CHAPTER 20
Adding Exits
In Part 2, it became obvious that most systems didn’t work all that well when trad-
ed from the short side, so from now on we will place most of our focus on the long
side. However, when testing the stops, we will test both sides simultaneously and
use the settings that work the best, on average, on the long side only. The reason
for this is basically the same as for why we want to test a system in as many mar-
kets as possible, and why we want every market, profitable or not, to influence the
final parameter settings to an equal degree. Namely, we never know when a prof-
itable market will start behaving as a unprofitable market (and vice versa).
Therefore, we want to prepare ourselves by letting the unprofitable markets influ-
ence the variable settings so that when less-profitable periods come along, at least
they won’t ruin us completely.
The same reasoning goes for the long and short trades. We never know when
the now-profitable long side will start to behave as the less-profitable short side.
Therefore, we want to prepare for that by letting the short side influence the vari-
able settings from the very beginning. This will lower the profit potential for all
long trades, but also make it more likely that a bad period will only end in a mod-
est drawdown instead of a complete disaster.
Yet another reason for not testing the long and the short sides separately from
each other is a more pragmatic one. Namely, it will double the number of test runs
necessary, to 4,000 runs per stock, or as many as 40,000 runs per stock, had I
included the random-testing procedure as well.
For each system, four stops and exit versions will be examined. Two versions
will be based on a regular percentage-stop method, and two versions will be based
on the true-range method. One version from each of these two methods will work219Copyright 2003 by The McGraw-Hill Companies, Inc. Click Here for Terms of Use.