|
PureBytes Links
Trading Reference Links
|
Steven,
I wonder if all this statistical work, despite how cool and technical it
sounds, is really worth the effort and would yield significantly different
results than a much simpler approach. For example, the Cynthia Kase model
uses SD to calculate a two bar stop while others use all kinds of
qualifying rules, while some prefer tried and true methods like MA and
volatility. Some even like fitting distributions and on and on. Arrrgh.
Calgon, take me away!!
Here's a method that takes into account most of the concepts presented so
far-- take a 30 day average of the range and multiple by 2. There's your
average two bar stop, your volatility based indicator and any smoothing you
wish to factor in. Done, simple and it works. Now you can spend your time
doing other things which is most likely why you trade in the first place.
I think it's more valid than pulling at least 30 samples over the past 5
years because price action like this is about the last 90 days not what
happened 5 years ago.
Or to simplify it more, place your stop above the high 2 days ago.
It seems that the Kase method of selling software is to write a book
promoting the idea then sell the math for a ridiculous sum in software.
Why would anyone buy her books if she's not going to provide a complete
solution? The formulas are part of the overall process and she's more or
less misleading the customer into thinking there are going to get a
complete answer when they buy her book. I would learn from the math she
uses too not just the ideas and that's what I expect to get when I buy her
book. Needless to say, her book has the most dust on my shelf.
-----Original Message-----
From: Conrad Bowers [SMTP:cpbow@xxxxxxxxxxxxx]
Sent: Sunday, July 26, 1998 9:49 AM
To: Bill Vedder
Cc: Steven Buss; omega-list@xxxxxxxxxx
Subject: Re: Stop Placement Concept Question
Bill Vedder wrote:
>
> Steven Buss wrote:
> >
> > There appear to be at least 3 concepts for thinking about stop place
ment:
> >
> > 1 Place a stop at a fixed dollar amount based on how much money I
can
> > afford to lose <G>
> > 2 Place a stop at some market price level that is determined by a
> > technical indicator (e.g., a level at which support/resistence is
broken
> > (i.e., a breakout), moving average crossover, momentum indicator turn,
etc.)
> > 3 Place a stop at a fixed dollar amount based on trading system
> > backtesting and then doing MAE analysis (see Sweeny, "Campaign Trading"
for
> > a discussion of MAE, Ruggerio for TS code that exports data for Excel
graph
> > of MAE, Omega's new Portfolio Maximizer does MAE) (Also, Pierre
sometime in
> > the last several months posted some TS code related to MAE)
> >
> > Are there other CONCEPTS around which stop placement may be thought
about
> > that I've missed?
> >
> > Obviously, listing #1 above is a joke for the oft stated reason that
the
> > market doesn't care about my personal capital requirements.
> >
>
> > I have a question related to concepts 2 and 3 above.
> >
> > - One reads that trading systems should first be backtested without
> > stops. I assume that what is meant by this recommendation is that
stops are
> > coded into the system based on technical indicator concepts (i.e., by
the
> > kind of concepts in #2 above). Am I right in this assumption?
> >
> > Steven Buss
> > Walnut Creek, CA
> > sbuss@xxxxxxxxxxx
> > "There's nothing more practical than good theory."
>
> I can think of two more; stops based on time and stops based on
statistical
> methods.
>
> I don't know how time periods are selected for time based stops. Others
on the list
> are much more knowledgable about this method and may care to speak to
this.
>
> I've not tried the statistical method (and it's been a long time since I
read the
> paper) but it was based on measuring the size of all reactions against
prominent
> trends and fitting a distribution to the data. The authors used an
exponential
> distribution. You then selected a probability level and calculated the
size of the
> stop from the distribution. If interested, I can dig around and find the
reference.
>
> Regards,
> Bill Vedder
Cynthia Kase in "Trading with the odds" describes (partially) a
statistically derived stop. Her stop is placed at a distance to
withstand "2-bar reversals". She measures a mean value and the standard
deviation and places the stop at (mean + n*sigma) away fro the price
action. However the exact formulas are not given. I'm not sure if she
is measuring historical reversals and averaging them or just saying a
2-bar reversal is 2 times the current avg-true-range. If the latter,
you would just get the average and sigma of ATR and then you could
calculate the stop. (This was an interesting book, but i had the same
"complaint" in various places: while the general idea was explained,
actually formulas or procedures were rare; i guess to have that you have
to buy her software. Maybe that's only fair, but....) In any event,
she recommends using 3sigma for staying with a trend and 1 sigma if
warning signs of the end of the trend exist.
Al Gietzen in "Real Time Futures Trading" does use a stop that is:
avgRng/2 + n*sigma. He uses an exponential avg to get the average range
and suggests n should be 1.8-2.5. He smooths the stop and then places
the stop this distance away from the daily price midpoint. It should be
noted that this was in context of a system trading cycles and he
anticipated many trades would be reversed by a signal, not stopped out,
so these stops (or at least the suggested n) may have to be modified if
the stops are the primary exit.
I have to wonder if it's worth the trouble to do statistical analysis to
the point of finding sigmas. Sure, you are finding the distribution of
the last 15 or 30 days or whatever but the next ten days may be a
different "population". Certainly basing stops on current volatility
makes sense. But is calculating the stops to be a distance of avgrng +
n*sigma away, any better than just using n*avgrng?
Colin Alexander in Five Star Futures Trades, suggests chart-based stops
but with certain modifications when the stop would be too far away. For
example, never more than a limit move away, or in markets without
limits, never more than n*margin. He suggests looking at 60 min. charts
to see if there is a chart based stop on that timeframe near where you r
desired distance is. Of course, now we are getting into a considerably
more discretionary mode or at least one that is harder to test with TS.
Conrad Bowrers
|