|
PureBytes Links
Trading Reference Links
|
John,
Q: "when do we get into overoptimization? "
A. Almost everyday (sometimes a good thing, sometimes a bad
thing)
Nothing is a substitute for solid TA. My suggestions for anyone
using MetaStock is to:
1. Buy Edwards and McGee's book (the old testament)
2. Buy a Steve Nison book (the new testament)
3. Buy any number of books on system building
4. Read, study, apply the knowledge, win money, lose money,
start over
Harvey Pearce (yesterday's post) has covered the subject very
well and there's a lot written on optimization in both Future's
Mag. and TASC.
We all have access to a powerful piece of software that has the
ability to run many simulations in a short period of time.
There are many traps involved in system's trading and there is
no substitute for common sense TA. Having said this, I've
extracted a few pesos from many markets over the last 25 years
using the "automatic trading approach". Do I want to reoptimize
when they start to turn sour? Yes. Do I want to continually
search for a better mouse trap? Yes. Is this a good thing?
Maybe not. Are there any correct answers? I dunno.
I would not even be suggesting a simple, systematic approach
unless it worked over a broad range of securities and futures.
But, like fishing, why use the same bait for each type of fish
in the lake? I know, over the years, that I've satisfied my own
curiosity about using different parameters in different markets.
I use the same bar chart analysis and candlestick analysis on
every thing that trades. But, let me ask the "list": does
everyone use the default numbers for RSI that MetaStock has
assigned to the default? If so, why? Do you think Welles
Wilder uses the same period for equities and futures in his RSI
calculation? When does a stock or a market turn bearish? After
it breaks a 20 day moving average...a 50 day moving average...or
a 200 day moving average. Try sticking with a commodity, in a
long position, until it breaks it's 200 day average and I
suggest you might have a slight problem (bankruptcy).
So, all this astute analytical work that we do, ends up being
optimized in some ways.
In the last couple of weeks, I've posted two systems (with
optimizing formulas). Both these systems show the largest
returns in silver and stock indices. Are they tradable in their
optimized form? Due to the enormous drawdowns, in both
commodities, at times, I would say: no. Money management
principles are always more important than the approach you use
to trade. These formulas work over a broad array of securities
and futures (but, not all of them). They also work better when
they are optimized.
John, all you concerns are real problems and your points are
well taken. All aspects of TA have their blind alleys,
pitfalls, and false truths. System development and optimizing
is no different. It's always important to be careful out there.
When I "dangled" the formulas on the list, my entire goal was
to stimulate discussions on system development (plus, I can't
write these formulas worth a darn). I find that formula
discussions are a vast improvement over discussing the
inadequacies of Equis.
Thanks for your thoughtful and probing response,
Steve Karnish
CCT
Steve Karnish
CCT
----------
> From: John Manasco <manasco@xxxxxxxxxxxx>
> To: metastock@xxxxxxxxxxxxx
> Subject: Re: message interruptus
> Date: Sunday, August 30, 1998 1:10 PM
>
> Steve et al
>
> I have been watching this thread for awhile and would like to
ask a
> question on optimization. You state that lumber doesn't dance
to the
> same rhythm as DELL which is very true, but when do we get
into
> overoptimization? I have thought about trying to optimize my
parameters
> over each stock I trade but feel that is tantamount to
admitting my
> system is not very robust. I trade stocks and not commodities
so I
> understand the need for optimizing commodity systems but I
wonder how
> much stock systems need to be optimized. I notice Jim Greening
lets a
> stock sort of pick its own favorite indicators and pays
attention to
> them. I'd like to hear (read) more discussion about this.
>
> John Manasco
>
> Steve Karnish wrote:
> >
> > James,
> > I'll let the formula wizards handle your technical question,
but
> > wouldn't it make more sense to "optimize" the formula for
each
> > security or futures contract? Lumber doesn't dance to the
same
> > rhythm as DELL. Please keep in mind the basic laws of
> > optimizing (see Harvey Pearce's excellent discussion from
> > yesterday).
> > Steve Karnish
> > CCT
> >
|