Читать книгу Encyclopedia of Chart Patterns - Thomas N. Bulkowski - Страница 82
Statistics
ОглавлениеTable 6.2 shows general statistics for the big M. Notice that the tables are not split into up and down breakouts. That's because valid big Ms don't have upward breakouts, at least not in this universe.
Number found. I uncovered most big M patterns manually by searching through more than 1,700 stocks (finding big Ms in 772 stocks) from mid‐1991 to September 2019. I also automated finding them and checked each one to be sure they adhered to the ID guidelines. That process helped to unearth the stinkers like the one shown in Figure 6.3, where the pattern has a nice left‐side rise but a mutant right‐side decline.
Reversal (R), continuation (C) occurrence. All of the patterns acted as reversals of the uptrend by definition.
Average decline. Both bull and bear markets show performance that beats the average of all other chart patterns, but not by an amount that Mom would like to hear about, especially at three in the morning.
Standard & Poor's 500 change. Using the same hold time from the breakout to the ultimate low as the associated big M, we find that the big M dropped substantially more than the S&P 500 index. To put it another way, you can see how the general market helped the individual stocks perform.
Days to ultimate low. The median time to reach the ultimate low is a month for bull markets and about three weeks for bear markets. The table shows the average, which is longer in both bull and bear markets.
Speed trap? I checked price velocity and found that bear markets see price drop almost twice as fast as bull markets.
How many change trend? As a measure of how well the chart pattern works (trends), I counted how many big Ms saw price drop more than 20% after the breakout. The thinking here is that a 20% move represents a trend change. For example, if the markets are bullish, a drop of 20% turns the market bearish. I applied the same measure to chart patterns.
Table 6.2 General Statistics
| Description | Bull Market | Bear Market |
|---|---|---|
| Number found | 2,090 | 569 |
| Reversal (R), continuation (C) occurrence | 100% R, 0% C | 100% R, 0% C |
| Average decline | –17% | –22% |
| Standard & Poor's 500 change | –2% | –11% |
| Days to ultimate low | 59 | 40 |
| How many change trend? | 32% | 51% |
Table 6.3 Cumulative Failure Rates
| Maximum Price Decline (%) | Bull Market | Bear Market |
|---|---|---|
| 5 (breakeven) | 298 or 14% | 46 or 8% |
| 10 | 465 or 37% | 81 or 22% |
| 15 | 374 or 54% | 84 or 37% |
| 20 | 288 or 68% | 70 or 49% |
| 25 | 232 or 79% | 85 or 64% |
| 30 | 147 or 86% | 58 or 75% |
| 35 | 96 or 91% | 41 or 82% |
| 50 | 152 or 98% | 79 or 96% |
| 75 | 36 or 100% | 24 or 100% |
| Over 75 | 2 or 100% | 1 or 100% |
As the table shows, higher numbers of big Ms see price drop more than 20% in bear markets as opposed to bull ones. This makes intuitive sense (it's like a receding tide will lower all boats). I consider values above 50% to be exhilarating.
Table 6.3 shows how failure rates climb for small drops after the breakout. For example, 298 big Ms or 14% of the patterns in bull markets failed to see price drop more than 5% after the breakout. Over a third (37%) failed to see price drop more than 10%. In bear markets, the pattern performs better, with 37% failing to see price drop more than 15%.
You can use this table to help determine the chance of price making an extended decline. Want to make 50% after shorting your big M? Good luck with that. Only 2% in bull markets see price drop that far (that is, 98% fail to see price drop that far).
Table 6.4 provides breakout and post‐breakout statistics.
Breakout direction. As I mentioned, if a big M breaks out upward, then you have a case of mistaken identity. Valid big Ms only break out downward.
Yearly position, performance. You will see the big M pattern appear anywhere in the yearly price range. Bull markets seem to find them hiding near the yearly high almost half the time. Bear markets see the breakout price hibernating in the middle of the yearly price range most often.
Mapping performance over the yearly price range doesn't see any statistically significant differences.
Pullbacks. A pullback occurs after a downward breakout. Price must return to the breakout price within a month by definition. The table says that price drops 7% to 10% in 6 days and then returns to the breakout price in a round‐trip total of 12 days.
Table 6.4 Breakout and Post‐Breakout Statistics
| Description | Bull Market | Bear Market |
|---|---|---|
| Breakout direction | 100% down | 100% down |
| Performance of breakouts occurring near the 12‐month low (L), middle (M), or high (H) | L –17%, M–17%, H –16% | L –23%, M –22%, H –22% |
| Pullback occurrence | 67% | 63% |
| Average time to pullback bottoms | –7% in 6 days | –10% in 6 days |
| Average time to pullback ends | 12 days | 12 days |
| Average decline for patterns with pullbacks | –16% | –20% |
| Average decline for patterns without pullbacks | –19% | –27% |
| Percentage price resumes trend | 60% | 55% |
| Performance with breakout day gap | –17% | –23% |
| Performance without breakout day gap | –16% | –22% |
| Average gap size | $0.85 | $0.48 |
Patterns without pullbacks perform better than do those with pullbacks. That behavior is not unique to big Ms; we've seen it in other species of chart patterns. After a pullback completes, it resumes dropping 55% to 60% of the time on average.
Gaps. Breakout day gaps don't help or hurt performance much. Notice that the gap size in bull markets is almost twice the size of bear market gaps. Hmm. That's a surprise.
I measured the median gap size (19 cents in bull markets and 21 cents in bear markets) and used those as the difference between tall and short gaps. I found that in bull markets, gaps taller than the median saw price drop an average of 19% after the breakout. Short gaps lost just 16%.
In bear markets, the results were 25% decline for tall gaps and 21% decline for short gaps.
Table 6.5 shows size statistics for the big M.
Height. We see that regardless of the market condition (bull or bear), tall patterns outperform short ones, but not by much. To compute this, measure the height of the pattern from the highest peak to the lowest valley between the two peaks and divide the result by the price of the lowest valley. If the result is larger than the percentage shown in the table, then you have a tall big M.
Width. Width doesn't offer any helpful performance improvement, although narrow patterns' performance is slightly better (but probably statistically insignificant). Measure the width of the pattern from peak to peak and compare it to the 26‐day median. If the result exceeds the median, then you have a wide pattern.
Table 6.5 Size Statistics
| Description | Bull Market | Bear Market |
|---|---|---|
| Tall pattern performance | –18% | –23% |
| Short pattern performance | –16% | –22% |
| Median height as a percentage of breakout price | 12.4% | 15.1% |
| Narrow pattern performance | –17% | –23% |
| Wide pattern performance | –16% | –22% |
| Median width | 26 days | 26 days |
| Short and narrow performance | –16% | –22% |
| Short and wide performance | –15% | –22% |
| Tall and wide performance | –18% | –22% |
| Tall and narrow performance | –18% | –24% |
Height and width combinations. Upward breakouts show better performance if the big M is tall (regardless of width). Tall and narrow patterns outperform in bear markets.
Table 6.6 shows that volume doesn't play a key role in performance of big Ms.
Volume trend. Volume trends downward every two out of three trades on average.
Rising/Falling volume. There's no performance difference between big Ms with a rising or falling volume trend (as measured from peak to peak using linear regression).
Breakout day volume. Only bear markets show better performance after heavy breakout volume. Heavy breakout volume typically leads to better performance for other chart patterns (not so much with downward breakouts, though, so it's weird that big Ms are different).
Table 6.6 Volume Statistics
| Description | Bull Market | Bear Market |
|---|---|---|
| Volume trend | 67% down | 68% down |
| Rising volume trend performance | –17% | –22% |
| Falling volume trend performance | –17% | –22% |
| Heavy breakout volume performance | –17% | –23% |
| Light breakout volume performance | –17% | –20% |
Table 6.7 How Often Stops Hit
| Description | Bull Market | Bear Market |
|---|---|---|
| Pattern top | 2% | 0% |
| Middle | 16% | 11% |
| Pattern bottom | 70% | 68% |
Table 6.7 shows a table new to this edition: how often price hits a stop.
I split the pattern into three pieces: top, middle, and bottom. The top is the higher of the two peaks (B in Figure 6.3). The bottom of the pattern is the low at the valley between the two peaks (point C). The middle is, well, midway between the top and bottom.
In bull markets, price reached the top of the pattern just 2% of the time on the way to the ultimate low (after the breakout). Although bear markets show 0%, I actually found two samples where price reached the top.
If you wish to place a stop, then price will rise up to the middle of the pattern 16% of the time in bull markets, or recover (perhaps during a pullback) to the bottom of the pattern 70% of the time. Bear markets show similar performance.
In Table 6.8 I mapped the breakout date into one of three decades, which the table shows, split into performance in the top portion of the table and failures in the bottom portion. Because bear markets only happened in the 2000s, they are not included in the statistics.
Performance over time. The 1990s showed the best performance from big Ms, but not so's you'd notice unless you traded a lot of them.
Failures over time. Failures happened at about the same rate over all decades studied, with the 2000s showing the most and the 1990s showing the least. These “failures” are breakeven failures, that is, failure of the stock to drop more than 5% after the breakout.
Table 6.8 Performance and Failures Over Time for Bull Markets
| Description | Bull Market |
|---|---|
| 1990s | –18% |
| 2000s | –15% |
| 2010s | –16% |
| Performance (above), Failures (below) | |
| 1990s | 13% |
| 2000s | 16% |
| 2010s | 14% |
Table 6.9 Busted Patterns
| Description | Bull Market | Bear Market |
|---|---|---|
| Busted patterns count | 693 or 38% | 101 or 18% |
| Single bust count | 466 or 67% | 57 or 56% |
| Double bust count | 15 or 2% | 3 or 3% |
| Triple+ bust count | 212 or 31% | 41 or 41% |
| Performance for all busted patterns | 38% | 34% |
| Single busted performance | 55% | 55% |
| Non‐busted performance (big W) | 46% | 30% |
Table 6.9 shows how busted patterns performed. As you will recall, a busted big M happens when price breaks out downward, drops less than 10% (an arbitrary amount), reverses, and closes above the top of the big M (above the higher of the two peaks). The new, upward breakout busts the downward one. Multiple busts can follow.
Busted patterns count. Because the sample count for big Ms was high, I found quite a number of busted ones, from a sprinkling in bear markets (18%) to over a third in bull markets (38%). That makes intuitive sense because the bear market trend will help take price lower. It reminds me of watching a barge move down river. The water level drops near shore and gets all churned up. As a kid, I was afraid I'd be pulled under while swimming or even canoeing (even though the barge was across the river, about a quarter mile away).
Busted occurrence. Single busts happened most often followed by triple busts. While you may find that odd (that double busts don't come in second), you join me in my confusion. I've seen this trend in other pattern types and don't have an explanation for why it happens. A double bust would send price lower (that is, the first bust sends price upward and the second bust turns price downward), if that is any help.
Busted and non‐busted performance. Grouping single, double, and three or more busts together sees price rise 38% after busted downward breakouts in bull markets. Separating out single busts, we see price rise an average of 55% in bull markets (that's measured from the top of the big M to the ultimate high). That's a nice return if you can capture most of it.
I used big Ws as a proxy for a non‐busted big M. They don't see price perform as well as their busted sisters.
If you can find a busted big M, buy it and pray that price continues rising for a long time (a single bust). The 55% rise compares to a 46% average gain for big Ws (which break out upward). See the table for bear market results.
