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

Re: [RT] Neural Models Predicting S&P Bottom ?? - Retry



PureBytes Links

Trading Reference Links

Hi Don,

The two previously posted neural networks are not like the kind one expects
to see, that is, they are not "trading models" that suggest when to go long
and when to short.  These two models are what I would call my "specialty
models" that focus in one area only, "When do I have a decent probability of
going long?"  You can see they both answer that question quite well.  Their
singals are either "off" (below the red zero axis line - not indicating a go
long signal) or "on" (passing above the red zero line - indicating a go long
signal such as tonight's posts).  I personally tend to like the top model of
the two posted (about 14 out of 15 correct picks if you allow for short-term
up movements as well as major bottoming situations, hopefully 15 of 16
correct after tomorrow).  If you are suggesting drawing trendlines on the
indicators (signals), I would not suggest that.  Again, the interpretation
of the signals is basically binary (is it time to go long or not ("on" or
"off')?).  An example of specialty "Shorting Models" would be like the two
posted below, also 18 months of out-of-sample signals.  The first shorting
model attempts to detect weakness in the market.  I call it my "trap door"
model in that the yellow neural signal clings to +1.0 until it detects
market weakness.  The yellow neural signal then starts to fall from its +1.0
perch, thus opening the door for the market to fall.  This particular model
is often a leading to coincident indicator.  On certain occasions the "door"
will open up as the market begins to move in a sideways pattern prior to
falling, again indicating market weakness.   I know the detail is not in the
chart, but zoomed views show that it is quite reliable in its task of
predicting.  The second shorting model attempts to predict minor/major
market topping periods.  This model's yellow neural signal hugs -0.41 until
it thinks the market has reached a topping area.  It then spurts up from
its -0.41 home base.  Visual analysis shows that the height of the signal
jump is not correlated well with the pending market fall.  In other words,
whenever the neural signal jumps up (filtering out / ignoring the very tiny
jumps), the idication is that a minor/major topping event is near.  Large or
small, who knows?

Regards,
Brian



-------------------------- eGroups Sponsor -------------------------~-~>
eGroups eLerts
It's Easy. It's Fun. Best of All, it's Free!
http://click.egroups.com/1/9698/0/_/152424/_/971317438/
---------------------------------------------------------------------_->

To unsubscribe from this group, send an email to:
realtraders-unsubscribe@xxxxxxxxxxx



Attachment Converted: "c:\program files\qualcomm\eudora\attach\1113.gif"

Attachment Converted: "c:\program files\qualcomm\eudora\attach\2223.gif"