it becomes a breeze to find the values for the input variables that suit you and the
system best. Usually I test the system for three input variables at a time.
For example, for this book I tested the stop-loss and profit target distances
together with the maximum allowed trade length. I then compared the variables
two at a time, disregarding the third variable for the moment. In this way, I got 10
different test runs per variable combination and market tested. When looking at the
surface chart, I always try to come up with the variable combination that seems to
work best on average, over all test runs and markets tested.
An alternative way is to test for only two variables at a time and substitute
for the third variable a random-number generator that will produce a unique
sequence of trades each time you run the system through the market. The expert
exits system was developed this way for the June 2002 issue of Active Tradermag-
azine. Be forewarned, however, that this method is very time consuming.
Using surface charts to examine several different stop and exit levels for five
of our systems, it turned out that the average true range concept worked best for
most systems. Interestingly, however, the percentage concept, which worked best
for the most short-term systems, also happened to work the best on both sides of
the market (both long and short trades).
The testing was done on both sides of the market for all systems tested,
although in the end we would only apply it to the long side. Similarly, the exits
were applied to all systems before the relative strength filters were added. We did
this because I believe that the more market conditions and circumstances that are
allowed to influence the final parameter setting, the more robust and reliable the
system will be when traded on future data.
Another way might have been first to apply the filters and then test only the
side of the markets that would be traded. This would probably have boosted the
profits considerably on the historical data, as the system would have been more
curve fitted to those exact conditions. The drawback would have been that the sys-
tem would have been less robust and reliable for real-life trading in the future,
where everything still can happen.
While testing the stops and exits, we found several examples of how chang-
ing from a stop loss to a profit target resulted in completely different characteris-
tics for a system. This is because a system often produces a series of entry signals
that might or might not result in a trade, depending on if we’re in a trade already
or not. Altering the way we exit the market also alters which signals will be avail-
able to enter on, which in turn will alter the most optimal exit levels and trade
lengths, and so on. These differences were further exaggerated when adding the
three systems saved for filtering, in Chapter 21.
The research in Chapter 21 also revealed that using a relative-strength filter
doesn’t necessarily add to the bottom line and the robustness of the combined
strategy. We identified several reasons. One of these reasons was that trading the
stock that seems to be the strongest both in the intermediate-term and short-term284 PART 3 Stops, Filters, and Exits