Written by and copyright © 20052017 by Thomas N. Bulkowski. All rights reserved. Disclaimer: You alone are responsible for your investment decisions.
See Privacy/Disclaimer for more information.
This butterfly pattern has nothing to do with options. The butterfly pattern was coined by Larry Pesavento, if you believe the Internet. It's a variation of the Gartley (or Gartley 222) chart pattern. I review it here.

Bearish Butterfly

Important Results for the Bearish Butterfly
Overall performance rank (1 is best): Not ranked
Break even failure rate: 15%
Average drop: 13%
Throwback rate: Not studied
Percentage meeting price target: Up to 74% (74% drop to point X)
The above numbers are based on just 117 perfect trades in both bull and bear markets. See the glossary for definitions.


Bearish Butterfly Identification Guidelines
See the below figure. It's called a butterfly because if you connect XB and BD with lines, it looks like the wings of a butterfly.
I did not find any sources that identified how the pattern is constructed except to say that the AD leg (not CD) should be longer than XA by
1.272 to 1.618 times.
However, they hinted that it's a bearish Gartley with a long move (AD), using similar Fibonacci ratios. I assumed
the ratios of the Butterfly were the same as the Gartley, but that could be wrong. I tested a loose version of the butterfly with more relaxed ratios as
well as the strict variation.
Characteristic  Discussion 
XA  Price drops from X (see figure on the right, not drawn to scale) to valley A. This is typically a large downward move to accommodate the retraces that follow. 
AB  Price retraces from valley A to peak B. The retrace should be near 61.8% of the XA move. For testing, I used a range of 61% to 62.6% (0.8 around 61.8%) 
BC  After peaking at B, price drops to C. The BC move retraces 61.8% or 78.6% of the AB rise. For testing, I chose to interpret this as a range of acceptable values. 
CD and AD  The final leg of the pattern sees price rise from valley C to peak D. The AD move is sometimes 127% or 161.8% of the XA move. 
Rise  Point D must be above X. 

The Bearish Butterfly Retraces

Bearish Butterfly Trading Tips
Trading Tactic  Explanation 
Fibonacci  Use the length of the XA move to help predict the price at which the stock will turn at D. AD should be 1.27 or 1.62 times as long as XA. Be flexible since this will rarely occur. 
Short  Once price turns at D, short the stock. 
Stop  I chose a close above D as the stop location. 
Measure Rule  The target price zone is between peak X and valley A. The associated figure shows how often price reaches those targets. 
For an explanation of the measure rule, refer to the figure on the right.
I tested the pattern and measured the drop from the high at D to either a 20% trend change (price rises 20% from a valley) or a rise above D.
I found that 74% of the patterns I looked at reached X, 15% made it to A, 43% made it to B, and 19% reached C. Twentysix percent failed to reach X (the start of the XA target area).

The Measure Rule

Notice that many of the patterns stop in what's called the corrective phase of the measured move up (BC). This is a congestion region that often stops price. See
"Completion" in the trading tactics table of the measured move up for more information.
Bearish Butterfly Testing Methodology
To test the performance of the bearish butterfly, I programmed my computer to find all peaks and valleys within 5 days. In other words, I looked for the highest high (peak)
from 5 days before to 5 days after (11 days total) the peak. I did the same for valleys.
Then it was just a matter of finding turns that fit the butterfly profile.
Since finding an AB retrace of exactly 61.8% would make the pattern too rare, I used a range of 61% to 62.6% (that is, plus or minus 0.8 around the 61.8% value).
The other ranges worked well, but didn't allow many patterns. I also removed the AD v. XA move rule that AD should be 1.272 or 1.618 times XA. As long as AD was longer than XA, I
included it. I let the pattern tell what ratio occurred most often and how well it worked.
Using daily price data from January 1990 to April 1, 2013, I found 108 stocks with butterflies (but I used over 1,000 stocks in the search). That gave me 117 individual patterns.
I did not split performance into bear or bear markets.
Bearish Butterfly Performance Statistics
Table 1: Performance of Strict and Loose Butterfly Patterns
Measure  Strict  Loose 
Measure rule: reaches X  74%  82% 
Measure rule: reaches A  15%  20% 
Measure rule: reaches B  43%  51% 
Measure rule: reaches C  19%  26% 
Break even failures  15%  17% 
Average length (days)  82  83 
Samples  117  3,436 
I show the "strict" butterfly retraces in the figure for reference. The loose interpretation allows more flexibility in the ratios. AB retrace of XA for the strict
version is 61% to 62.6%. The loose version uses a range of 50% to 78.6%. The BC/AB retrace widens to 50%  78.6%, too. For both the strict and loose versions, AD (not CD) must be at least as long as XA.
I show the performance results of tests on both types in Table 1.
The table shows how often price, after turning at D, reaches X, A, B, and C for the strict and loose interpretations (the "measure rule" lines). The numbers are similar with the loose
interpretation outperforming the strict variation.
Failures differed slightly. The break even failure rate represents moves down from D that fail to drop at least 5% before rising. This is different than the 26% that fail to reach X.
Bearish Butterfly Trading Example
Let's take a look at a trading example.
I show the butterfly on the daily chart of Genworth Financial. X is at 13.72, A is at 10.61, B is at 12.53, C is 11.10 and D is 14.77. Not shown is the exit at 5.26 on August 8, 2011,
for a drop of 64%.
The AB retrace of XA is 62%. The BC/AB drop is 74%, near the top of the range between 61.8% and 78.6%. The AD/XA move is 134%, near the low end of the 127% to 161.8% range (which
I left open ended).
After D, the stock drops until reaching the "ultimate low" (the lowest low before price rises 20%, which is not shown).
Would you be able to spot this as a butterfly if the labels were not attached?
 Thomas Bulkowski
Written by and copyright © 20052017 by Thomas N. Bulkowski. All rights reserved. Disclaimer: You alone are responsible for your investment decisions.
See Privacy/Disclaimer for more information.
If it's not broke, fix it until it is.
