|
PureBytes Links
Trading Reference Links
|
Paul Altman wrote:
>
> I'm experimenting with trailing volatility stops and I'm immediately
> confused by the enormous resulting size of the stops, based on Tharp's
> suggestion to experiment with a 3 * ATR trailing stop, as a starting point.
>
> 1) Let's look at QCOM: a recent 10 day weighted moving average of the ATR
> is, according to my calculations, ~8% at the close of Fri 4/7/00. OK, it's
> a volatile stock, but a trailing stop of 24%?!! Doesn't that seem like an
> awfully large stop? [snip]
I'm not a systems xpert by any means but I've looked a bit at atr based
stops on commodity charts. Here's my impressions:
I think it's a trend following stop, if set to 3 atr. In other words it
attempts to stay in during minor retracements. If you're a swing trader
you don't use this big a stop (or you use it on a smaller timeframe).
It assumes that in the course of a trend, short reversals occur, perhaps
several days in a row. Each of these days might take the stock down 1
atr. So if you want to ride the whole trend w/o getting bumped out
during the correction, you have to have a stop that's bigger than atr.
If you don't mind getting bumped out more often you could try 1 atr,
say. But not only do you give up 1 atr when you're stopped out, but you
give up subsequent movement that occurs in the right direction
until/unless you get back in. If it happens several times in a trend
that the 3atr stop just kept you in, you're better off with the 3atr.
Still having said that, I think a 1 atr stop will keep you in most
trends for a good while.
So, you could compromise. Nobody says you have to have the initial stop
at 3atr. You could, as an arbitrary example, use 1 atr for the initial
stop and 3 atr after you have profits. Now unless you entry signal is
particularly prescient, using 1 atr might give you more losses (though
smaller).
> 3) If the whole point of an x*ATR stop is to make sure that you're not
> stopped out due to the normal fluctuations of noise, isn't x=3
> extraordinarily generous? If you very loosely assume that TR's tend to
> straddle the closing price, then really only _half_ of the TR should
> represent the normal downward fluctuations. So x=3 means that you're
> setting a stop very approximately _6_ times larger than the average normal
> downside fluctuation in a given day.
Stock might move up 10 days, down 3 in a row, and then up 10. question
is are you trying to just trade the first upmove or stay in for both?
> 2) Additionally, assuming for a second that QCOM remains as volatile for a
> while into the future, you'd have to have a profit of 32% before you'd even
> be assure of breaking even (after a possible 24% drop back to your original
> buy price). After a 20% or 30% profit, shouldn't you be allowed to tighten
> your stop so that you actually are assured of keeping some of it?
>
One of the most frustrating parts of longer term trendfollowing is that
phase where you have a decent profit but your stop doesnt' protect any
yet. However, anytime you move your stop closer, you risk getting out
of the trade only to see it go on to $$$. If your system is "supposed"
to be in the mega trade every once in a while and you defeat that by a
close stop you might be screwing up your system (unless you've tested
the close stop as well).
The market does't know where you got in. So in reality, if a 3 atr stop
is good for trailing your profits it may well be the best for staying in
a trade from the start. It seems like if your entry is a "simple" entry
like moving average, this would be true. Since you never know if the mkt
might decide to do a short retrace after you get in.
However, maybe your entry conditions somehow make that untrue. Suppose
your entry rules get in only after a retracement has just occurred.
Maybe you've found that if a 2nd retracement occurs right after you get
in, that's a bad sign. If that's true, why ride it out? Use a closer
stop at the beginning.
Also despite that fact that the $ is the same stuff whether it is losses
or open profits, I feel there is some change in the statistical picture
when you move your stop to breakeven. (After all you could do an
infinite # of breakeven trades and not damage your acct.) From the
standpoint of market motion, your likelihood of an unneeded stopout,
etc. - a "breakeven" stop is unjustified. But maybe from a money
management standpoint it's a good idea.
You hear pretty often from trendfollowing system traders/writers that
uncomfortably-far stops are the most profitable. It seems a number of
them use a closer initial stop (or at least a $ maximum), but their
profit taking stop is typically awfully far away at times. Of the few
systems i've bought or subscribed to, one used a 50% profit give back as
a stop, another as I recall used 2-atr but moved above breakeven only if
a significant profit (a big multiple of atr) was accrued, and a third
used what appeared to be a slow moving average. In contrast one other
system subscription I took had fairly distant (fixed point value)
initial stops, but *very* aggressive breakeven and profit preserving
stops. The result was a lot of premature stopouts and they had to get
back in the next day (with the larger entry stoploss!).
You're "allowed" to do anything with your stops. Just test it, if you
can. Attached is a very simplistic atr envelope indicator. If you plot
it on the same scale as your prices you can play with different atr
multiples. In actual trading you probably wouldn't let your stop go
"backwards". Some better versions of this indicator are available that
prevent the line from going backwards. (I've got it defaulted rather
"tight" but it's only 1 of 2 components to my system stop (which still
needs some testing).
Conrad Bowers
Attachment Converted: "f:\eudora\attach\Atrstop.ela"
|