could help. It also could be a good idea to add a trailing stop that, for example,
never lets you lose more than half your profits, or that tightens as your profits
increase. To avoid tying up valuable resources in trades that go nowhere, a time-
based stop could be added that exits all trades after a certain number of days, no
matter what.
As usual, it also would be a good idea to add a trend filter, such as a long-
term moving average, and allow the system to trade short by reversing the logic. It
has been proven that a trend filter—allowing only those trades that go with the
direction of the underlying trend—improves performance in most systems.
Finally, note that many of the stocks traded in this example weren’t tradable
until a few years ago, which explains the exceptionally long flat period halfway
through the testing period (not shown). Had we been able to test the same 30
stocks throughout the entire period, it’s highly likely that it would have increased
performance considerably. This is indicated by the latest drawdown, which comes
nowhere near the current drawdown in the NASDAQ, and which would probably
have been smaller still had we traded the market from the short side as well.Revising the Research and Modifying the System
Testing this system anew over the period January 1990 to February 2002, on the
30 stocks making up Dow Jones index, the 30 stocks with the highest market
capitalization in NASDAQ 100, and five different stock indexes (Dow
Industrial, S&P 500, NASDAQ 100, S&P MidCap, and Russell 2000) show that
the results on the long side hold up well. An average profit factor of 1.48 and an
average risk factor of 0.28 are not bad. A risk–reward ratio of 0.96 means that
we can be pretty sure that the average profit will not turn negative when the sys-
tem is traded for real. Also, of all the markets tested, more than 80 percent were
profitable, which is very good indeed. Table 14.1 shows the combined results
from all markets tested.
As Table 14.2 shows, adding the short side to the equation seems not to have
been such a good idea. The number of profitable markets sank to 61.5 percent, the
profit factor is just barely high enough to compensate for slippage and commis-
sion, and the risk factor is a mere 6 cents per dollar risked. The risk–reward ratio
for the entire portfolio, which measures the risk-adjusted return by dividing the
average profit per trade by the standard deviation of the average profit per trade,
also is too low, at 0.04, thus indicating that the trades produced by this system are
all over the place.
One big negative with this system is its average trade lengths of 23 and 12
days, respectively. This is way too long, considering that the trade is based on a
three-day pattern and the fact that it should be a short-term strategy. Looking
through the charts, it is easy to find trades that last for as long as six months and
longer. We will talk more about this in Part 3, where we discuss various types ofCHAPTER 14 Harris 3L-R Pattern Variation 165