system was to make it more aggressive than it was originally. More often than not,
when testing an old system on new data, the “best” way to improve it, or to curve
fit it to the previously unseen data (if we want to be really critical), is to make it
more difficult for the system to enter into a trade. The fact that we didn’t have to
do that this time is a good sign that this very basic entry technique does a good job
of identifying high-probability trading opportunities and that it will continue to
work in the future as well.
The two major disadvantages of the system, as of now, are that the average
trade length is about twice as long as we would want it to be, and the time spent
in a trade is way too long. We really need to get the time spent in a trade down by
something like 10 to 15 percent.
When revising the research on the Harris 3L-R pattern variation system
(Chapter 14), I found it had one major flaw built into it. This had nothing to do
with the original logic, created by Michael Harris, but was completely of my own
doing when adding the stops and exits to go with the entry technique.
However, when altering the exit rules, it also turned out the system traded a
little too infrequently. To come to grips with that, I had to make a few modifica-
tions, such as getting rid of trading on the open the day after the signal and not ask-
ing for two consecutive lows (highs) before the market turns and breaks higher
(lower). Because both modifications not only increase the results, but also relax
the criteria for getting into a trade, it is fair to say that this new version of the sys-
tem actually is less curve fitted than the original one, and therefore also more
robust and more likely to hold up in the future.
As a general observation, it is evident that I have not done a particularly good
job of making these systems profitable on the short side. However, there are sev-
eral reasons for that. First, when many of these systems were featured for the first
time in Active Tradermagazine, the bear market wasn’t developed enough to test
the systems and, in many instances, I decided to make them long only because that
is what most of my readers are most familiar with.
In revising the research, I was stuck with the original idea and logic behind
the systems and, therefore, had to work with what I had. Maybe, just maybe, I
could have done a better job with the short side as well, had I been able to start
from scratch. Granted, some of you still might say that I could have done a better
job with what I had, but the question is, how far should I have taken the research
to do so? Sure, I could have indulged in more testing and optimizations, but I have
other topics that I need share with you as well. The purpose behind all this is to
give you a glimpse inside my head while I work with different system concepts for
Active Tradermagazine, so that you can go out and do a better job yourself. I can’t
provide you with the definitive answers, because there are none.
Another important observation is that many of these systems are quite com-
plex in their nature, which contradicts what I wrote in Trading Systems That Work,
that I like simplicity, and that preferably a system should not have more than three188 PART 2 Trading System Development