so without letting the total time spent in a trade increase to a point where we limit
our diversification possibilities by being fully invested after only a few trades.
Even without the benefit of hindsight and the original trend filter, the
volume-weighted average system (Chapter 13) has held up well throughout the
long bear market, although a combined analysis of Tables 13.1 and 13.3 indicates
that the average profit per trade on the long side has decreased over the last sev-
eral years while, at the same time, the average trade length has gotten a little short-
er. Both these findings suggest that the thrusts to the upside are not as large and
forceful in a downtrend as they are in an uptrend.
The different lookback periods for the short and the long sides, given the
results in Tables 13.5 and 13.6, confirm that short-term down moves generally are
more explosive in nature than short-term up moves. Another, more philosophical
way to look at it, is that the shorter lookback period for the down side indicates
that a market discounts and then forgets bad news quicker than it does good news.
The RS system No. 1 (Chapter 9) is one of three systems that, later in the
book, will be used as a (trend) filter rather than as an outright entry trigger. As
such, it needs to have a slightly longer average trade length to open up a window
for the other systems to trigger an entry. However, because this normally also
means that the time spent in the market will increase, you have to be careful that
the system doesn’t spend too much time in the market and thereby make itself use-
less. In this case, the average trade length on the long side comes out to approxi-
mately 20 days.
With the time spent in the market at 70 percent, the system will spend an
average of about 10 days out of the market for every 30-day period. Putting this in
relation to, for example, the meander system, when using RS system No. 1 as a fil-
ter for the meander system, the long side of the meander system will decrease its
time in the market to approximately 30 percent and its number of trades per mar-
ket to approximately 130. This still feels like a little too much for optimal diversi-
fication, but we will use these numbers to see what we can learn from them.
At first glance, the code for the relative-strength bands system (Chapter 11)
looks very complicated, with a lot of calculations and comparisons between mar-
kets. However, if you stop and think about it, it really isn’t all that difficult. All
we’re really doing is comparing the trend of several different markets—something
we do in our heads all day long anyway, at least if we also tinker around with the
more subjective kind of technical analysis. This is only a way to formalize it all,
so that we can digest more information from more markets. Nothing more, noth-
ing less.
If for nothing else, one important thing we learned from this system was that
testing only on three markets doesn’t even come close to producing results that
will be reliable and robust in the long run. However, by expanding the research to
include several markets, we had enough observations to allow us some room for
tinkering with the system. As it turned out, this system was much better off get-186 PART 2 Trading System Development