As of 08/22/2019
Indus: 26,252 +49.51 +0.2%
Trans: 10,073 34.32 0.3%
Utils: 841 +0.59 +0.1%
Nasdaq: 7,991 28.82 0.4%
S&P 500: 2,923 1.48 0.1%

YTD
+12.5%
+9.8%
+17.9%
+20.4%
+16.6%

27,200 or 24,900 by 09/01/2019
10,700 or 9,550 by 09/01/2019
850 or 790 by 09/01/2019
8,200 or 7,400 by 09/01/2019
3,000 or 2,750 by 09/01/2019

As of 08/22/2019
Indus: 26,252 +49.51 +0.2%
Trans: 10,073 34.32 0.3%
Utils: 841 +0.59 +0.1%
Nasdaq: 7,991 28.82 0.4%
S&P 500: 2,923 1.48 0.1%

YTD
+12.5%
+9.8%
+17.9%
+20.4%
+16.6%
 
27,200 or 24,900 by 09/01/2019
10,700 or 9,550 by 09/01/2019
850 or 790 by 09/01/2019
8,200 or 7,400 by 09/01/2019
3,000 or 2,750 by 09/01/2019
 
Initial release: 6/20/2018. Rank updated 7/25/19.
This article describes my analysis of the bearish AB=CD pattern as described by publicly available information and common sense rules to determine valid patterns. Additional rules may or may not improve performance. I tested the pattern using only the below identification guidelines.
The bearish AB=CD is a measured move up chart pattern except that the turns are located using Fibonacci ratios. By using three different measures, I found that the pattern isn't bearish at all. It's bullish between 61% and 77% of the time.
However, the pattern correctly predicts point D 69% of the time. By using the next higher Fibonacci ratio in the measure rule, you can boost the accuracy somewhat, to 74%. I'll explain this in Trading Tips.
Bearish AB=CD

Important Bull Market ResultsOverall performance rank for up/down breakouts (1 is best): 19 out of 56/38 out of 53
Break even failure rate: 17%/27%
Average rise/decline*: 43%/13%
Throwback/pullback rate: 66%/66%
Percentage meeting predicted point D: 69%/
The above numbers are based on at least 2,000 perfect trades. See the glossary for definitions. * As measured from the peak at point D 
Find four turns where the ratio of one leg to another is close to the Fibonacci numbers listed in the prior table. What does "close" mean? In my tests, I used 1% because this pattern occurs so frequently, a larger percentage is not necessary. See Trading Tips for an explanation of this. Testing used a reciprocal of the Fibonacci retrace to predict D. That excluded some of the ratios listed in the table, BCD row, because they are not reciprocals of the ABC retrace (meaning I excluded 314% and 224%). 
Bearish AB=CD Ratios

The following explains how to use the measure rule to predict point D. Why is this important? Because if you can identify a valid ABC turn, you can determine how far price will rise (to D). This method works 69% of the time, according to my tests but you can increase this to 74%.
Find the ABC retrace which obeys one of the Fibonacci ratios listed in the previous table. Because the CD leg is supposed to equal the length of leg AB, if we know the length of AB, we can compute point D.
For example, if point A is at 25, B is at 35 and C is at 28.82. Point C retraces 61.8% of the AB move. The length of AB is 10. Add this to the low at C gives a target of 38.82.
In this case, the BC leg is 1210.98 or 1.02. Multiplying this by 2 gives 2.04 and adding it to the low at C gives 13.02. That value becomes the D target.
This method works the same as the prior one, 69% of the time.
You can increase the accuracy of the measure rule by using the next higher Fibonacci number in the formula to find point D.
In this example, instead of using 50%, use 61.8%. The multiplication number would change from 2 (1/0.5) to 1.618 (1/0.618). So we have 1.02 * 1.618 or 1.65. Add that value to the low at C (10.98) gives 12.63, closer than the prior method's 13.02. This projection method works 74% of the time.
The traditional measure I use to determine whether a chart patterns is bullish or bearish is to determine the breakout direction. A close above the top of the pattern (above point D, refer to the image on the right) means an upward breakout. A close below the bottom of the pattern (below point A) means a downward breakout. Using this measure, I found that 61% of bearish AB=CD patterns breakout upward, not downward. That means the pattern is bullish, not bearish.
However, this measure is misleading for this pattern. Why? Because price ends the pattern at D, the high of the pattern, a close above D is much easier to achieve than a drop all the way down to close below A.
Another measure is to look at patterns which meet the measure rule target (that is, they climbed up to the predicted turning point at D). I used a 5% window to gauge whether the stock turned when it was supposed to (within 5% of D). If it did, then I concluded a bearish reversal had occurred.
How many actually turned within 5% of point D? Just 23%. In other words, it's more accurate to say that price will continue rising 77% of the time.
Perhaps the 5% window is too small. I opened it to 10%. How many bearish AB=CD patterns turned within 10% of the target (for those that reached the target)? Answer: 36%. That's still less than half (meaning the pattern is bullish, not bearish).
In the third method, I visually inspected 125 patterns with upward breakouts only (that is, price met or exceeded the measure rule target) to see if price turned down at D or continued rising (that is, made a meaningful drop or not). It's a very subjective analysis, so results tend to vary. I found that 81 or 65% showed price continuing higher. In other words, the pattern is bullish 65% of the time.
You can find your own bearish AB=CD patterns and gauge this for yourself. And as I said, additional identification rules might improve performance.
Consider a trade in AES corporation, the chart of which is shown on the right.
Point A has a low price of 10.23. B has a high of 11.32. And C has a low of 10.65. Does this qualify as a potential bearish AB=CD pattern?
Let's find out. The BC/BA retrace is (11.3210.65)/(11.3210.23) or 61.5%, which is close (within one percentage point) to the target ratio of 61.8%
So it's a valid pattern. What is the predicted price of D, which is to say, how far will price rise?
If the pattern works as it's supposed to, the stock will peak at 11.73. How do I know? Let's go through the calculation.
The reciprocal of the ABC retrace is 1/0.618 or 1.618. Multiply this by the BC leg and add it to the low at C. We get (11.3210.65)*1.618 + 10.65 = 11.73
A simpler approach is to compute the length of the AB leg and add it to the low at C. We have 11.3210.23 = 1.09. Add 1.09 to the low at C gives 1.09+10.65=11.74. The differences between the two methods are because of round off.
The stock closed at D at 11.74.
In this case, the stock reached the target but look what happened. Price kept rising. Oops. As I mentioned, my tests showed price continuing rising past D the vast majority of the time, and this is an example.
 Thomas Bulkowski
See Also

AB=CD is a registered trademark of Scott Carney.
Support this site! Clicking any of the books (below) takes you to Amazon.com. If you buy ANYTHING while there, they pay for the referral.
Diplomacy  the art of letting someone have your way.