As of 10/15/2020
Indus: 28,494 19.80 0.1%
Trans: 11,989 +101.34 +0.9%
Utils: 875 +0.03 +0.0%
Nasdaq: 11,714 54.86 0.5%
S&P 500: 3,483 5.33 0.2%

YTD
0.2%
+10.0%
0.4%
+30.6%
+7.8%

29,300 or 27,400 by 11/01/2020
12,000 or 11,000 by 11/01/2020
915 or 840 by 11/01/2020
12,800 or 11,300 by 11/01/2020
3,700 or 3,400 by 11/01/2020

As of 10/15/2020
Indus: 28,494 19.80 0.1%
Trans: 11,989 +101.34 +0.9%
Utils: 875 +0.03 +0.0%
Nasdaq: 11,714 54.86 0.5%
S&P 500: 3,483 5.33 0.2%

YTD
0.2%
+10.0%
0.4%
+30.6%
+7.8%
 
29,300 or 27,400 by 11/01/2020
12,000 or 11,000 by 11/01/2020
915 or 840 by 11/01/2020
12,800 or 11,300 by 11/01/2020
3,700 or 3,400 by 11/01/2020
 
Research updated 3/27/2020.
My book, Swing and Day Trading, shown on the left, discusses several swing and day trading setups.
If you click on the above link and then buy the book (or anything) while at Amazon.com, the referral will help support this site. Thanks.
$ $ $
This article discusses a swing trading setup based on price retracing a portion of a prior rise and it also explores Fibonacci retracements.
On 3/26/2020, I rewrote the test to see if Fibonacci retraces (38%, 50%, 62%) are any more likely to appear than other retrace values. They don't.
I used bull market data only from 3/6/2009 to 2/12/2020 in 489 stocks using daily price data. I excluded any retrace that started with price at or below $5 per share.
For the peaks and valleys, I used 2, 5, 10 and 20 days in my tests (that is, I ran 4 types of tests). That means, for example using a 5 day period, that I found peaks/valley that were the highest/lowest price bar from 5 days before to 5 days after the peak/valley, for 11 days total. Samples ranged from 10,075 (using 20day widths)to over 65,000 (using 2day widths).
Using the peaks and valleys, I measured the depth of the retrace then sorted the results.
I found that the three Fibonacci retrace values of 38%, 50%, and 62% were no more likely to appear than any other number from 1% to 100%.
This updated test confirms what I found and describe below.
The figure on the right shows an example of price retracing a portion of the prior rise.
After trending downward (most of which is not shown), price begins a recovery at A, 19.51. Price rises to B in a nice straightline run to 24.07. Then the stock drops, bottoming at C, or 21.75.
The move from A to B is 4.56 (24.07  19.51) and the drop from B to C measures 2.32 (24.07  21.75). The ratio between the two numbers is the retrace value: 2.32/4.56 or 51%. In other words, price has dropped about half of the AB move.
I created a test to find and measure these riseretrace patterns and discovered how often they reached an arbitrary chosen target and exceeded that target. This article discusses those results and the method behind the test.
In the discussion that follows, I use these definitions.
I found every minor high that was higher than any peak within 5 days (before and after), and the lowest valley that was lower than any other low within 5 days (before and after). The 5 day number resulted in significant turning points as the above chart shows. Each asterisk (*) represents a 5 day turning point. Once I found the turning points, then it was just a matter of connecting the dots.
I measured the rise from valley to peak and the following retrace from peak to valley and various other points.
In the tests that follow, The buy occurs at C (shown in the first chart), which is the lowest low in the retrace. You can think of this as the perfect entry. The exit occurs when price reaches the target or drops at least 5% or more below the buy price. If price has not gapped lower, a stop loss order is assumed to close out the trade.
I used 23,921 samples from 576 stocks with data starting on January 1, 1995 to October 20, 2009. Not all stocks covered the entire period.
In an earlier test using inverted and ascending scallop chart patterns and manually found patterns, a Fibonacci retrace of 50% and 61% became visible. With this automated test, those peaks disappeared. In fact, only the "catchall" end points showed any variation above 1% of the totals. The frequency distribution of the counts resembled a bellshaped curve.
In other words, there were no spikes at 38%, 50% and 62% as one would expect. The results suggest that using Fibonacci retracements offers no advantage than using any other number as a turning point.
Since the 24,000 sample database did not show Fibonacci retrace peaks of 38%, 50% or 62%, I decided to filter the results by an upward trend leading to the start of the rise. I used the slope of a linear regression line of the prior 20 closing prices leading to the start of the rise. That is approximately a month's worth of data. If the slope was upward, then I included the sample.
Here's what I found
In other words, the results are somewhat inferior to the full sample set.
I did find that the most winning trades occurred when the rise is between 21% and 38% and the retrace between 45% and 75%.
The lessons from this study are manifold and they are listed below.
 Thomas Bulkowski
See Also

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.
Good news is just life's way of keeping you off balance.