Читать книгу Encyclopedia of Chart Patterns - Thomas N. Bulkowski - Страница 39

Imagine Dave likes to bowl and he plays three games per match. Over a month he competes five times and wins three out of five matches with one tie. If we tally his win/loss record not for the 15 individual games, but for the five matches, we find he wins 75% of the time (three of four contests with one tie). We might conclude that he's a good bowler or those competing against him are not. I built the following tables just like I described with Dave. They show how often an aspect of a chart pattern leads to better or worse performance.

Table 1.1: Reversals versus Continuations shows the results of the first contest: Which types of patterns perform better, those acting as reversals or continuations? Before I answer that, what is a reversal and a continuation?

A pattern acting as a reversal happens when price enters and exits a chart pattern from different directions (down going into the pattern and exiting out the top, or rising into a pattern and breaking out downward). Patterns acting as continuations see price enter and exit the chart pattern in the same direction (down going into the pattern and breaking out downward, or rising into a pattern and exiting upward).

For example, Figure 1.7 shows double bottom AB acting as a reversal because price drops into the pattern and leaves it going upward. The downtrend reverses. The rectangle at C in Figure 1.2 is a continuation pattern. Price rises into the start of the rectangle (from B) and exits out the top. That is, price continues in the direction of the prevailing price trend.

Table 1.1 Reversals (R) versus Continuations (C)

	Bull Market, Up Breakout	Bear Market, Up Breakout
Winner	Reversals (55%)	Reversals (73%)
Performance	43% R, 42% C	30% R, 26% C
	Bull Market, Down Breakout	Bear Market, Down Breakout
Winner	Continuations (87%)	Continuations (75%)
Performance	–14% R, –16% C	–22% R, –23% C

Winner. Table 1.1 tells us that reversals work best after upward breakouts and continuations work best after downward breakouts. For example, I found 55% of chart patterns acting as reversals outperformed continuation patterns in bull markets after upward breakouts. Almost all (87%) of the chart patterns after downward breakouts in bull markets showed patterns acting as continuations beating those which acted as reversals.

Performance. This line in the table shows the average gain or loss for the contests. For example, reversals in bull markets after upward breakouts gained 43%. Continuation patterns saw price rise 42%. Notice that the differences between the contests are often narrow. Continuations, for example, lead by one or two percentage points. The narrow lead is a warning that the indicator is weak as a predictor of future performance.

Table 1.2: Height. Do tall patterns outperform short ones? Tall or short is a measure of the height of the chart pattern divided by the breakout price.

Winner. This line shows how chart pattern performance varies with height. I contend that tall patterns outperform short ones, and we find that belief is true in all market conditions and breakout directions. For example, tall patterns win all of the contests (100%) after downward breakouts in bull markets.

Performance. The performance averages are wide, too, with tall patterns gaining an average of 46% and short ones gaining just 39% (bull market, up breakout). I believe height is a key indicator of future performance, so you'll want to trade tall patterns and avoid short ones.

Table 1.3: Width. Do wide patterns outperform short ones? I measure width from the start to the end of the pattern.

Winner. The numbers in the table tell how well width works as an indicator of chart pattern performance. Contests from patterns in bull markets perform better when they are wide. Bear market patterns show a tie.

Performance. Most performance differences are marginal (one or two percentage points). For example, after downward breakouts in bear markets, wide patterns see price drop 23% on the way to the ultimate low, but narrow ones see price drop an average of 22%. Thus, width is not a strong predictor of performance, but it can give you an edge in bull markets (where performance differences are wider).

Table 1.2 Height: Tall (T) versus Short (S)

	Bull Market, Up Breakout	Bear Market, Up Breakout
Winner	Tall (89%)	Tall (89%)
Performance	46% T, 39% S	30% T, 25% S
	Bull Market, Down Breakout	Bear Market, Down Breakout
Winner	Tall (100%)	Tall (87%)
Performance	–17% T, –13% S	–23% T, –21% S

Table 1.3 Width: Wide (W) versus Narrow (N)

	Bull Market, Up Breakout	Bear Market, Up Breakout
Winner	Wide (81%)	Tie (50%)
Performance	45% W, 40% N	28% W, 27% N
	Bull Market, Down Breakout	Bear Market, Down Breakout
Winner	Wide (85%)	Tie (50%)
Performance	–16% W, –14% N	–23% W, –22% N

Table 1.4 Breakout Day Gap (G) versus No Gap (N)

	Bull Market, Up Breakout	Bear Market, Up Breakout
Winner	Gap (68%)	Tie (50%)
Performance	45% G, 42% N	28% G, 28% N
	Bull Market, Down Breakout	Bear Market, Down Breakout
Winner	Gap (85%)	Gap (57%)
Performance	–16% G, –15% N	–23% G, –22% N

Table 1.4: Breakout day gap. Do gaps that occur on the day of breakout help performance?

Winner, Performance. After downward breakouts in bull markets, I found that 85% of the chart pattern types showed a gap helping performance. The chart patterns dropped an average of 16% versus 15% for those not showing a breakout gap.

Notice that the percentages on the performance line are close, suggesting gaps are not a strong performance indicator, either. Because I measure performance using the opening price the day after the breakout day gap, you can participate in the better performance that a gap may provide.

Table 1.5: Throwbacks and pullbacks. Throwbacks and pullbacks are features I love. When they appear, they invariably hurt performance. The table shows the blood. A throwback happens after an upward breakout from a chart pattern when price soars but then returns to the breakout price (or comes close to it) within a month. Pullbacks are the same except the breakout is downward.

Winner, Performance. For example, 97% of the time I found that throwbacks hurt performance after upward breakouts in bull markets. In contests of patterns with downward breakouts in bear markets, all of them (100%) showed that pullbacks hurt performance. The average decline of those contests was 26% for those patterns not showing a pullback (N) and 20% for those that did pull back (P). For a downward breakout, that's a wide difference.

Table 1.5 Throwbacks (T) and Pullbacks (P) versus None (N)

	Bull Market, Up Breakout	Bear Market, Up Breakout
Winner	Throwbacks (97%)	Throwbacks (89%)
Performance	40% T, 48% N	25% T, 33% N
Occurrence	64%	65%
	Bull Market, Down Breakout	Bear Market, Down Breakout
Winner	Pullbacks (91%)	Pullbacks (100%)
Performance	–14% P, –17% N	–20% P, –26% N
Occurrence	64%	63%

Occurrence. The Occurrence line in the table shows how often a throwback or pullback occurred, on average. Throwbacks and pullbacks happen almost two out of three times (63% to 65%).

Table 1.6: Rising or falling volume. The table shows how patterns behave if volume is rising or falling from the start of the pattern to the end, found using linear regression.

Winner. For upward breakouts, performance improves just over 60% of the time if volume is rising. Downward breakouts are mixed with bull markets in patterns showing falling volume doing best but bear markets show a tie.

Performance. The performance numbers are close, though, especially for downward breakouts. For example, patterns with rising volume (R) saw price climb 44% to the ultimate high (bull market, up breakout). Patterns with falling volume (F) showed gains averaging 42%. The volume trend is not a good predictor of future performance.

I think technical analysts put too much emphasis on volume. (Consider that for every share sold, one is bought. If institutions are selling massive amounts of shares, then other institutions are buying those shares.)

Table 1.7: Heavy versus light breakout day volume. I compared breakout day volume with the prior month.

Table 1.6 Rising (R) versus Falling (F) Volume

	Bull Market, Up Breakout	Bear Market, Up Breakout
Winner	Rising (62%)	Rising (61%)
Performance	44% R, 42% F	29% R, 27% F
	Bull Market, Down Breakout	Bear Market, Down Breakout
Winner	Falling (56%)	Tie (50%)
Performance	–15% R, –15% F	–22% R, –22% F

Table 1.7 Heavy (H) versus Light (L) Breakout Day Volume

	Bull Market, Up Breakout	Bear Market, Up Breakout
Winner	Heavy (79%)	Heavy (79%)
Performance	43% H, 41% L	29% H, 26% L
	Bull Market, Down Breakout	Bear Market, Down Breakout
Winner	Heavy (67%)	Light (55%)
Performance	–15% H, –15% L	–22% H, –22% L

Winner, Performance. The table shows that heavy breakout day volume suggests better performance most of the time. For example, I found that 79% of the time after upward breakouts in bear markets, high volume led to better performance by 29% (for heavy volume) versus gains averaging 26% for patterns with light breakout day volume.

You will notice that the percentage difference is not great, especially after downward breakouts (which show ties). Breakout volume is not as good a predictor of performance as many believe.

Table 1.8: Trend change. This table is different from the others. It shows how often price continues rising or falling more than 20% after the breakout. My thinking is that if a chart pattern shows large post‐breakout moves, then it should be easier to make money trading them. Conversely, a pattern that has lots of small moves may be difficult to trade profitably.

Occurrence. Only in bull markets do the majority (55%) of chart pattern types see price rise more than 20%. The worst performance (28%) comes from chart pattern types with downward breakouts in bull markets. That poor performance makes intuitive sense because the market trend is upward but the breakout is downward. It's like swimming against the current, so you'd expect the swimmer to struggle.

Notice that downward breakouts in bear markets place second (49%). It's another indication that trading with the trend (upward breakouts in bull markets and downward breakouts in bear markets) leads to better performance.

Table 1.8 Trend Change

	Bull Market, Up Breakout	Bear Market, Up Breakout
Occurrence	55%	46%
	Bull Market, Down Breakout	Bear Market, Down Breakout
Occurrence	28%	49%

Table 1.9 Single Busted Patterns (S) versus Proxy (P)

	Bull Market, Up Breakout	Bear Market, Up Breakout
Winner	Single bust (97%)	Single bust (75%)
Performance	–22% S, –15% P	–24% S, –22% P
Single bust occurrence	53%	76%
	Bull Market, Down Breakout	Bear Market, Down Breakout
Winner	Single bust (94%)	Single bust (79%)
Performance	53% S, 41% P	40% S, 28% P
Single bust occurrence	70%	63%

Table 1.9: Single busted patterns versus proxy. The final table in this chapter shows how well busted patterns perform against non‐busted ones.

Winner. Single busted patterns outperform their non‐busted counterparts a good portion of the time. That's especially true for chart patterns in bull markets. They win more than 90% of the contests. Bear markets also win contests, but at a more sedate pace.

Performance. I compared the performance of single busted patterns with their proxy (P), the non‐busted pattern. In most cases, the performance difference is quite wide.

For example, in bear markets after downward breakouts, single busted patterns with downward breakouts saw price rise an average of 40% above the top of the chart pattern. The non‐busted patterns saw price climb just 28%, on average.

Occurrence. This line tells how often a single busted pattern happens (versus double or more than two busts). The higher the number, the easier it'll be to trade a busted pattern that wins big. That means you are more likely to trade a single busted pattern than a pattern that double or triple+ busts.

Bull markets with upward breakouts show fewer single busts (53%) than the others. I think that's because when the pattern busts, price will be heading lower and that's going against the bullish current. It invites a double bust.

The chapters that follow look at individual chart patterns. I use statistics to help discover how they behave, and I share my findings with you.