|
PureBytes Links
Trading Reference Links
|
I agree that the expert systems may be the exception to the general
rule. The DVI (Detection, Verification and Interpretation) technique
I use may be categorized as an expert system (a collection of un-
optimized systems used to arrive at a consensus signal) with adaptive
parameters. An optimized system behaves exactly like an un-optimized
system out-of sample if it is not re-optimized. Then why does an
optimized system works sometimes and fails often? How robust is it?
It is the difference between training a mouse to navigate a
particular maze and training a mouse to navigate any maze... The
mouse that has been trained to naviagate a particular maze may do it
faster than the mouse that has been trained to navigate any maze, but
when the former mouse enounters an entirely different maze (a market
with different personality) it fails, hence it requires re-training
(re-optimization). The answer may be in the number of parameters
used by the system, i.e., how robust it is. The more optimized
parameters it uses the more re-optimization it needs to successfully
cope with this change in market condition, as opposed to adaptive
parameters (many parameters but the values fixed for each system so
that all of the underlying instruments use the same periods for each
system) which require no re-optimization and is able to detect
accurate signals regardless of the market size, condition or nature.
MAs are perhaps the most widely used form of analysis because it has
been proven by elaborate forecasting modeling that most investments
follow a MA process. The key to successful market timing is
identifying when the prices are significantly lower or higher than
the average price. This discrepancy between the current and the
average price identifies a possible trading opportunity. Generally,
the price will return to the average price at some point in the
future - hence proving the MA process itself...
rgds, Pal
--- In amibroker@xxxxxxxxxxxxxxx, "Fred" <fctonetti@xxxx> wrote:
> In general I would agree with the concept that the liklihood of
good
> out of sample or real time system performance is indirectly
> proportional to the number of parameters, but there ARE exceptions
to
> this general rule at least as applied to what some would term to be
> expert systems. A case in point is a system I use every day and
has
> had for years as good out of sample, real time performance as it
had
> in sample. It has 27 parameters. I haven't reoptimized in at
least
> 18 months because there was no reason to.
>
> --- In amibroker@xxxxxxxxxxxxxxx, "palsanand" <palsanand@xxxx>
wrote:
> > My primary trading system is very simple in the sense that it has
> > only one parameter (the lookback MA period), and it generally
leads
> > to a robust system. It is objective (non-optimized and
parameters
> > are never subject to change) and unique in the sense it is both
> trend-
> > following (continuation of a previous trend or breakout) and
> counter-
> > trend following (start of a new trend) at the same-time. The
trend
> > and counter-trend following features of the MA based systems are
> > highly desirable because it resists becoming outdated as markets
> > change character (personality). Independent of the fundamentals
> > driving the market, these features are designed to capture
extended
> > runs in both bullish and bearish directions. The Verification
and
> > Interpretation of the signals Detected by this system was a great
> > challenge, but I was able to use good filters (OB/OS) and use
> > Volatility (non-trend component of a signal) to optimize my entry
> > points. A robust system can be categorized as one that stands a
> good
> > chance (probability) of working in the future as it has worked in
> the
> > past, i.e, it is tough (able to handle all markets and conditions
> > regardless of size and nature) and long-lasting (durable).
> >
> > As discussed in a recent STOCKS & COMMODITIES article by Jeffrey
> Owen
> > Katz and Donna McCormick, a rule of thumb in evaluating trading
> > systems is that the more parameters a system has, the less robust
> the
> > system is. Taken to the extreme, many fitting parameters can be
> used
> > to curve-fit past data to eliminate all whipsaws while
maintaining
> > good performance. The potential of such a system working in the
> real
> > world is nil...
> >
> > rgds, Pal
> > --- In amibroker@xxxxxxxxxxxxxxx, "Dave Merrill" <dmerrill@xxxx>
> > wrote:
> > > agreed. if the fact that a trading system did well in the past
> has
> > no
> > > bearing whatsoever on whether it does well in the future, how
can
> > we know
> > > anything at all about the future performance of a proposed
> trading
> > system?
> > >
> > > dave
> > >
> > > The gambler”Ēs fallacy is a fallacy because the gambler
ignores
> > the
> > > independence of the outcomes and looks for patterns that do not
> > exist. If
> > > we have designed trading systems based on recognition of
patterns
> > that
> > > precede profitable trading opportunities, and if those patterns
> are
> > > persistent, then we no longer have random, independent
outcomes.
> > Our
> > > trading systems do have serial dependencies and upward sloping
> > equity
> > > curves. So analysis of the equity curve provides an indication
> of
> > the
> > > health of the trading system.
> > >
> > >
> > >
> > > Howard
------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/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/
|