Stocks & Commodities V.

18:9 (12-20): Chart Pattern Surprises by Thomas Bulkowski

NOVICE TRADER

What Can We Learn From A Close Examination Of

Chart Patterns? Read This For A Few Ideas

Chart Pattern Surprises

What chart patterns are re-
ally reliable, portending
significant moves? This
S&C author did some re-
search, and we present the
results from his survey of Double top?
some 15,000 patterns.

I n exploring and
answering ques-
tions about the
stock data we fol-
low and the tech-
Confirmation point

nical tools we use, Declining volume trend

sometimes we make star-
tling discoveries. These dis-
coveries may help you im- FIGURE 1: DOUBLE TOP? If you think you see a double top, wait for the breaking of the neckline to increase the
prove your investment per- possibility of a successful trade.
formance or just add to your
general knowledge. I spent a year studying more than a double top. Here, prices trend upward beginning in
15,000 chart patterns while I was working on my book, December 1994 on their way to the first peak in May
Encyclopedia Of Chart Patterns. Of the dozens of 1996. The two peaks are close to the same price,
surprises I uncovered, here are a few that you may find 175/8 and 171/4, with a decline between them that
revealing and useful. appears relatively smooth (but then, an irregular
appearance is common). Prices decline between the
■ ABOUT DOUBLE TOPS two peaks by 43%, well above the 10% to 20%
• The failure rate of a double top is 65%, but minimum. Volume is higher on the left peak than on
waiting for confirmation cuts it to just 17%. the right as expected, with an overall downward trend
denoted by the descending volume line. After the
Figure 1 shows a twin-peak formation on a weekly right top, prices head down, but only reach 111/8
scale that many would say is on its way to becoming before reversing course and making a new high near
24 a year later.
Of the 1,280 twin-peak formations I looked at in
five years of daily price data, 826, or 65%, acted just
by Thomas Bulkowski like the one shown. I do not consider them to be double
Copyright (c) Technical Analysis Inc.
 Stocks & Commodities V. 18:9 (12-20): Chart Pattern Surprises by Thomas Bulkowski

tops because prices do not decline below the

Price Rise vs Trough Separation
50% confirmation point, which is the lowest point
between the two peaks. If you wait for prices to
close below the confirmation point, the failure
40% rate drops to just 17%. This means that 83% of
the time, prices continue moving down by more
PRICE RISE

than 5%.
30% It’s easy to explain the high failure rate of
twin-peak formations to decline below the con-
firmation point: It’s called a bull market. I
20%
conducted the study using data from mid-1991
to 1996, a time when the Standard & Poor’s 500
nearly doubled.
10%
21 35 49 63 77 91 105 119 Three other surprises for double tops I discov-
TROUGH SEPARATION (DAYS) ered are that tops closer together, tops with deep
troughs, and double tops with high-volume
FIGURE 2: DOUBLE BOTTOM. Bottoms close together perform better than those spaced
widely apart.
breakouts all show larger price declines; in other
words, they are better-performing double tops.

■ ABOUT DOUBLE BOTTOMS

• In a double-bottom formation, bottoms close
together perform better.

Figure 2 shows a frequency distribution graph

of the price rise versus the separation between
the two bottoms of 542 double-bottom forma-
FIGURE 3: HORN. On the weekly
scale, two price bars sticking out be-
tions. Bottoms that occur three weeks apart
low the bottom of a price formation have rises averaging 42%, while those about
make a “horn,” a formation with a four months apart have gains averaging 23%.
strong upside potential. The line is irregular, but the downward trend is
unmistakable. Why? My guess is that it is
much easier to visualize a double bottom with
bottoms three months apart than those spaced
a year apart.
If bottoms that are closer together perform
better, what happens when the weekly separa-
tion narrows to just a week or less? Answering
that question is how I define two formations I
call horns and pipes. On the weekly scale, a
horn is a bottom with two bars that define the
low separated by a single bar (Figure 3). The
Partial rise C lows are very close to each other, and the price
B
ranges of the two bars largely overlap.
A
A pipe bottom is similar to a horn bottom,
except there is no intervening bar between
them. The performance of the four variations
Partial decline — horn/pipe, top/bottom — is respectable,
with failure rates below 20%, and in the case of
pipe bottoms, an outstanding upside average
rise of 47%.

■ BROADENING TOPS
FIGURE 4: BROADENING TOPS. A partial rise or decline suggests that the breakout will be downward • A partial rise/decline from a broadening top
or upward, respectively. signals a downward/upward breakout 65%/
86% of the time.

Copyright (c) Technical Analysis Inc.

 Stocks & Commodities V. 18:9 (12-20): Chart Pattern Surprises by Thomas Bulkowski

BRUCE WALDMAN

Figure 4 shows two broadening tops, where prices enter each angled broadening formations as well. Watch for them the next
formation from the bottom. (To test the performance of tops time you trade; you’ll be able to get in at a better price than you
and bottoms, I define broadening bottoms to have prices could waiting for the breakout.
entering from the top.) The left formation shows a partial rise
at point A and another on the approach to the top trendline.
Broadening tops with partial rises correctly predicted a down- Chart patterns are behavioral records
ward breakout 65% of the time. of market action, ones that repeat time
The right formation shows a partial decline that correctly after time. These 10 patterns, distilled
predicts an upside breakout 86% of the time. If you think that
point B is a partial rise, consider what the pattern would look like
from thousands, are those that can
before point C occurred. The top trendline would touch point B give a trader the edge he needs.
until the two days at point C forced a redraw of the formation.
You could also consider point C as a premature breakout from
the smaller broadening top — a hint of things to come. ■ SYMMETRICAL TRIANGLES
Partial declines and rises not only occur in broadening tops • Pullbacks from symmetrical triangles are more likely to
and bottoms, but they occur in broadening wedges and right- occur after a high-volume breakout.

Copyright (c) Technical Analysis Inc.

 Stocks & Commodities V. 18:9 (12-20): Chart Pattern Surprises by Thomas Bulkowski

I split symmetrical triangles into tops and

bottoms the way I did with broadening for-
mations, just to discover whether they act
differently from each other. Figure 5 shows a
symmetrical triangle bottom, where prices
enter the formation from the top. The day
before the breakout, the stock traded less than
a million shares. On the breakout day, vol-
ume zoomed to 5.7 million shares. Two days
after the breakout, prices pulled back to the
Pullback lower formation boundary before resuming
their descent.
The same situation applies to symmetrical
triangle bottoms with upside breakouts. A
Breakout breakout on high volume is more likely to
retrace than a breakout on low volume. I used
Breakout
to think that a low-volume breakout was
volume more likely to throw back — retrace —
simply because there was less upside pres-
sure forcing prices higher. Now I think that a
high-volume breakout satiates buying de-
FIGURE 5: BREAKOUT. A pullback from a symmetrical triangle bottom is more likely to occur after a mand, leaving a dearth of investors willing to
high-volume breakout than a low-volume breakout. buy at higher prices. As a result, prices round
over and head back down until buying de-
mand catches up with selling pressure.
If you are a nimble trader who buys an upside
breakout, consider taking profits as prices round
over within a week or two of the breakout, then
reenter the position after the throwback is com-
pleted and prices begin rising again.
And another surprise for symmetrical tri-
angle bottoms: After a high-volume breakout,
prices tend to move further than a breakout
accompanied by low volume. This applies to
Ascending scallop
both upside and downside breakouts.

■ SCALLOPS
• Consecutive scallops† in a trend tend to
get shorter and narrower.

Figure 6 shows three ascending scallops in

a row. (A scallop is a chart formation in
which the price dips momentarily, forming
a shallow cup, before resuming its upward
course.) The first is about four and a half
months wide and $8.50 deep. The middle
scallop is about a month wide and $4.50 tall.
The right scallop is about two weeks wide
FIGURE 6: SCALLOPS. Scallops, or retracements, shorten and narrow as trends persist. and $3 high. The scallop on the right marks
the high; prices tumbled from 35 to 173/8 in
18 months. The lesson? If you are about to
take a position in a stock that shows a nar-
row or short ascending scallop — especially
when others appear in the same trend — you
might be buying near the top.

Copyright (c) Technical Analysis Inc.

Stocks & Commodities V. 18:9 (12-20): Chart Pattern Surprises by Thomas Bulkowski

■ BREAKOUTS
• High-volume breakouts show a
larger gain.

As I mentioned earlier, a high-volume

breakout propels prices farther. This
not only applies to symmetrical tri-
angle bottoms, but to other formations
as well. Figure 7 shows a complex
head-and-shoulders bottom that has
volume of only 55,900 shares the week
before the breakout. In the three weeks Neckline
afterward, the volume climbs steadily
until it hits a high of 415,000 shares. RS
LS
This stock reached a high of 393/8 in LS RS
mid-September 1995, a climb of 117% Head
from the breakout price (where prices
pierced the neckline).
Another surprising finding is that
complex head-and-shoulders bottoms
with a downsloping neckline, like that
shown in Figure 7, perform better FIGURE 7: HIGH-VOLUME BREAKOUT. Trading lore confirmed! High-volume breakouts do better than low-
than those with a rising neckline. volume ones.

■ TRIPLE BOTTOMS
• The third bottom of a triple bottom
predicts performance.

Figure 8 shows a triple bottom with the

first bottom at 173/4, the middle at 177/
1
8, and the right bottom low at 17 /2.
Volume is highest on the first bottom
and weakest on the last one. My statis-
tical review of 122 triple bottoms shows
that when the third bottom low is above
the center trough low, the formation is
more likely to outperform. Triple bot- Bottom Bottom
Bottom
toms with that configuration show
gains averaging 48%, versus 31% for
patterns with a third bottom below the
level of the second, like that shown.
The triple bottom pictured in Figure 8
has a gain of only 14%.

FIGURE 8: TRIPLE BOTTOM. A lower right bottom of a triple bottom indicates that this formation is likely to
■ FLAGS
underperform.
• Of 35 bullish chart pattern varia-
tions examined, the high, tight flag performs best. ■ HEAD-AND-SHOULDERS
• Of the 32 bearish chart pattern variations I looked at, the
The average gain is 63%, handily beating the 38% average rise complex head-and-shoulders top performs best.
for all bullish patterns. Figure 9 shows an example of a high,
tight flag. The stock doubles in about six weeks, from a launch The average decline is 27%, above the 21% average decline for
point low of 15.69 to over 30 before it meets resistance at the all bearish formations. The complex head-and-shoulders pat-
flag. The stock eases upward for several weeks before taking tern shown in Figure 10 seems to have at least two of every-
off in the new year. The stock reached a high of 1047/8, a climb thing: two heads, two left shoulders, and two right ones. This
of nearly 170% from the flag high of 391/16. Together, that’s a formation shows prices piercing the neckline at 27, but they
568% gain in just four months. pull back into a diamond top before ultimately reaching a low

Copyright (c) Technical Analysis Inc.

Stocks & Commodities V. 18:9 (12-20): Chart Pattern Surprises by Thomas Bulkowski

The average decline is 27%,

above the 21% average
decline for all bearish
formations. The complex
head-and-shoulders pattern
shown in Figure 10 seems to High, tight flag
have at least two of
everything: two heads, two
left shoulders, and two right
ones.

of 171/16 in March 2000 for a 37% decline.

Another surprise: Complex head-and-
shoulders tops with downsloping necklines,
like the one shown, result in marginally better
performance than those with level necklines,
27% versus a 26% average decline. FIGURE 9: HIGH, TIGHT FLAG. The best-performing formation: the high, tight flag generally presages
a solid upthrust.
SUMMARY
Chart patterns are behavioral records of Dual head
market action, ones that repeat time after
time. These 10 patterns, distilled from thou-
sands, are those that can give a trader the
RS
edge he needs. LS
RS
Thomas Bulkowski has a bachelor’s degree LS
in computer engineering and uses his soft-
ware engineering skills to further his full- Diamond reversal
time investment activities. .

SUGGESTED READING
Bulkowski, Thomas [2000]. Encyclopedia
Of Chart Patterns, John Wiley & Sons.

†See Traders’ Glossary for definition

FIGURE 10: HEAD-AND-SHOULDERS. Shown is a complex head-and-shoulders top with a downsloping

neckline, an example of the best-performing bearish formation.

Copyright (c) Technical Analysis Inc.