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

Objective functions (was RE: [amibroker] Re: Optimization -- again)



PureBytes Links

Trading Reference Links

"Where your test data starts determines the final test results just as
much as your system does."

It does ? Maybe for what you trade, how you trade it, the size of the 
sample and/or number of trades and what you used as a yardstick to 
measure how good it is, but I can't say as I ever observed this 
except of course when the sample was too short, like it typically is 
with walk forward optimization.

--- In amibroker@xxxxxxxxxxxxxxx, "palsanand" <palsanand@xxxx> wrote:
> Hi,
> 
> Here is an interesting observation on system testing:
> 
> Say you run a system test over 10,000 bars of data, then print out 
a 
> chart of the system's equity line. Then repeat the test, but start 
> 100 bars later. Let's say two trades were included in those 100 
bars, 
> so they've been dropped. Now print the second equity line and 
compare 
> it to the first. You'd get exactly the same equity line, but 100 
bars 
> shorter. Right? Wrong! 
> 
> When I do this I get a radically different equity line on the 
second 
> test, i.e., they are not near-mirror images of each other. My hunch 
> is that a form of the chaotician's "butterfly-effect" has arisen: 
> changing any given trade's market position (long, short, flat) will 
> effect in a chain reaction all the subsequent trades in complex and 
> unexpected ways. Here dropping the first two trades could very well 
> change the system's market position when the third trade is 
> calculated, and so on. 
> 
> I believe this observation has profound and unfortunate 
implications 
> for the robustness of system testing. It's a second and more subtle 
> problem that lies behind the mere curve-fitting/optimization 
problem. 
> 
> If dropping a couple of early trades will always effect later 
trades, 
> then there's no truly "neutral" starting point with any test data. 
> Where your test data starts determines the final test results just 
as 
> much as your system does. 
> 
> The success or failure of many different mechanical systems is 
> predicated to a surprising and varying degree on the sequence of 
> events just prior to the first actual trade generated by the 
system. 
> 
> The trade setup and timing of the first trade can have a profound 
> effect on the subsequent trading results. The circumstances and 
> timing of entry into the first trade can sometimes make a huge 
> difference in the overall trading performance. 
> 
> Regards,
> 
> Pal
> 
> --- In amibroker@xxxxxxxxxxxxxxx, "Fred" <fctonetti@xxxx> wrote:
> > That IS what I was trying to say.  I suspect because equity feed 
> back 
> > is like looking in a rear view mirror, great for letting us know  
> > where we were and how we could have adjusted the past to make it 
> > better, but that's about it.  
> > 
> > --- In amibroker@xxxxxxxxxxxxxxx, "Dave Merrill" <dmerrill@xxxx> 
> > wrote:
> > > don't think I get what you mean here fred.
> > > 
> > > you can't be saying that metrics on the equity curve of a 
trading 
> > strategy
> > > or its parameters aren't useful, right? that's the only thing 
we 
> > have to
> > > judge the effectiveness of our methods and settings.
> > > 
> > > so you must be saying that equity feedback isn't a useful 
concept,
> > > regardless of how you measure "good" equity. do I have that 
right?
> > > 
> > > if so, as I've said, my experience agrees -- none of the 
> indicators 
> > I've
> > > tried are wonderfully profitable when auto-optimized this way. 
I 
> > just cannot
> > > for the life of me understand why that's the case, if backtests 
> > tell us
> > > anything useful about future performance.
> > > 
> > > if I've misunderstood completely, my apoligies (:-)
> > > 
> > > dave
> > >   Like a lot of other things that sound like they SHOULD work, 
I 
> > have
> > >   never found metrics related to equity curve feedback to be of 
> much
> > >   value in the determination of system parameter values.
> > > 
> > >   --- In amibroker@xxxxxxxxxxxxxxx, "Dave Merrill" 
<dmerrill@xxxx>
> > >   wrote:
> > >   > interesting as usual howard (:-). one piece I wanted to 
drill 
> > into
> > >   a bit.
> > >   >
> > >   > I wonder what the effect of using performance measures that
> > >   concentrate on
> > >   > certain things at the expense of others actually is.
> > >   >
> > >   > for example, my auto-optimization stuff currently uses 
simple
> > >   profit per bar
> > >   > to choose parameter values. my gut-level assumption was 
that 
> > since
> > >   it was
> > >   > ignoring drawdown (among other things), the resulting 
systems 
> > might
> > >   have
> > >   > higher drawdown than I was comfortable with, but that 
profit 
> per
> > >   bar should
> > >   > be as good as the trading method could produce.
> > >   >
> > >   > maybe that's not the case. maybe by choosing a more balanced
> > >   success metric,
> > >   > not only would the other factors not considered by my 
> simplistic
> > >   first pass
> > >   > metric be improved, but profitability might be improved as 
> well.
> > >   >
> > >   > is this something you've investigated or thought about? 
> anyone 
> > else?
> > >   >
> > >   > dave
> > >   >   Note ? it is perfectly valid to have different objective
> > >   functions for
> > >   > different purposes.  For example, I might be modeling the 
> > behavior
> > >   of a
> > >   > sector, say oil services, with the intent of trading 
> individual
> > >   stocks based
> > >   > on what I learn.  In this case, I want to identify periods 
of
> > >   rising prices
> > >   > with careful attention to turning points, but without much 
> > interest
> > >   in
> > >   > overall profit.  On the other hand, I might be modeling 
> > individual
> > >   high beta
> > >   > tech stocks, in which case my model includes several stop 
loss
> > >   techniques
> > >   > and I care most about avoiding drawdowns.
> > >   >
> > >   >
> > >   >
> > >   >   Thanks,
> > >   >
> > >   >   Howard
> > > 
> > > 
> > >         Yahoo! Groups Sponsor
> > > 
> > > 
> > > 
> > >   Send BUG REPORTS to bugs@xxxx
> > >   Send SUGGESTIONS to suggest@xxxx
> > >   -----------------------------------------
> > >   Post AmiQuote-related messages ONLY to: 
amiquote@xxxxxxxxxxxxxxx
> > >   (Web page: http://groups.yahoo.com/group/amiquote/messages/)
> > >   --------------------------------------------
> > >   Check group FAQ at:
> > > http://groups.yahoo.com/group/amibroker/files/groupfaq.html
> > > 
> > >   Your use of Yahoo! Groups is subject to the Yahoo! Terms of 
> > Service.


------------------------ Yahoo! Groups Sponsor ---------------------~-->
Rent DVDs Online - Over 14,500 titles.
No Late Fees & Free Shipping.
Try Netflix for FREE!
http://us.click.yahoo.com/Tq9otC/XP.FAA/3jkFAA/GHeqlB/TM
---------------------------------------------------------------------~->

Send BUG REPORTS to bugs@xxxxxxxxxxxxx
Send SUGGESTIONS to suggest@xxxxxxxxxxxxx
-----------------------------------------
Post AmiQuote-related messages ONLY to: amiquote@xxxxxxxxxxxxxxx 
(Web page: http://groups.yahoo.com/group/amiquote/messages/)
--------------------------------------------
Check group FAQ at: http://groups.yahoo.com/group/amibroker/files/groupfaq.html 

Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/