As of 05/01/2024
Indus: 37,903 +87.37 +0.2%
Trans: 14,865 -44.88 -0.3%
Utils: 904 +10.56 +1.2%
Nasdaq: 15,605 -52.34 -0.3%
S&P 500: 5,018 -17.30 -0.3%
|
YTD
+0.6%
-6.5%
+2.5%
+4.0%
+5.2%
|
 39,325 or 37,000 by 05/15/202415,820 or 14,600 by 05/15/2024925 or 865 by 05/15/202416,700 or 15,200 by 05/15/20245,250 or 4,850 by 05/15/2024 |
As of 05/01/2024
Indus: 37,903 +87.37 +0.2%
Trans: 14,865 -44.88 -0.3%
Utils: 904 +10.56 +1.2%
Nasdaq: 15,605 -52.34 -0.3%
S&P 500: 5,018 -17.30 -0.3%
|
YTD
+0.6%
-6.5%
+2.5%
+4.0%
+5.2%
| |
|
39,325 or 37,000 by 05/15/202415,820 or 14,600 by 05/15/2024925 or 865 by 05/15/202416,700 or 15,200 by 05/15/20245,250 or 4,850 by 05/15/2024 | ||
Initial release: 10/31/2023.
When price peaks, will the candle shape, color, or volume signal a big drop? I decided to find out and this article describes the results.
I did not find any combination that would be worth exploring for a trading setup to avoid a large decline. However, small declines (20% or more) show more promise than do large declines (50%).
Let me give you a visual of what I'm trying to do.
The figure shows a chart of 3D Systems (DDD) on the monthly scale. From the January 2014 peak of 97.28 (log scale), the stock plummets to a low of 3.55 in October 2023 (not shown).
The two vertical red lines show where the peak and following three price bars line up with volume.
The inset shows the four price bars, starting with the peak at candle 1 to the last bar, at candle 4. These are the four bars that I looked at in my study.
In this study, I found a peak, checked for a large drop, and looked at the peak price bar and the following three bars, for a total of four bars.
In the first test, I found looked for price to drop at least 50% below the peak, measured from high to close to avoid one-bar flash crashes. I stopped looking for the low if price climbed above the original peak (that is, if price climbed above bar 1).
I excluded stocks that spun off operations (which causes a gap on the chart) or were priced below $5 at the peak.
I show results that use data from as early as 1973 to October 2023. I used 453 stocks in the analysis, taken from the portfolio I use on a daily basis and selected decades ago for performance.
Table 1 shows three factors that I measured using monthly charts: height (taller or shorter than the 4-month average taken before the peak), body color (white means a close above the open, black means a close below the open, ties are discarded), and volume (versus a prior month).
| Metric | Peak | Peak+1 | Peak+2 | Peak+3 |
| Height: Tall | 74% | 74% | 74% | 74% |
| Height: Short | 73% | 73% | 74% | 73% |
| Body color: Black | 74% | 73% | 74% | 74% |
| Body color: White | 74% | 75% | 73% | 74% |
| Volume: Light | 73% | 74% | 73% | 73% |
| Volume: Heavy | 77% | 75% | 76% | 75% |
The benchmark decline was 74% (not shown in the table), meaning I looked for a minimum decline of 50% and found that they averaged 74% from a peak (high to close).
For example, if the first price bar was taller than the prior 4-month average, the associated drop from the peak averaged 74%. The next price bar (Peak+1), if it was tall, and succeeding tall price bars (Peak+2 and Peak+3) all saw price drop a similar amount, also measured from the Peak price bar (high to close).
The next line down shows the loss after short price bars appeared. For example, if the Peak was short, price dropped an average of 73% below the peak. If bar Peak+1 was short, the stocks averaged declines of 73%, as measured from the high price at Peak to the lowest close before price made a new high (if it ever did).
The table shows that performance between short or tall price bars didn't vary much.
Candle color, which is determined by the position of the opening and closing prices, didn't predict how far price would fall, either.
Heavy volume on the peak and/or three succeeding price bars did show performance improvement (a larger drop). We can conclude that heavy volume (at least twice the prior month) may indicate a large drop coming.
Table 2 shows the performance of stocks with combinations of candle color. In an earlier study, I found that three black candles represented a sell signal on the monthly chart, as part of a trading setup. My hope is that the results in this study would reinforce the setup's findings. They do not. The setup may work as designed (you find price rising following a trend line before it peaks and shows three black candles).
| Metric | Loss | Samples |
| Benchmark | 74% | 1,156 |
| Two black candles | 73% | 442 |
| Three black candles | 74% | 287 |
| Four black candles | 74% | 160 |
| Two white candles | 75% | 59 |
| White then black | 74% | 506 |
| Black then white | 75% | 149 |
Again, I looked at four price bars, starting with the peak before price declined at least 50%. In this test, I use consecutive candles to see if colors helped predict a larger decline.
The benchmark decline after the peak measured 74% and that came from 1,156 examples.
The next line down in the table looked for the peak candle to have a black body as well as the next candle. That combination showed price dropped an average of 73% from 442 samples.
The remainder of the table didn't show much of a variation whether you were looking at two, three, or four candles with the same color or different color combinations. I did not test all possible permutations of candle color for the four price bars.
It could be that candle color, height, or volume patterns are more effective for smaller drops (such as 20% below a peak) or a range (such as 50% to 60% below a peak). Let's explore drops of at least 20% below a peak.
In this test, I changed the 50% down figure to 20%. Did it make a difference in results?
| Metric | Peak | Peak+1 | Peak+2 | Peak+3 |
| Height: Tall | 46% | 46% | 47% | 47% |
| Height: Short | 43% | 44% | 44% | 44% |
| Body color: Black | 46% | 45% | 46% | 47% |
| Body color: White | 45% | 47% | 44% | 43% |
| Volume: Light | 45% | 45% | 45% | 45% |
| Volume: Heavy | 50% | 49% | 51% | 53% |
Table 3 shows results similar to Table 1 except that the benchmark decline is 45%. That's the average decline from a peak that shows price dropping at least 20%.
Tall candles (taller than the average of the prior 4-months) show further declines than short ones. The difference between a tall candle and a short one is less than a few percentage points.
Black candles suggest a bigger decline than white candles, but the performance difference is small (at most 4 percentage points).
Heavy volume (twice or more) the prior month's volume suggests a larger decline. Notice that candle Peak+3 shows an 8 percentage point difference but for larger declines (50% or more), the difference is only two percentage points (from Table 1). I find it odd that the fourth candle in the series has the highest performance difference.
| Metric | Loss | Samples |
| Benchmark | 45% | 4,700 |
| Two black candles | 46% | 1,756 |
| Three black candles | 47% | 1,009 |
| Four black candles | 50% | 469 |
| Two white candles | 47% | 218 |
| White then black | 45% | 2,163 |
| Black then white | 47% | 563 |
Table 4 shows the same measure as Table 2, except as it applies to smaller declines, those of 20% or more.
Four black candles shows the largest average decline of 50%. The other combinations don't deviate much from the benchmark 45%.
-- Thomas Bulkowski
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.
Legal notice for paid links: "As an Amazon Associate I earn from qualifying purchases."
|
My Stock Market Books
|
My Novels
|
on 