ting rid of several of the original rules and input variables. With fewer variables
tested on more markets, the end versions of this system are much more likely to
hold up in the future.
In this case, we ended up with both a long and a short version of the system.
Both versions were profitable, but unfortunately this profitability also came at the
price of very volatile results, as illustrated by the low risk–reward ratios. For the
remainder of the book, this system will be used as a (trend) filter for the other
entry techniques, which should help bring down the time spent in the market by
approximately 50 percent for most of them.
If we thought that the code for the relative-strength bands system was com-
plicated, then what must we think about the code for the rotation system (Chapter
12)? Again, it really isn’t that complicated at all. All that we’re doing is compar-
ing the strength of the trend between several markets. But this time, instead of
looking at the distance between the relative-strength line and its moving average,
we’re looking at the slope of the moving average. This is very much like having
several charts in front of you, comparing them all with each other, trying to come
up with the chart that has the most strength and also is likely to continue strong in
the future.
Because the original version of the rotation system was tested in a way sim-
ilar to the relative-strength bands system, we again had to expand the number of
markets tested to come up with statistically reliable results. This time, however, the
original results had held up much better since the system was originally featured.
However, because all markets were selected at random, the chance also exists that
the results from both of these relative-strength systems can be either better or
worse than a more thorough examination would reveal. The possibility for that is,
however, something most statistical examinations have to live with. The best we
can do is make the sample large enough to minimize the possibility.
Note also that I decided to change the logic for the short side halfway
through the research process—just like that—and without really starting the
research anew. Of course, the correct thing to do would have been to back up and
step through the entire research process again, but because we’re not using the
word “research” here in its most scientific form, but rather as a word for describ-
ing my thought process, I decided to just move on as if nothing happened. When
it comes to the entry techniques, I am not so much interested in coming up with
the optimal solution, but rather a solution that makes sense to me. Right or wrong?
We can argue about that later.
For the remainder of the book, this system too will be used as a filter for the
other entry techniques, which should help bring down the time spent in the mar-
ket by approximately 50 percent for most of them.
The hybrid system No. 1 (Chapter 8) was one of the systems that held up best
over time. But, as was the case with many of the other systems, it did not fare that
well on the short side. Interestingly, it turned out that the best way to modify thisCHAPTER 16 Evaluating System Performance 187