> > The purpose of OOS is to make sure there is no over fitting of 

>> data. So all

>> > optimization is done on In sample data. It is always possible to 

>> have more

>> > than 1 set of optimized parameters, because of  different 

> markets, 

>> or

>> > different selection of stocks, different parameters being 

> optimized 

>> or even

>> > different fitness criteria. I currently run my optimzation nearly 

>> 3000

>> > tickers, generating thousands of trades. There needs to be a 

>> systematic way

>> > of choosing the "right" system and I strongly argue that OOS has 

> a 

>> major

>> > role in that. I think it is not invalidating the OOS because, the 

>> amount of

>> > data mining is very small compared to insample. I think by just 

>> examing it

>> > casually, OOS is under utilised, at least in my case. From what I 

>> hear, a

>> > lot of people only optimize on individual or just a few tickers 

> and 

>> the

>> > degree of freedom is comparatively low. What I do here is to run

>> > optimization over the whole ASX exchange, past and present. 

>> > Finally, I think you should consider automating the process in IO 

>> and allow

>> > the user a  choice. 

>> > Cheers

>> > Paul

>> > 

>> > 

>> >   _____  

>> > 

>> > From: amibroker@xxxxxxxxxxxxxxx 

> [mailto:amibroker@xxxxxxxxxxxxxxx] 

>> On Behalf

>> > Of Fred Tonetti

>> > Sent: Thursday, 8 May 2008 12:12 AM

>> > To: amibroker@xxxxxxxxxxxxxxx

>> > Subject: RE: [amibroker] Re: Fitness Criteria that incorporates 

>> Walk Forward

>> > Result

>> > 

>> > 

>> > 

>> > 

>> > 

>> > 

>> > Paul,

>> > 

>> > 

>> > 

>> > One other word of caution .

>> > 

>> > 

>> > 

>> > If you are using OOS testing to drive the selection process of 

>> parameters to

>> > be used in sample then you run the risk of invalidating the OOS.

>> > 

>> > 

>> > 

>> > I could have automated this process in IO but I didn't for 

> exactly 

>> this

>> > reason.

>> > 

>> > 

>> > 

>> > 

>> >   _____  

>> > 

>> > 

>> > From: amibroker@xxxxxxxxxxxxxxx 

> [mailto:amibroker@xxxxxxxxxxxxxxx] 

>> On Behalf

>> > Of Paul Ho

>> > Sent: Wednesday, May 07, 2008 8:37 AM

>> > To: amibroker@xxxxxxxxxxxxxxx

>> > Subject: [amibroker] Re: Fitness Criteria that incorporates Walk 

>> Forward

>> > Result

>> > 

>> > 

>> > 

>> > Hi Fred

>> > Yes, I want to use the composite fitness to compare different 

>> systems 

>> > and or use it as a feedback in deciding on different parameter 

> sets 

>> > of the same system, This is not too dissimilar to how sensitivity 

>> > analysis is incorporated into the fitness criteria. The only 

>> > difference is that sensitivity analysis during optimization, and 

>> walk 

>> > forward is done after a new fitness high is found. Instead of 

> using 

>> > the insample fitness as the selection criteria for the best fit 

>> > system, the composite is criteria is used to choose among the 

>> various 

>> > peak values in one system or in different systems.

>> > 

>> > What you said "the capability to automatically reoptimize when 

> some 

>> > condition related to the performance metrics occurs during the 

> out 

>> of 

>> > sample period i.e. MDD goes beyond some static threshold or when 

> it 

>> > goes beyond some relationship to the same" is particularly 

>> > interesting. Because you are addressing a similar problem but 

> using 

>> a 

>> > different method, in your case, you change the time frame and 

>> > reoptimize. In my case, I am looking at refining my fitness 

>> criteria 

>> > so I might end up in choosing a different optimized parameter set 

>> in 

>> > the same time frame.

>> > 

>> > Paul.

>> > 

>> > --- In amibroker@xxxxxxxxx <mailto:amibroker%40yahoogroups.com

>> ps.com, Fred

>> > Tonetti <ftonetti@> wrote:

>> > >

>> > > Paul,

>> > > 

>> > > 

>> > > 

>> > > I understand what you are saying but I'm not sure what you do 

>> with 

>> > the

>> > > combined fitness when you get it . Do you use it to compare 

>> > different

>> > > systems to each other ?

>> > > 

>> > > 

>> > > 

>> > > Personally from the perspective of multiple automated WF's I am 

>> more

>> > > interested in . When to reoptimize .

>> > > 

>> > > 

>> > > 

>> > > IO already has the capability to reoptimize based on:

>> > > 

>> > > 

>> > > 

>> > > - Some static amount of time occurring during the OOS i.e. 

>> > > 

>> > > 

>> > > 

>> > > //IO: WFAuto: Rolling: 2: Weeks

>> > > 

>> > > 

>> > > 

>> > > - or in some undefined amount of time based on some number of 

>> > long/short

>> > > entries/exits etc i.e. 

>> > > 

>> > > 

>> > > 

>> > > //IO: WFAuto: Rolling: 2: LongEntrys

>> > > 

>> > > 

>> > > 

>> > > What I've been playing with recently is something a little 

>> > different that is

>> > > also based on a variable amount of time in the OOS i.e. the 

>> > capability to

>> > > automatically reoptimize when some condition related to the 

>> > performance

>> > > metrics occurs during the out of sample period i.e. MDD goes 

>> beyond 

>> > some

>> > > static threshold or when it goes beyond some relationship to 

> the 

>> > same or

>> > > different performance metrics of in sample.

>> > > 

>> > > 

>> > > 

>> > > For example . 

>> > > 

>> > > 

>> > > 

>> > > Assume the In Sample Performance Metrics are prefaced by IS and 

>> Out 

>> > of

>> > > Sample Performance Metrics are prefaced by OS then one should 

> be 

>> > able to

>> > > write ( in terms of IO Directives )

>> > > 

>> > > //IO: WFAuto: Rolling: Condition: OSMDD > 10 or OSMDD > 0.75 * 

>> ISMDD

>> > > 

>> > > 

>> > > 

>> > > In reality I suspect this is what most people actually do i.e. 

>> find 

>> > some

>> > > yardstick(s) that tell them their system is broken or about to 

> be 

>> > broken and

>> > > then reoptimize at that time.

>> > > 

>> > > 

>> > > 

>> > > 

>> > > 

>> > > _____ 

>> > > 

>> > > From: amibroker@xxxxxxxxx <mailto:amibroker%40yahoogroups.com

>> ps.com

>> > [mailto:amibroker@xxxxxxxxx <mailto:amibroker%40yahoogroups.com

>> ps.com] 

>> > On Behalf

>> > > Of Paul Ho

>> > > Sent: Tuesday, May 06, 2008 10:41 AM

>> > > To: amibroker@xxxxxxxxx <mailto:amibroker%40yahoogroups.com

>> ps.com

>> > > Subject: [amibroker] Fitness Criteria that incorporates Walk 

>> > Forward Result

>> > > 

>> > > 

>> > > 

>> > > Howard calls it the objective function. Fred calls it Fitness. 

>> What 

>> > I 

>> > > meant by Fitness Criteria is a mathematical function on which 

>> > fitness 

>> > > or goodness of the system is judged, and is used as an 

> objective 

>> > > criteria to compare different systems, as a score in 

>> optimization. 

>> > > 

>> > > My currrent question is - So why not incorporate the fitness in 

>> > walk 

>> > > forward analysis into our fitness criteria? What I am talking 

>> about 

>> > > is to formalise the visual inspection process. I am not 

> proposing 

>> > to 

>> > > use out of sample data for optimization purposes. Rather the 

>> > > parameter set that has been previously optimized is forward 

>> tested 

>> > > and a fitness is obtained and incorporated into the original 

>> > criteria 

>> > > to form a composite fitness. 

>> > > 

>> > > For example. My current composite fitness is the geometric 

>> average 

>> > of 

>> > > In sample fitness and Out of sample fitness divided by the 

>> standard 

>> > > deviation (?) of In sample and out of sample fitness. 

>> > > 

>> > > Are there anybody doing something is this area? What are your 

>> > > thoughts?

>> > > 

>> > > If you are wondering why not use visual inspection. My plan is 

> to 

>> > use 

>> > > the computer to do most of the work and thats why I need a 

>> fitness 

>> > > criteria.

>> > > 

>> > > Cheers

>> > > Paul.

>> > >

>> >





> ------------------------------------


> Please note that this group is for discussion between users only.


> To get support from AmiBroker please send an e-mail directly to 

> SUPPORT {at} amibroker.com


> For NEW RELEASE ANNOUNCEMENTS and other news always check DEVLOG:

> http://www.amibroker.com/devlog/


> For other support material please check also:

> http://www.amibroker.com/support.html

> Yahoo! Groups Links


> <*> To visit your group on the web, go to:

>     http://groups.yahoo.com/group/amibroker/


> <*> Your email settings:

>     Individual Email | Traditional


> <*> To change settings online go to:

>     http://groups.yahoo.com/group/amibroker/join

>     (Yahoo! ID required)


> <*> To change settings via email:

>     mailto:amibroker-digest@xxxxxxxxxxxxxxx 

>     mailto:amibroker-fullfeatured@xxxxxxxxxxxxxxx


> <*> To unsubscribe from this group, send an email to:

>     amibroker-unsubscribe@xxxxxxxxxxxxxxx


> <*> Your use of Yahoo! Groups is subject to:

>     http://docs.yahoo.com/info/terms/


__._,_.___

Please note that this group is for discussion between users only.

To get support from AmiBroker please send an e-mail directly to
SUPPORT {at} amibroker.com

For NEW RELEASE ANNOUNCEMENTS and other news always check DEVLOG:
http://www.amibroker.com/devlog/

For other support material please check also:
http://www.amibroker.com/support.html




Your email settings: Individual Email|Traditional
Change settings via the Web (Yahoo! ID required)
Change settings via email: Switch delivery to Daily Digest | Switch to Fully Featured
Visit Your Group | Yahoo! Groups Terms of Use | Unsubscribe

__,_._,___