|
PureBytes Links
Trading Reference Links
|
The chart referenced below represents my best neural network model for
timing the market to "Go Long / Exit Short" the S&P 500 Index. In addition
to the model's noteworthy timing characteristics, it offers some handy
probabilistic statistics (provided recent past probabilities are indicative
of future probabilities). For instance, note that in strong rising markets
the neural indicator pegs out around a value of 0.95 and remains there
(even when there is obvious noise in the data series). Based upon the 172
weeks charted (October 20, 1995 to January 29, 1999), the indicator was at
or above 0.94 for 73 of those weeks or 42% of the time. When the indicator
was above 0.94 for the period charted, the probability of a higher index
one week later calculated at 61% and 70% for two weeks later (The current
indicator value as of January 29, 1999 is 0.9552). Most importantly, when
the indicator "jumped" above 0.80 from a value smaller than 0.70
(indicating Market Entry), the probability of a rising S&P Index value 5
weeks later was 93% (13 of 14 such occurrences in the charted period).
This suggests the powerful nature of the model to anticipate long-term
up-trends quite handily. In addition, market entry typically lagged an
intermediate low by only one week (much better than channel analysis would
tend to indicate based upon visual inspection of some of the turning
points). My goal now is to develop a "Go Short / Exit Long" neural model
and a "Non-trending" neural model and to use all three in concert to
develop a viable trading strategy. - Brian
The chart could be found at http://pweb.netcom.com/~blee3/splong1.gif
|
