Chapter
Variables
Most of the time, when building a system, we talk about making the system
robust, meaning it should be as little curve fitted as possible to increase its chances
of holding up in the future as well as it has in the past or when developed.
However, just using one word for this important consideration—the word robust—
is not enough if we want to get a firm grip on how to go about increasing our
chances for future success. If you ask me, I think there are at least three things we
need to consider, all of which fit within the term robust, as we have used it so far.
Instead of just talking about robustness, I would like to propose that we talk about
robustness, stability, and consistency. Let’s start with the term stability, which is a
concept we’ve stressed throughout Part 3.
Stability means that the system should work equally as well, on average and
over time, using several different variable settings. For example, if you use a nine-
day moving average for a system, there is no way for you to find out in advance if
you might have been better off with a 10-day average over a certain time period.
If that would have been the case, your nine-day setting should not have worked that
much worse, and definitely not have produced catastrophic results, whereas the
10-day setting would have shown a profit. In the long run and over several subpe-
riods, it should not have mattered that much which of the two settings you would
have gone with.
Another example is when the best variable setting, in hindsight, turned out to
be way off from the one you used. In that case, the best setting might have result-
ed in a decent profit, but you suffered a rather severe drawdown. That drawdown
should not, however, have been of such magnitude that it forced you out of the
game. The best way to assure stability, as I have defined it here, is to make the sys-
271
Copyright 2003 by The McGraw-Hill Companies, Inc. Click Here for Terms of Use.
