Way of the turtle

Luck and Time
Time tends to favor the truly excellent traders rather than merely
the lucky ones. If there are 16 traders who are lucky for 10 years,
after another 15 years their performance is much more likely to be
closer to average. Conversely, if you consider only 5-year track
records, the number of seemingly excellent traders who are actually
only lucky traders skyrockets. This is because over shorter measure-
ment periods the extent of the random effects is also much greater.
What happens to variance in our test if we use a shorter time
frame, perhaps only the 3.5 years from January 2003 through June
2006? For this period, the average performance for the random
Lies, Damn Lies, and Backtests
•
161

---

entry system was a 35 percent return with a MAR ratio of 1.06. The
real systems did considerably better. The Triple Moving Average
system returned 48.5 percent with a MAR ratio of 1.50. The
Bollinger Breakout system returned 52.2 percent with a MAR ratio
of 1.54. The Dual Moving Average system returned 49.7 percent
with a MAR ratio of 1.25.
As for the random tests, how many lucky traders emerged from
the 100 tests in the simulation? How many beat our best system’s
performance purely on the basis of luck? Seventeen out of 100 had
a MAR ratio better than 1.54; of those 17 tests, 7 had a return
higher than 52.2 percent. The very best random trader returned
71.4 percent with a drawdown of 34.5 percent for a MAR ratio of
2.07. All this is something to think about the next time you are look-
ing at a three-year track record with excellent performance.
When you are looking at a short-term track record, you should
realize that much of what you are seeing is attributable to luck. If
you want to know whether a particular trader is one of the lucky
average or one of the excellent few, you need to dig deeper than
the track record and focus on the people behind the track record.
Good investors invest in people, not historical performance.
They know how to identify traits that will lead to excellent per-
formance in the future, and they know how to identify traits that
are indicative of average trading ability. This is the best way to over-
come random effects. The good news for those who are doing his-
torical simulations is that it is fairly easy to detect when backtest
results probably are due to random effects rather than to a system’s
edge. Chapter 12 will discuss ways in which this can be done, but
first let’s look at two more reasons why backtest results do not
match reality.