Way of the turtle

Measuring What You Cannot See
There are many ways to quantify risk, which is one way to factor in
the pain you would have encountered while trading a particular
system. Here are some common measures that I find useful:
96
•
Way of the Turtle

---

1. Maximum drawdown: This is a single number that
represents the highest percentage loss from peak to
subsequent equity low during the course of a test. In Figure
7-4 this would be the 65 percent drawdown that was due to
the price shock of the 1987 crash.
2. Longest drawdown: The largest period from a peak in equity
to a subsequent new peak. This is a measure of how long it
would take to regain new equity highs after a losing streak.
3. Standard deviation of returns: This is a measure of the
dispersion of returns. A low standard deviation of returns
indicates that most returns are near the average; a high
standard deviation indicates that returns vary more from
month to month.
4. R-squared: This is a measure of smoothness of fit to the line
that represents the CAGR%. A fixed-return investment such
as an interest-bearing account would have an R-squared
value of 1.0, whereas a very erratic set of returns would have
a value lower than 1.0.
The Flip Side of Risk: Rewards
There are many ways to quantify a reward, which in the case of a
particular trading system relates to the amount of money you might
expect to earn when trading that method. Here are some common
measures that I find useful:
• CAGR%: The compound annual growth rate, also known as
the geometric average return, reflects the rate of growth that
when compounded equally over the specific period would
By What Measure?
•
97

---

have resulted in identical ending equity. For simple interest-
bearing accounts this is equal to the rate of interest itself.
This measure can be affected greatly by a single period of
high returns.
• Average one-year trailing return: This is the measure of the
average return for a rolling one-year period. This measure
gives a better sense of what the typical return might be in any
specific one-year period. It is relatively less sensitive to a single
period of high returns for tests of more than a few years.
• Average monthly return: This is the average of each single
month’s returns over the period of the test.
In addition to these single number measures, I find it useful to
examine the equity curve itself as well as a graph that highlights the
distribution of monthly returns, as in Figure 4-4 back in Chapter
4. I also like to examine the individual monthly returns over time,
as shown in Figure 7-5, which gives the monthly returns for the
Donchian Trend system from 1996 to June 2006.
I find that a graph like the one in Figure 7-5 gives a pretty good
indication of the relative pain versus reward one can expect and is
much more revealing than a single figure or set of figures.