Introduction
Following is a proposal to adopt a new strategy for the hedge fund
and a supplemental strategy for the discretionary fund. The
objective of this strategy is to construct a consistent philosophy
and trading plan based on an observational model about how markets
work in the context of the Hanseatic model's characteristics.
How Markets Work
One can argue endlessly about the sources of price change in
markets. Fascinating, and fortunately, unnecessary. The purpose here
is to suggest a model of price behavior based on a lot of study and
some statistical characteristics of our model. We're trying to
combine a few fundamental principles into a consistent process. The
major objective of this process is simply to be in harmony with the
market's movements, patterns and environments. First, in general
terms, this is the basic observational model of market environments.
One can argue endlessly about the
sources of price change in markets. Fascinating, and fortunately,
unnecessary. The purpose here is to suggest a model of price
behavior based on a lot of study and some statistical
characteristics of our model. We're trying to combine a few
fundamental principles into a consistent process. The major
objective of this process is simply to be in harmony with the
market's movements, patterns and environments. First, in general
terms, this is the basic observational model of market environments.
One of the most compelling books I've read about price behavior is
The Path of Least Resistance. The author, Robert Fritz, is a
musician and there is no mention of markets in this book. He talks
about the underlying structures (of anything) determining the path
of least resistance. Some structures lead to oscillation and some
structures lead to resolution. Similarly, markets spend a lot of
time oscillating in trading range environments, where the dominant
price characteristic is to move away from the mean (smooth price)
and generate a buy or sell signal which represents a return to one
end or the other in the oscillating structure. But market forces,
whatever their composition, do not permit price to resolve outside
the oscillating structure, hence a return to the smoothed price, and
then a deviation in the opposite direction. At some point as the
market moves through time, some combination of fundamental and/or
technical factors create a structure that resolves (into a trend).
Then the path of least resistance for the market is to move away
from the smooth price in a sustained way. Market psychology and
momentum make if difficult for signals counter to the trend to
extend beyond the smoothed price of the larger time dimension
(weekly). Instead, volatility declines and the market enjoys a
benign period of "sponsorship," that is interrupted by minor
excesses (i.e. figure 8 type patterns) and corrections of these
excesses.
At some point in time, excess develops in the larger time dimension
that is relieved by either a correction in time, i.e. trading range,
or resolution and sustained trend in the opposite direction.
This is a general philosophy of price behavior in markets as I have
observed them over the years. It applies, I believe, to all time
dimensions and with rare exceptions, to all markets. Is this
philosophy perfect and do prices always behave in the framework of
these two dominant structures? No and no. As we all know, the
reality of markets is that they can behave in ways that are quite
messy and very frustrating to rigid rules. But I nevertheless think
that it is important to have a market philosophy, and that a trading
plan absent that philosophy is not likely to be successful.
The Hanseatic Model
With the relatively slow smoothing in our exponential moving
average, the buy and sell signals are essentially a trend-following
model. However we have never used the model in this way. Rather, to
the extent we use the signals at all, it is very selective and then
sometimes in the opposite direction, i.e. buying on a sell signal,
or what we refer to as a countersignal. Nevertheless, because I want
to incorporate these signals in the trading process I am proposing,
it is important to know the characteristics of the signals.
Using the Ten Year Notes as an example, this is a summary
description of all buy and sell signals for this market: Since
October, 1993 there has been a total of 123 buy and sell signals.
29.3% have been profitable, 70.7% unprofitable. The profit factor is
0.70 and the total gain for all trades is -26.98%. Profit factor
(PF) is the sum of the gains divided by the sum of the losses (PF =
1.0 is break even). If we look at the best winning trades, say the
top 5% which is the best six trades in this case, the gain from
these trades are 52% of the total gain. Of 123 total trades, only 36
were profitable, and 5 of the 36 winning trades accounted for more
than half the gain. The biggest single trade contributed 20% of the
total gain.
The most basic conclusion we can draw is that most of the signals
lose and that relatively few trades account for most of the gain.
Said another way, the majority of trades belong to the oscillating
trading range environment and a minority of trades resolve into
trends. The table on the next page shows some relevant statistics
about the signals for a wide variety of markets.
Some conclusions we can reach from
the signal analysis are as follows:
-
Overall, about one-third of all
signals are profitable, the remaining two-thirds of signals are
unprofitable. U.S. financial markets have been profitable less
than 30% over the last ten years. A few commodities have had good
signals over 35% of the time.
-
The profit factor over a wide range
of futures contracts is essentially 1.0 or breakeven, and negative
for U.S. financial markets.
-
On Average, about 45% of the total
gain for a market is accounted for by only 5% of the total trades.
In the S&P 500, 10 out of a total 202 daily buy and sell signals
made 51 % of the total gain, or 100% of the total gain of 197%.
About 74% of the total gain, or 146% of the total gain of 197% is
derived from only 10% of all the trades.
-
Not discernable from the table are
two other perhaps obvious but nonetheless salient facts. First,
losing trades are relatively short and winning trades last much
longer. Losing trades last about 6-8 days on average, while the
best trades last from 25-100 days. Second, 30-40% of the maximum
gain during a buy or sell signal is given up as the market
reverses course and transitions to the opposite signal.
Trading
Plan
Our goal is to develop a trading plan that is consistent with both
our observational philosophy and the nature of our basic model. The
objectives of the trading plan are as follows:
-
Participate in every major trend in
every market or stock that we choose to trade.
-
Function in the oscillating or
trading range markets through the use of countersignals.
-
Try to observe and trade the
excesses that develop in every trend using the figure 8 and other
patterns of that character.
-
Conserve capital.
Now, returning to the observational
model shown earlier, the diagram below is an expanded version that
incorporates the Hanseatic model.
Note: B* and S* refer to new signals.
This is how I view market price flow incorporating the
characteristics of the model and some of the patterns we have
developed. Another way to view the process is as a sequence of
market patterns or fractals.
We have quite a bit of experience in trying to develop models that
are similar in character to the process above. That is, selling into
strength in the trading range environment, taking selected signals
that have better probabilities straight, and trying to recognize the
transition from a trading range environment to a trended
environment. My conclusion is that while it is relatively easy to
model a base version of the process, try to capture all of it makes
the model very complex.
I am proposing that we us the discipline inherent in the process to
trade a wide variety of exchange traded funds, individual stocks and
futures. That we anchor ourselves to the process, but not a
systematized model of the process. When one analyzes a market using
this process, there is only one task at each part of the sequence.
In the countersignal environment, the only thing to observe is
whether or not the signal has developed any trend characteristics.
Most of the time it doesn't. In a trend environment the only purpose
one has in the daily review of the model is to determine whether one
of the "excess" patterns has developed. This is quite different and
more efficient that analyzing a market from the perspective of both
daily and weekly models and all their parameters.
Application to the hedge fund
Below are a list of exchange traded funds (ETF) which have good
liquidity and which taken together represent a composite of the
equity markets' dominant themes, i.e. growth and value, its
structure in the form of large, middle and small capitalization
stocks and the sectors and groups which compete for leadership. |
Futures would be traded with relatively small
positions using only the daily model.
Implementation
Each day after the market closes, each market traded by this process
would be assigned a position code on a spreadsheet. CTR B*, CTR S*,
JB (join buy), JS, 8s (figure 8 sell), etc. This would be done
manually at first. Then we can find some better ways to do this.
For all 50 of the ETFs and commodities, I would estimate that it
should take no more than 30 minutes. If it takes longer, we're doing
it wrong. Initially I would be the one charged with doing the
signals. But I would start training Brian, Russ and others from the
beginning.Other Considerations
What I am proposing obviously entails some observation and judgment,
albeit within the framework of a defined process. My intention is to
have the process itself shoulder the burden of the trading decisions
rather that the trader. Inevitably there will be errors of judgment
and stops will be an integral part of the process.
Does this mean that we stop trying to design systematic models for
the futures markets or equities? No! We are about to finish the NDX
models mentioned on the R&D agenda for instance, and that will be
implemented in a systematic way. But, as a practical matter, the
actual experience of trading this process on a daily basis will give
us a better chance at programming more and more of the process.
|