two strongest of them from the long side. By adding a trend filter to each market,
results should improve considerably.
The asymmetrical stops are designed to distinguish between different market
characteristics that depend on the long-term trend. For example, because markets
generally are more volatile when they fall than when they rally, the trailing stop is
placed further away for the weakest of the markets than it is for the two strongest
ones (to avoid getting stopped out too often).
The trailing stop will only stop out the quickest and most adverse moves
against the trade. If the market moves slowly against the trade for a prolonged peri-
od, the trade won’t be exited until the price reaches the stop-loss level.Revising the Research and Modifying the System
The reasoning for this system is pretty much the same as for the relative-strength
bands, discussed previously. For the original system, I only looked at the three
indexes: S&P 500, NASDAQ 100, and DJIA. Therefore, I once again had to
expand the research for this book by adding two more indexes and 20 stocks to
achieve statistically reliable results. The code below is set to calculate the relative
strength between one of the indexes and the four indexes it’s compared to.
All stocks were once again picked at random from a population of 30 stocks
in each subgroup. Each stock that was to be traded was compared to eight other
stocks from the same group. For example, from the NASDAQ 100 group, I ran-
domly picked QUALCOMM (QCOM) to be compared against Altera (ALTR),
Bed Bath & Beyond (BBBY), Chiron (CHIR), Dell (Dell), Flextronics (FLEX),
Immunex (IMNX), KLA Tencor (KLAC), and Nextel (NXTL). Only QCOM
would be traded. The other markets are only there to construct a randomly created
relative-strength indication for QCOM to generate trading signals. In total, this
will give us 25 different trading sequences to analyze. Because the original model
was asymmetrical in nature, with different rules for long and short trades, each
side will be analyzed individually.
Table 12.1 shows the results of the original model, traded from the long side
only. As you can see, this is a good start, with very high profit and risk factors. The
major negative is that most standard deviation measures are a little too high.
For the original testing of this system, I decided to always go long, the two
strongest indexes out of the three making up the portfolio. During this research
process, I decided to go long two out of five indexes and long a stock if it belonged
to the four strongest in its group of nine stocks. I settled for these numbers, and
decided to keep them constant, because during the many years of bull market, the
system most likely will seem to perform better the more markets traded, and if
that’s the case, we might just as well trash this relative-strength system complete-
ly. However, during this stage of the research process, what seems to be best isn’t
always what will work best in the end. By limiting the number of markets tradedCHAPTER 12 Rotation 139