Encyclopedia of Chart Patterns. Thomas N. Bulkowski
Чтение книги онлайн.
Читать онлайн книгу Encyclopedia of Chart Patterns - Thomas N. Bulkowski страница 44
DC/BC extension. If you invert the ratio found in the last step, you use it to find the price of D. In this example, we found the closest Fibonacci number to be .5, so we'd expect point D to be twice as far away. To put it another way, let's pick a point D where the ratio of DC to BC is 2. The high at point D is 67.36, so the equation is (67.36 – 60.32)/(63.85 – 60.32) or 1.99 (or about 2).
We found turn ABC to obey one of the numbers listed in the table, and we also found point D using a Fibonacci extension (one of them listed in the table), so we found a valid AB=CD pattern.
In this example, price turns down at D, just like it's supposed to. However, the drop is brief (to E).
Hills and valleys. I excluded any pattern that had a peak or valley outside of the turns as described in the table.
Volume. Although it may not look like a downward volume trend in this example (F), linear regression says it recedes. In fact, you'll see volume trending downward in most AB=CD patterns and other chart pattern types, too. If volume trends upward, that's fine. Don't throw away a pattern because of an unusual volume trend.
Duration. I imposed a 6‐month limit to the length of most chart patterns, including the AB=CD. It's an arbitrary limit.
Focus on Failures
Figure 2.3 shows an example of a failed bearish AB=CD pattern (labeled as turns ABCD). The pattern fails in multiple ways. The first is that price doesn't make it up to the predicted point D.
Turn A has a low price of 134.82, B has a high of 148.28, and C has a low of 139.95. That gives a BC/BA retrace of .618, so the turn qualifies as a valid AB=CD. It predicts that point D should be at 153.41 using the formula: D = (B – C)/Ratio + C.
As the figure shows, point D falls well short of the target, which I show as F. Instead, price rises only to D before dropping to E. Imagine that you wanted to trade the anticipated rise to point D by buying the stock soon after turn C. You placed a stop a few pennies below C, and you would have been stopped out at E, which reached a low of 139.79, slightly below the low at C.
As I mentioned in the Identification Guidelines, there can't be a low below C on the way to the calculated point D. Point E stops the search for D because it's below the low at C. If you ignore that rule, then you have discovered the second failure type.
Figure 2.3 Price is supposed to turn down at F, but the search for point D ends when point E is below C.
The second way this AB=CD fails is when price continues rising instead of turning at D. If we ignore the lower low violation at E and assume price climbs to F, where turn D should be (153.41), then look what price does. It continues rising, doesn't it? So price fails to turn at the new target D and moves to G, peaking at 158.77, well above the 153.41 target. Shorting the stock at F would have tested a trader's courage against a rising price trend when the stock climbed to G.
Look back at Figure 2.2 where it shows another example of how the pattern fails to see price decline much after D. Price drops from 67.36 (point D) to 65.22 (point E), a drop of 2.14 points or 3%. Could you make money shorting the stock at D, knowing that if you traded it perfectly, you'd make 3%?
If you owned the stock (long) and sold at D thinking price would drop, you'd be happy that the stock dropped to E, but your joy would turn to sadness when the stock continued climbing up to F and beyond. It would say you'd made a mistake.
Of course I chose Figure 2.3 to highlight the failure of this pattern to perform as expected. That's what the Focus on Failures section is supposed to do. In the next section, we'll see what the numbers say about how this pattern behaves.
Statistics
Table 2.2 shows general statistics for the bearish AB=CD and tailored to the Fibonacci pattern. That means you won't find explanations for the table entries in the Glossary. Most are self‐explanatory.
Number found. If you can program your computer to find them, you'll discover that they come out like worms after a heavy downpour. They were so plentiful that I limited the number catalogued per stock.
I found the first pattern in February 1990 and the most recent in February 2020, finding them in 884 stocks. Not all stocks covered the entire period, and some stocks no longer trade.
Table 2.2 General Statistics
Description | Bull Market | Bear Market |
---|---|---|
Number found | 2,649 | 696 |
Breakeven failure rate | 26.3% | 10.2% |
Average decline after D | –12.7% | –21.6% |
Volume trend | 54% Downward | 58% Downward |
Performance Up/Down volume | –13%U, –13%D | –19%U, –24%D |
Breakeven failure rate. For those patterns that see price make it up to D and reverse there, this is a measure of how often price fails to drop more 5% (below the high at D). The bull market value is high (ranking fourth out of five, where one has the lowest failure rate), but the bear market rate, at 10.2%, is the worst of the five bearish Fibonacci patterns I looked at.
Average decline after D. This is a measure of the drop after point D, for those patterns seeing price make it up to point D and reverse there. As one would expect, the drop in bear markets is larger than in bull markets. If you were to trade the bearish AB=CD pattern perfectly and frequently, this is how much you could make on average. Commissions were not included.
Volume trend. I used linear regression from the start to end of the pattern and found it trends downward most of the time, but it's near random.
Performance Up/Down volume. I checked performance when volume was trending up or down. This applies only to those patterns that turned down at D. The bull market shows no performance difference, but in bear markets, the performance difference is wider and substantial. Patterns with downward‐sloping volume see price drop an average of 24% compared to a 19% decline for those with up‐sloping volume.
Trading Tactics