Читать книгу Encyclopedia of Chart Patterns - Thomas N. Bulkowski - Страница 134
Statistics
ОглавлениеTable 10.2 shows general statistics for right‐angled and descending broadening formations.
Number found. I found 1,150 patterns in 668 stocks from August 1991 to October 2019. However, that included bear market patterns, and there were not enough of those to include in the statistics. Not all of the stocks covered the entire period, and some no longer trade. Notice that the pattern favors upward breakouts, judging only by the numbers.
Table 10.2 General Statistics
| Description | Up Breakout | Down Breakout |
|---|---|---|
| Number found | 601 | 335 |
| Reversal (R), continuation (C) occurrence | 38% R, 62% C | 58% R, 42% C |
| Reversal, continuation performance | 45% R, 41% C | –14% R, –17% C |
| Average rise or decline | 43% | –15% |
| Standard & Poor's 500 change | 12% | –2% |
| Days to ultimate high or low | 244 | 51 |
| How many change trend? | 54% | 27% |
Reversal (R), continuation (C) occurrence. Upward breakouts act as continuation patterns most often, but downward breakouts act as reversals. Recall that a reversal has price leaving the chart pattern in a direction opposite the way it entered. Continuation patterns have price exit in the same direction as the prevailing price trend.
Reversal/continuation performance. Upward breakouts show better performance from reversal patterns, but downward breakouts favor continuations.
Average rise or decline. Shown in the table is the average rise or decline. Nothing spectacular here.
Standard & Poor's 500 change. I compared the date of the breakout to the ultimate high or low posted by the chart pattern and compared those to the index. The index shows it didn't perform nearly as well as the broadening pattern over the same holding period.
Days to ultimate high or low. Upward breakouts take an average of about 8 months to reach the ultimate high, but downward breakouts end faster, in about 2 months. The yardage traveled is different, though, because upward breakouts rise 43% and downward breakouts drop 15%. If you do the math, we find that the downward breakout should have reached bottom in 85 days, not 51. So, price dropped faster than it climbed.
How many change trend? I added this measure to help decide which chart patterns lead to outsized gains (a gain or loss of more than 20%). The results shown in the table are mid‐list numbers. The upward breakout number is good (more than half the patterns see decent gains), but downward breakouts fall well short of a 20% loss (just 27% drop that far).
Table 10.3 shows various failure rates for the two breakout directions. How do you make sense of the table? It is easier to understand by example. I found that 21% of the patterns with upward breakouts failed to see price rise more than 5%. Over half (51%) of the patterns failed to see price rise more than 25% after an upward breakout.
Table 10.3 Cumulative Failure Rates
| Maximum Price Rise or Decline (%) | Up Breakout | Down Breakout |
|---|---|---|
| 5 (breakeven) | 125 or 21% | 77 or 23% |
| 10 | 65 or 32% | 72 or 33% |
| 15 | 46 or 39% | 50 or 59% |
| 20 | 43 or 46% | 45 or 73% |
| 25 | 29 or 51% | 29 or 81% |
| 30 | 44 or 59% | 21 or 88% |
| 35 | 28 or 63% | 15 or 92% |
| 50 | 65 or 74% | 17 or 97% |
| 75 | 55 or 83% | 9 or 100% |
| Over 75 | 101 or 100% | 0 or 100% |
Table 10.4 Breakout and Post‐Breakout Statistics
| Description | Up Breakout | Down Breakout |
|---|---|---|
| Breakout direction | 64% up | 36% down |
| Performance of breakouts occurring near the 12‐month low (L), middle (M), or high (H) | L 46%, M 49%, H 41% | L –19%, M –14%, H –12% |
| Throwbacks/pullbacks occurrence | 64% | 69% |
| Average time to throwback/pullback peaks | 5% in 6 days | –7% in 6 days |
| Average time to throwback/pullback ends | 12 days | 12 days |
| Average rise/decline for patterns with throwbacks/pullbacks | 43% | –15% |
| Average rise/decline for patterns without throwbacks/pullbacks | 42% | –15% |
| Percentage price resumes trend | 71% | 55% |
| Performance with breakout day gap | 35% | –18% |
| Performance without breakout day gap | 44% | –15% |
| Average gap size | $0.45 | $0.37 |
You read downward breakouts in a similar manner. Almost a quarter of the patterns (23%) saw price drop no more than 5%. Half (59%) didn't see price drop more than 15%.
Table 10.4 shows breakout‐related statistics.
Breakout direction. Price breaks out of the pattern upward almost twice as often as downward.
Yearly position, performance. I sorted the breakout price into one of three buckets, each a third of the yearly high–low range. Then I checked performance for patterns in those three buckets. The table shows that there's not a big performance difference for upward breakouts. Clearly, though, you'll want to avoid trading patterns within a third of the yearly high (price rises 41%).
Downward breakouts also show those near the yearly high underperforming those in the other two buckets. It suggests people should avoid momentum trading this chart pattern and focus on bottom fishing for candidates.
Throwbacks and pullbacks. Throwbacks and pullbacks occur almost two‐thirds of the time, and it takes stocks less than 2 weeks to return to the breakout price.
I didn't see any significant performance difference for patterns with and without throwbacks or pullbacks. With other chart patterns, if a throwback or pullback occurs, performance suffers (on average).
Price resumes the upward move 71% of the time (which is quite good) and the downtrend 55% of the time (which needs improvement, but it's not bad).
Gaps. Most of the time (for other chart patterns), breakout day gaps help performance but we see that behavior in this pattern only after a downward breakout.
The rumors you've heard are true: Table 10.5 shows size‐related statistics.
Height. Tall patterns perform substantially better than short ones. How do you use this result? Compute the pattern's height from the price of the top trendline to the last touch of the lower trendline (the lowest low in the pattern). Divide the difference by the breakout price. If the result is above the median shown in the table, then you have a tall pattern; below the median means it is a short one. Invest only in tall patterns unless you feel confident of your assessment.
Table 10.5 Size Statistics
| Description | Up Breakout | Down Breakout |
|---|---|---|
| Tall pattern performance | 46% | –18% |
| Short pattern performance | 39% | –13% |
| Median height as a percentage of breakout price | 9.7% | 10.4% |
| Narrow pattern performance | 41% | –15% |
| Wide pattern performance | 45% | –16% |
| Median width | 50 days | 42 days |
| Short and narrow performance | 40% | –13% |
| Short and wide performance | 37% | –10% |
| Tall and wide performance | 48% | –18% |
| Tall and narrow performance | 43% | –18% |
Width. Wide patterns outperform narrow ones but not by an amount worth telling Mom about. I used the median length to separate narrow patterns from wide ones.
Height and width combinations. The worst performance comes from patterns that are both short and wide. You will want to avoid those. The best performance happens for patterns both tall and wide. The performance differences are large enough that you should pay attention to them when contemplating a trade.
Table 10.6 shows volume‐related statistics.
Volume trend. Volume trends upward most often, but the direction is near random.
Rising/Falling volume. Broadening patterns with rising volume trend show better performance after the breakout than when volume recedes (as measured from the start to the end of the pattern).
Breakout volume. Technical analysts seem to place a lot of emphasis on heavy breakout volume, but as the table shows, the results I found are not statistically significant (for heavy or light breakout volume versus performance).
Table 10.7 shows how often price reaches a stop location. I split the pattern in half (I sedated it ahead of time) and checked how often price returned to the various parts on the journey from the breakout to the ultimate high or low.
Table 10.6 Volume Statistics
| Description | Up Breakout | Down Breakout |
|---|---|---|
| Volume trend | 58% up | 53% up |
| Rising volume trend performance | 45% | –16% |
| Falling volume trend performance | 39% | –14% |
| Heavy breakout volume performance | 43% | –15% |
| Light breakout volume performance | 42% | –16% |
Table 10.7 How Often Stops Hit
| Description | Up Breakout | Down Breakout |
|---|---|---|
| Pattern top | 78% | 2% |
| Middle | 26% | 17% |
| Pattern bottom | 5% | 74% |
For example, broadening patterns with upward breakouts saw price return to the top of the pattern 78% of the time. It suggests you don't want to place a stop‐loss order there. If you stick the stop order at the bottom of the pattern, it'll avoid being hit 95% of the time, but the resulting loss may be higher than you can tolerate. So do check both the position of the stop and the size of the loss should the stop trigger. A volatility stop may be just what you're looking for. I've hidden some in the Glossary (see “Volatility stop”).
Table 10.8 shows the performance over three decades.
Performance over time. For upward breakouts, the 2000s posted stellar gains. Downward breakouts suffered during the same time (and that makes intuitive sense).
Failures over time. Failures (failure of price to rise or fall more than 5% after the breakout) after upward breakouts happened most often in the 2010s, and downward breakouts also show a lot of failures in that decade, but the 2000s edged them out. There were two bear markets during the 2000s, but I used a coffee filter to remove them from the statistics.
Table 10.9 shows statistics for busted patterns.
Table 10.8 Performance and Failures Over Time for Bull Markets
| Description | Up Breakout | Down Breakout |
|---|---|---|
| 1990s | 35% | –16% |
| 2000s | 53% | –13% |
| 2010s | 35% | –17% |
| Performance (above), Failures (below) | ||
| 1990s | 19% | 17% |
| 2000s | 19% | 27% |
| 2010s | 25% | 25% |
Table 10.9 Busted Patterns
| Description | Up Breakout | Down Breakout |
|---|---|---|
| Busted patterns count | 175 or 29% | 144 or 43% |
| Single bust count | 85 or 49% | 100 or 69% |
| Double bust count | 54 or 31% | 5 or 3% |
| Triple+ bust count | 36 or 21% | 39 or 27% |
| Performance for all busted patterns | –13% | 43% |
| Single busted performance | –22% | 60% |
| Non‐busted performance | –15% | 43% |
Busted patterns count. I'm always surprised at how many patterns fail to move more than 10% away from the pattern, then reverse and shoot out the other side of the pattern. Here, we have over a quarter (29%) to nearly half (43%) bust the up or down breakout, respectively.
Busted occurrence. I sorted the number of busts into one, two, or more than two busts (triple+). Single busts happen most often. Notice that triple+ busts place second after a downward breakout. I've seen this in other patterns, too.
Busted and non‐busted performance. Look at the performance after a single busted downward breakout: Broadening patterns see price rise an average of 60%. When I read that, my jaw swung open, my dentures fell out, and I had to pick them up off the carpet. Single busts also outperformed after a busted upward breakout.
The 60% number is so startling that if you can find a busted downward breakout, then do consider trading the pattern.
Non‐busted patterns seem to have the edge (and it's slight) on performance if you disregard the single busted patterns and downward breakouts.
