Way of the turtle

162
•
Way of the Turtle

---

Optimization
An effect that I call the optimization paradox is another reason
for the divergence between tested results and actual results. This
paradox is the cause of much confusion, especially for traders who
are new to computer simulation. Optimization is the process of
determining which particular numeric values to use in trading
with a system that requires computations that employ specific
numbers. These numbers are called parameters. For example, the
number of days in a long moving average is a parameter; the num-
ber of days in a short moving average is another parameter. Opti-
mization is the process of choosing the best, or optimum, values
for those parameters. Many traders say that optimization is bad
because it leads to curve fitting and results in poor performance.
I say that’s bunk!
Optimization is good when it is done correctly because it is
always better to understand the performance characteristics of
changes to a parameter than to be ignorant of them. An examina-
tion of the changes in the performance measures of a parameter
often reveals signs that performance is due to random effects or
overfitting rather than to the edge of a system. Optimization is sim-
ply the process of discovering the impact on the results of varying
a particular parameter across different values and then using that
information to make an informed decision about which specific
parameter value to use in actual trading.
Traders who believe that optimization is bad or dangerous think
that way because they do not understand the optimization paradox
and because they have seen the effects of improper optimization
that leads to what is known in statistics as overfitting.
Lies, Damn Lies, and Backtests
•