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

[amibroker] Re: testing multiple systems simultaneously



PureBytes Links

Trading Reference Links

I'm pinching myself.

I must be dreaming.

Two years of nagging the forum and I couldn't get anyone to agree with me before.

The best case use of system eq curve analysis is to compute the opt f eq curve and analyse that OR the partial opt f curve, if that is where you run.

Still.... everyone of us is managing a portfolio ... we are always in at least two assets ... cash and 1 system, unless we are not trading and are 100% in cash.


--- In amibroker@xxxxxxxxxxxxxxx, i cs <ics4mer@xxx> wrote:
>
> Hi Brian,
> 
> I have to strongly agree on drawdown not being a useful metric to 
> estimate risk. Its a chimera. It's a once off/unique trade sequence which is
> unlikely to occur again.
> 
> 
> 
> RZ
> 
> 
> 
> 
> ________________________________
> From: brian_z111 <brian_z111@xxx>
> To: amibroker@xxxxxxxxxxxxxxx
> Sent: Wednesday, 13 May, 2009 12:41:59 PM
> Subject: [amibroker] Re: testing multiple systems simultaneously
> 
> 
> 
> 
> 
> > Anecdotal observations from my trades are that it does smooth out >the 
> > equity curve rather  nicely.
> 
> For the benefit of others (IMO):
> 
> - portfolio eq is the only eq curve worth analysing BUT the math has already predicted it anyway, although seeing is believing!
> - eq curve analysis at the product (system) level is a waste of time ... better to plug the systems into a portfolio analysis and see what comes out at the P level ... then learn from that what type of systems are needed to produce 'good' portfolio curves.
> - system eq curve 'smoothness' can be predicted by Core Metric Evaluation
> - I will come up with a post, with a predictive ratio in it later ... working on it now on a part time basis ... those interested to race ahead should look at arithmetic mean and variance of the trade series as predictors of geometric mean variance ---> eq curve variance
> - drawdown is not the correct metric to use to estimate risk.
> 
> --- In amibroker@xxxxxxxxx ps.com, i cs <ics4mer@ > wrote:
> >
> > Hi Brian,
> > 
> > Yes, I have been using the optimal f on some of my portfolio management.
> > I haven't been using it more simply because I had no automation and the
> > workload did pile up. And I was trying a whole bunch of other leads 
> > which led to dead ends unlike the optimal f portfolio.
> > 
> > Anecdotal observations from my trades are that it does smooth out the 
> > equity curve rather  nicely. Talking about  300 observations over 6 months 
> > on intraday trades, although this is possible too small a sample.
> > 
> > 
> > For a gentle way into the process I would  recommend trading the 
> > optimal relationship at the portfolio level, and suboptimal at the 
> > product level - you sleep much easier!
> > 
> > As for other theories - I agree - there doesn't seem to be a lot of 
> > alternatives out there - and I am a bit cautious about the brute force
> > approaches offered by non-exhaustive optimisation. It's a great tool
> > and there are some essential uses for it, but if the only tool you
> > have is a hammer, everything starts looking like a nail. 
> > 
> > On the other hand, I think that theory and empirical observation
> > do go together well. One should push the other. Being too much 
> > in either camp certainly precludes you from seeing what "truths" are 
> > really out there. An example of  a very successful analysis technique
> > (signal predictive, not portfolio related) is the Hilbert-Huang algorithm. 
> > This is a type of Empirical Mode Decomposition (EMD) that was 
> > developed and used at NASA. (Huang did additional work to the 
> > original Hilbert equations). It is interesting that NASA describes the 
> > equation as "totally empirically proven" but as yet they still have no 
> > "theoretical basis whatsoever".
> > 
> > Anyway, dying to delve more into more portfolio work with AB but 
> > since I'm a newbie, I'm paying my dues at the moment by putting in 
> > time trying build some facility with loops loops loops loo sllps  llspolsl 
> > slls als  oops..
> > 
> > Robert Z
> > 
> > 
> > 
> > 
> > ____________ _________ _________ __
> > From: brian_z111 <brian_z111@ ...>
> > To: amibroker@xxxxxxxxx ps.com
> > Sent: Wednesday, 13 May, 2009 8:51:51 AM
> > Subject: [amibroker] Re: testing multiple systems simultaneously
> > 
> > 
> > 
> > 
> > 
> > > Surprised that nobody seems to have suggested modern portfolio >theory (MPT) in this context.
> > 
> > Yes, I was also surprised that no one, who is trading multiple systems, volunteered any info about the theoretical basis behind what they are doing (except Robert who uses opt f, but he didn't say if he extends that to the portfolio level).
> > 
> > I did mention Markorwitz, because of it's synchronization with Vince's portfolio models (The cover blurp, on his second book claims that "The Mathematics Of Money Management provides the missing element in modern portfolio theory that weds optimal f to the optimal porfolio"). 
> > 
> > I have only just got started on portfolio theory and so far I find that Vince's ideas stand up to rigorous testing and cross checking, contrary to what some people say e.g.  William R Gallacher, in his book "Winner Take All" says something to the effect that Vince is full of math theory but that he doesn't give one example of its application to a real trading situation and that his ideas have no relevance to 'real life' trading and don't work!
> > 
> > My instincts tell me that Vince's work is very good and highly relevant.
> > 
> > The best thing about his work is that the books track the development of his ideas sequentially so one can follow along as he develops the ideas, over time, and revisits his earlier 'mistakes' (he had over ten years of obsessive wrestling with the maths behind trading).
> > 
> > I find his logic very strong but he doesn't always connect the dots for us and I guess that puts a lot of people off i.e. his books are more aligned to teaching us the principles rather than providing us with ready made recipes ... a lot of people seem to fall over at that point!
> > 
> > I am looking for some alternative models, to cross check against Vince, and wasn't sure where to start.
> > 
> > It looks like Markowitz (+ PostModernPortfolio Theory?) is still the benchmark? ... so I will start there.... no wonder they gave him a Nobel prize if his theories survived over a decade in the modern era.
> > 
> > I haven't got started on R yet ... the links you gave are excellent ... R seems to be getting up quite a head of steam (except in the AmiBroker community) ... the portfolio examples are very serious indeed ... they would be a challenge for me if I do get into it at that level.
> > 
> > Thanks,
> > 
> > brian_z 
> > 
> > --- In amibroker@xxxxxxxxx ps.com, "vlanschot" <vlanschot@ ..> wrote:
> > >
> > > Surprised that nobody seems to have suggested modern portfolio theory (MPT) in this context (I may have missed something.) Although it may not be everybody's favourite, it still offers a well-researched and practised starting point. In that light, individual trading systems are nothing more than producers of return time series, i.e. individual portfolios or assets, and can thus be analyzed in a composite portfolio via MPT. 
> > > 
> > > In terms of practical applications see, for example, the presentation by Guy Jollin "R tools for Portfolio Optimization" here: http://rinfinance. quantmod. com/presentation s/
> > > From there you can explore the R-metrics universe of portfolio applications: http://r-forge. r-project. org/R/?group_ id=156. All this is now available to the AB user via the RMath plug-in. 
> > > 
> > > Oh, one more point. Relative performance is relative: cash can obviously act as the benchmark for those who operate in absolute return space.
> > > 
> > > PS
> > > 
> > > 
> > > --- In amibroker@xxxxxxxxx ps.com, "bh.hicks" <bh.hicks@> wrote:
> > > >
> > > > I concur 100% on finding an "optimal" solution and agree there are solid mathematical models in place which serve as an excellent point of departure.  I for one am not interested in re-inventing the wheel, just looking for the best way to test and implement them with the limited time and programming ability I have at my disposal.
> > > > 
> > > > The whole idea of optimum is a tricky concept though.  As I'm sure you are aware, the optimum solution is always changing and evolving and thus, also highly susceptible to curve-fitting.  While I consider myself a quantitative trader, I tend to approach things with "fuzzy-logic" and try and conceptualize of solutions as bad, better, good but never "best".  I know that is a somewhat unsophisticated approach but I know this is an area if my trading that can be made much "better", as defined by more money in my account with less risk.  A little better is good but a lot better is even better ;)
> > > > 
> > > > I really appreciate your participation (and other's) in this dialog and am excited about watching it evolve over time.  I also have a lot of new reading to do!
> > > > 
> > > > 
> > > > 
> > > > --- In amibroker@xxxxxxxxx ps.com, "brian_z111" <brian_z111@ > wrote:
> > > > >
> > > > > <snip>the goal of this thread was to find a way to exploit my real-time
> > > > > observations regarding lower account drawdowns when trading multiple strategies
> > > > > than what I would have experienced if traded them independently. Most of my
> > > > > opinions on this are admittedly 100% anecdotal but they are at least based on my
> > > > > real-time observations with capital at-risk.
> > > > > 
> > > > > I am sure there are probably better or more academically sound ways of doing
> > > > > this but my own personal preference is to develop the systems independently
> > > > > using whatever criteria I deem important and the only optimizations done at the
> > > > > multi-system level would be the tweaking of position size to allow me to trade
> > > > > each a bit more aggressively while staying below my pain threshold.<snip>
> > > > > 
> > > > > I am working on improving my knowledge of 'portfolio efficiency' so I appreciate the feedback provided by people like Angelo and yourself who have 'real life' experience of trading multiple systems.
> > > > > 
> > > > > Based on my theoretical considerations and 'bench testing, I think it is unlikely that we will find the optimum solution (whatever our objectives are) by:
> > > > > 
> > > > > - relying on observation and experimentation in real life trading
> > > > > - blindly optimizing our way to it
> > > > > - hitting on some lucky allocations
> > > > > 
> > > > > IMO portfolio efficiency is an excercise in maths and the math models already exist.
> > > > > 
> > > > > There is no need to mimic Vince, or anyone else, but I believe that a consideration of the math models will quickly move us closer towards our personal goals.
> > > > > 
> > > > > 
> > > > > So far no one has suggested any portfolio model other than Vince's and the reference by Angelo.
> > > > > 
> > > > > 
> > > > > 
> > > > > 
> > > > > 
> > > > > 
> > > > > --- In amibroker@xxxxxxxxx ps.com, "bh.hicks" <bh.hicks@> wrote:
> > > > > >
> > > > > > Howard,
> > > > > > I saw your post and thank you for participating in this discussion.  I am eagerly awaiting your new book. Please let me know when it is available for pre-sale!
> > > > > > 
> > > > > > All three of the systems I currently trade were developed as stand-alone systems (on a different platform) and I have been trading all three for a couple of years now (with constant tinkering).  2 of the 3 were profitable in '08 and the 3rd was down only about 6%.  The long/short index system had a record year last year at 77% although it is under-performing thus far this year. 
> > > > > > 
> > > > > > This approach has given me an average CAGR of slightly over 30% for the past 3 years since I gave up discretionary trading.  I am sure most people here pull down numbers way better than that but my only point is that this passed the point of being an academic exercise for me a couple of years ago. 
> > > > > > 
> > > > > > I am however an extremely conservative trader with very specific rules for how I position size.  I went into it a little bit here...
> > > > > > 
> > > > > > http://finance. groups.yahoo. com/group/ amibroker/ message/138123
> > > > > > 
> > > > > > but the goal of this thread was to find a way to exploit my real-time observations regarding lower account drawdowns when trading multiple strategies than what I would have experienced if traded them independently.  Most of my opinions on this are admittedly 100% anecdotal but they are at least based on my real-time observations with capital at-risk. 
> > > > > > 
> > > > > > I am sure there are probably better or more academically sound ways of doing this but my own personal preference is to develop the systems independently using whatever criteria I deem important and the only optimizations done at the multi-system level would be the tweaking of position size to allow me to trade each a bit more aggressively while staying below my pain threshold.
> > > > > > 
> > > > > > Thanks again for your participation and I am eager to follow the development of this topic and your new book.
> > > > > > 
> > > > > > regards.
> > > > > > 
> > > > > > 
> > > > > > 
> > > > > > 
> > > > > > 
> > > > > > 
> > > > > > 
> > > > > > 
> > > > > > 
> > > > > > 
> > > > > > 
> > > > > > 
> > > > > > 
> > > > > > 
> > > > > > --- In amibroker@xxxxxxxxx ps.com, Howard B <howardbandy@ > wrote:
> > > > > > >
> > > > > > > Hi bh --
> > > > > > > 
> > > > > > > It was suggestions such as the one you made earlier in this thread that
> > > > > > > prompted me to write the posting on portfolio construction.
> > > > > > > 
> > > > > > > You wrote:
> > > > > > > "----------- --------- -------
> > > > > > > Out of curiosity ... are you finding it easy to come up with noncorrelated
> > > > > > > systems?
> > > > > > > 
> > > > > > > Perhaps not "easy" but possible. Without going into too many details, the 3
> > > > > > > systems I currently trade do the following.
> > > > > > > 
> > > > > > > System A - a long/short mean-reversion system that trades the S&P (ES or
> > > > > > > SPY) using short-term overbought/oversold levels. Average hold time is 3-5
> > > > > > > days, buys and sells the close.
> > > > > > > 
> > > > > > > System B - a long only mean-reversion system that trades the entire stock
> > > > > > > market universe with a minimum price and liquidity requirements. It
> > > > > > > essentially buys short-term weakness on longer-term high relative strength
> > > > > > > stocks. Average hold time is 3-5 days, buys and sells the open.
> > > > > > > 
> > > > > > > System C - a short only system that trades the entire stock market universe
> > > > > > > with a minimum price and liquidity requirements. I am very protective of
> > > > > > > this one because a short-only system that has an edge over the past 20 years
> > > > > > > through any market climate is rare but this system buys the open and sells
> > > > > > > the close of the same day.
> > > > > > > 
> > > > > > > I need a longer-term system (2-3 week hold times) that buys strength rather
> > > > > > > than weakness to try and fill in the under-performance gaps during
> > > > > > > significant market rallies like what we are having now. I have found this
> > > > > > > extremely challenging to do quantitatively.
> > > > > > > ------------ --------- -"
> > > > > > > 
> > > > > > > The intent of my posting was to point out that there are several complex
> > > > > > > issues that should be considered when creating a portfolio of trading
> > > > > > > systems.
> > > > > > > 
> > > > > > > They begin with analysis of the individual systems, asking questions such
> > > > > > > as:
> > > > > > > 
> > > > > > > What are the optimum lengths of the in-sample periods for the three systems
> > > > > > > you are describing?
> > > > > > > Will the portfolio be created (the portfolio weights determined, and so
> > > > > > > forth) by re-selecting the parameter values for those systems as part of a
> > > > > > > portfolio?
> > > > > > > Or, by using the trade results that were obtained by running each
> > > > > > > separately?
> > > > > > > Are the results that were obtained by running each separately in-sample or
> > > > > > > out-of-sample?
> > > > > > > And so forth.
> > > > > > > 
> > > > > > > Best wishes on the project (no sarcasm intended) -- let us all know how it
> > > > > > > turns out.
> > > > > > > 
> > > > > > > Thanks,
> > > > > > > Howard
> > > > > > > 
> > > > > > > 
> > > > > > > On Sat, May 9, 2009 at 12:53 PM, bh.hicks <bh.hicks@> wrote:
> > > > > > > 
> > > > > > > >
> > > > > > > >
> > > > > > > > Graham,
> > > > > > > > Let me think about specifically what my needs are regarding this over the
> > > > > > > > next few weeks and I may contact you. I am a little reluctant to pursue a
> > > > > > > > "custom" solution as I think there are others on this board way more capable
> > > > > > > > of designing a better and more robust solution than I but it may make sense.
> > > > > > > > Thank you.
> > > > > > > >
> > > > > > > >
> > > > > > > > --- In amibroker@xxxxxxxxx ps.com <amibroker%40yahoog roups.com> , Graham
> > > > > > > > <kavemanperth@ > wrote:
> > > > > > > > >
> > > > > > > > > It is possible to do, you just need to write this into the advanced
> > > > > > > > > backtest coding, low level coding is advisable. Takes a bit of work
> > > > > > > > > and ability to balance balls like a juggler.
> > > > > > > > >
> > > > > > > > > --
> > > > > > > > > Cheers
> > > > > > > > > Graham Kav
> > > > > > > > > AFL Writing Service
> > > > > > > > > http://www.aflwriti ng.com
> > > > > > > > >
> > > > > > > > >
> > > > > > > > >
> > > > > > > > > 2009/5/6 bh.hicks <bh.hicks@>:
> > > > > > > > > > Ang,
> > > > > > > > > > Ah yes, I had gathered as much but thank you for your candor.  I have
> > > > > > > > owned AB for over a year now but have just recently began using it for more
> > > > > > > > than a general market scanner due to some limitations I ran into with
> > > > > > > > Traders Studio and a general frustration with the development cycle on that
> > > > > > > > platform.  Other than a few issues I have run into (like this one), I have
> > > > > > > > been very pleased with AB and also see myself moving away from Traders
> > > > > > > > Studio for 'most' things.
> > > > > > > > > >
> > > > > > > > > > This would certainly be an amazing addition to AB.  Of course, the
> > > > > > > > feature would probably be lost on 90% of users so I understand the
> > > > > > > > reluctance to go down that path but I would certainly be willing to pay as
> > > > > > > > much if not more than I paid for AB itself for a robust plugin that was able
> > > > > > > > to do this well.  I suspect many others would as well.
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > > --- In amibroker@xxxxxxxxx ps.com <amibroker%40yahoog roups.com> ,
> > > > > > > > "ang_60" <ima_cons@> wrote:
> > > > > > > > > >>
> > > > > > > > > >> --- In amibroker@xxxxxxxxx ps.com <amibroker%40yahoog roups.com> ,
> > > > > > > > "bh.hicks" <bh.hicks@> wrote:
> > > > > > > > > >> >
> > > > > > > > > >> > I am basically looking for a way to have AmiBroker run multiple
> > > > > > > > systems concurrently in order to examine how trading multiple non-correlated
> > > > > > > > strategies affect drawdowns.
> > > > > > > > > >>
> > > > > > > > > >>
> > > > > > > > > >> As you are migrating from another software I happen to know a bit, the
> > > > > > > > short answer: as of today, Amibroker is not capable of "built in multiple
> > > > > > > > systems testing" as TraderStudio' s trading plan approach.
> > > > > > > > > >>
> > > > > > > > > >> This matter was raised long ago by myself and others: if you are a
> > > > > > > > registered user, you can check suggestion #406 in the feedback center, dated
> > > > > > > > 16 August 2006.
> > > > > > > > > >>
> > > > > > > > > >> In this list you can find very good programmers, claiming they are
> > > > > > > > able to get the multisystem/ multimarket approach to work programming it by
> > > > > > > > scratch, but - to my knowledge - I've never seen a public (that is.... free)
> > > > > > > > code able to do what I think is needed (and that's includes the two links
> > > > > > > > provided in this discussion).
> > > > > > > > > >>
> > > > > > > > > >> Just for clarity, this doesn't want to sound as a critic: I own
> > > > > > > > TraderStudio too, but by now the software I most use for testing is - by far
> > > > > > > > - Amibroker.
> > > > > > > > > >>
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > > ------------ --------- --------- ------
> > > > > > > > > >
> > > > > > > > > > **** IMPORTANT PLEASE READ ****
> > > > > > > > > > This group is for the discussion between users only.
> > > > > > > > > > This is *NOT* technical support channel.
> > > > > > > > > >
> > > > > > > > > > TO GET TECHNICAL SUPPORT send an e-mail directly to
> > > > > > > > > > SUPPORT {at} amibroker.com
> > > > > > > > > >
> > > > > > > > > > TO SUBMIT SUGGESTIONS please use FEEDBACK CENTER at
> > > > > > > > > > http://www.amibroke r.com/feedback/
> > > > > > > > > > (submissions sent via other channels won't be considered)
> > > > > > > > > >
> > > > > > > > > > For NEW RELEASE ANNOUNCEMENTS and other news always check DEVLOG:
> > > > > > > > > > http://www.amibroke r.com/devlog/
> > > > > > > > > >
> > > > > > > > > > Yahoo! Groups Links
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > >
> > > > > > > >
> > > > > > > > 
> > > > > > > >
> > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> > 
> > 
> > 
> > 
> > 
> >       Need a Holiday? Win a $10,000 Holiday of your choice. Enter now.http://us. lrd.yahoo. com/_ylc= X3oDMTJxN2x2ZmNp BF9zAzIwMjM2MTY2 MTMEdG1fZG1lY2gD VGV4dCBMaW5rBHRt X2xuawNVMTEwMzk3 NwR0bV9uZXQDWWFo b28hBHRtX3BvcwN0 YWdsaW5lBHRtX3Bw dHkDYXVueg- -/SIG=14600t3ni/ **http%3A/ /au.rd.yahoo. com/mail/ tagline/creative holidays/ *http%3A/ /au.docs. yahoo.com/ homepageset/ %3Fp1=other% 26p2=au%26p3= mailtagline
> >
> 
> 
>    
> 
> 
>       Need a Holiday? Win a $10,000 Holiday of your choice. Enter now.http://us.lrd.yahoo.com/_ylc=X3oDMTJxN2x2ZmNpBF9zAzIwMjM2MTY2MTMEdG1fZG1lY2gDVGV4dCBMaW5rBHRtX2xuawNVMTEwMzk3NwR0bV9uZXQDWWFob28hBHRtX3BvcwN0YWdsaW5lBHRtX3BwdHkDYXVueg--/SIG=14600t3ni/**http%3A//au.rd.yahoo.com/mail/tagline/creativeholidays/*http%3A//au.docs.yahoo.com/homepageset/%3Fp1=other%26p2=au%26p3=mailtagline
>




------------------------------------

**** IMPORTANT PLEASE READ ****
This group is for the discussion between users only.
This is *NOT* technical support channel.

TO GET TECHNICAL SUPPORT send an e-mail directly to 
SUPPORT {at} amibroker.com

TO SUBMIT SUGGESTIONS please use FEEDBACK CENTER at
http://www.amibroker.com/feedback/
(submissions sent via other channels won't be considered)

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

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/