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

[EquisMetaStock Group] Re: Early trend stages



PureBytes Links

Trading Reference Links



Thank you for your kind words - just a few additional thoughts.  I am
a bit new to the group, but was absolutely amazed at the often very
specific answers you received to the original, in my view, fundamental
question.  I cannot tell you how many 'holy grail' systems I've tested
in the past couple of years, from any conceivable moving average to
neural networks and fancy hilbert stuff.  Not only test the stuff, but
optimize it to the data.  We do a fairly rigorous test of models and
make sure we do not overfit them, and the ones that really work are
really not easy to find.  Even when I googled the Tillson indicator
I very soon came across pages discussing how it did not work and how
you could refine it to make it work.

My point is this, I am a little bit wary of such specific, one-size-
fits-all answers to such a general question.  I will test the answers
that was posted by building models for some of them and subjecting it
to our huge database of shares, along with a huge array of other
models, and then will know from the numbers if those models do work
or not, how general they are, and so forth.

Back to the original question, and spending some fundamental thought
on it.  To catch a V turn either way must be very profitable and,
I've seen, can cause an otherwise good system to break down.  It
requires a trend system that can follow the one leg of the V and,
very very soon after (or before!) turning give a signal to put you
into the other leg of the V.  Such a system can be built, but over
time you will be in and out so many times you'll go bust in a wink.
To avoid this, you introduce smoothness, thus lag, and you are not
able to pinpoint the point of the V so easily.  You will probably
be in on the first leg of the V and - I would recommend - be out
via a stop loss level or trailing stop, long before the trading
system turns around.  This is from experience that I make this
statement.

If you know anything about signal processing, I had good results
with ARMA models and bad ones with AR models.  If I add the MA
component, the model improves quite a bit.  Now, this MA bit is an
error-correcting bit that models how to react when you were wrong.
It also implies a choppy market where it typically goes down after
going up the previous day (in the short term, but you can extend
the argument to a V if you wish).

We typically use some long term trend model (that will only turn
long after the market) to establish the primary trend and then use
all manner of shorter term devices to trade around this trend.

Anyhow, good luck with the trading and I'll try to post some results
of the models mentioned once I have tested them.

Regards
MG Ferreira
TsaTsa EOD Programmer and trading model builder
http://tsatsaeod.ferra4models.com
http://www.ferra4models.com




--- In equismetastock@xxxxxxxxxxxxxxx, Philip Schmitz <pschmi02@xxxx>
wrote:
> 
> What a showing! MG, Kevin, Whit, Steve and Sebastian, thank you very
much 
> for posting your thoughts. I've printed them out so that I can take
a very 
> close look. The opinions of others are invaluable when it comes to
putting 
> issues like this into perspective.
> 
> Part of the problem may be that I've been looking at the weekly
chart for 
> directional signals. That could be heightening my sense of lag and 
> unfortunately also my impatience.
> 
> Note to Steve: As far as the following is concerned . . . ahem . . . "A 
> real thing of beauty:  very smooth, but sensitive to significant 
> directional changes" .  .  . I think we better talk about that one
over the 
> phone.  (huge grin!)
> 
> Best regards and thanks again,
> 
> Philip








------------------------ Yahoo! Groups Sponsor --------------------~--> 
In low income neighborhoods, 84% do not own computers.
At Network for Good, help bridge the Digital Divide!
http://us.click.yahoo.com/EpW3eD/3MnJAA/cosFAA/BefplB/TM
--------------------------------------------------------------------~-> 

 
Yahoo! Groups Links

<*> To visit your group on the web, go to:
    http://groups.yahoo.com/group/equismetastock/

<*> To unsubscribe from this group, send an email to:
    equismetastock-unsubscribe@xxxxxxxxxxxxxxx

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