|
How and why the program
works:
Over
the long term, the market is influenced by two major factors--the expectation
of future earnings and the expectation of future interest rates.
This is a flexible relationship with large normal variations. These two
elements have much less of an effect when you look at the market one day
at a time. We prefer to work in the short term where investor
psychology is the primary influence.
Palisades
Research uses a statistical approach to market trading. The technique is
unique and proprietary. It relies on two basic elements, "money
flow" and "investor emotion".
Like Deep Blue, the IBM
program that beat World Chess Champion and Grandmaster, Gary Kasparov in
1997, our program relies on a brute force approach. The two programs
work in basically the same way. They apply weights to different choices
and solve for the best one. We use genetic algorithms to determine
the weights.
The analysis of money flow
looks at the movement of cash into or out of the market. This can come
from pension and wage activity or from shifts between market segments.
This is a kind of "causative" or "push" force.
The emotional element
examines how investors have reacted to short term market changes and to
the changes in factors that influence the market (like the price of
oil). This is not a "push" force, it is more a
"reaction to" stimulus. Emotional response tends
to remain consistent over time unless affected by drastic conditions
like those found in 1929 and 1987. Even then, change is slow to
occur and once things become stable again the old patterns reappear.
A major effect on the
market is whether or not the people that are investing the next day feel
optimistic based upon the way the market closed and its recent
behavior.
The stock market moves up
less often than people think. Over the last 50 years, it has gone up
only 53% of the trading days. This means that 94%
of the time the market goes nowhere. Focusing on each day,
individually as an opportunity, is very rewarding.
Initially the total effort of this
program went into forecasting the direction of the S&P500 index for
the single following day. We now target the NDX. This program makes no attempt to estimate the
amount of change beyond one day. It does not try to determine the
probable direction for the next week, or any other time period other
than for the next day.
Our longer term forecasts
use a different program with a different focus.
Our main computer program indicates
that the next day’s direction can be forecast with a 60% – 70%
reliability compared to 53% with buy and hold, but even this is very
difficult to maintain in real life. This seemingly small improvement is
actually very substantial. Nothing can approach 100%, because
there is nothing we can do about what happens after the market is closed
and during the next market day. Certainly, things will happen to
influence the market direction and all of them are beyond our control.
The best we can do is to make sure we have a strong enough signal to
overcome minor adverse influences. When the program gives an
"Money Market" or "OUT" signal, it doesn’t mean
that the next day's movement will be insignificant; it only means that
we don’t have a clear indication of direction. So we move into the
money market.
Over 50
years of data were used in the development of the
fundamental algorithm that drives this program. Yet, it was
written in a way that allows it to be responsive to changes that occur
over shorter time frames so it can adapt to its environment.
This
program will continue to evolve, since ceaseless
effort is underway to improve the program and make it even better.
Recently,
in response to the extreme down market early in 2001 we added a technique
borrowed from NASA.
Patience
is one of the keys to this method. We stay in the money market currently
about 20% -25% of the time, patiently waiting for those days with the highest
probability of gain. When we
are out of the stock market we do not expose ourselves to market
risk.
|
