106
•
APPLICATIONS OF TRADING STRATEGIES
A few results of empirical research may be helpful in under-
standing chart patterns. We are grateful for the efforts of Thomas
Bulkowski in sorting chart patterns according to statistical criteria.
Readers of his works will f ind a statistics summary that includes per-
formance statistics for all of the chart patterns
discussed in his ency-
clopedia (Bulkowski, pp. 654– 657). In the following paragraphs,
we refer to the findings of Bulkowski for the 3-point chart patterns we
have tackled (and some of the candlestick patterns discussed earlier).
Before showing the statistics in tabular form, we need to state
Bulkowski’s criteria.
Bulkowski
def ines a
failure rate
, which is the percentage of for-
mations that do not work as expected, including 5 percent failures. The
numbers apply to formations once they stage a breakout (conf irming
the formation).
Bulkowski also distinguishes between reversal or consolidation.
The letter “R” appears if the majority of formations act as reversals
of the price trend, and the letter “C” appears for consolidations (or
continuations, as we call them). If both
R and C appear in an entry,
then the chart pattern has no overriding majority of either type.
Furthermore, Bulkowski separates
throwbacks
from
pullbacks.
A throwback is an upside breakout that returns prices to the top of the
formation or trend line boundary. A pullback is a downside breakout
that returns prices to the bottom of the formation or trend line bound-
ary. Both occur after a breakout and return within 30 days. The per-
centages for throwbacks apply to formations with
upside breakouts
only; pullback percentages apply to downside breakouts only.
In addition, Bulkowski provides a measure for
average rise or de-
cline.
He typically measures the average rise or decline from the price
on the breakout day (using the daily high or low)
that is closest to the
formation. The ultimate high or low is the highest or lowest point before
a signif icant change in trend (typically a 20 percent price change,
which is measured high to low).
From these figures, Bulkowski computes a value for
likely rise or
decline
and tabulates a frequency distribution of the results. The most
likely rise or decline is the range with the
highest frequency and usu-
ally excludes the rightmost column.
Most important to our readership is Bulkowski’s
rank by score.
He separates the table entries into bullish (1 to 35) and bearish (1 to
32) formations and then ranks them by their score. The score is the av-
erage rise or decline times the most likely rise or decline divided by
c04.qxd 6/16/03 4:18 PM Page 106