Written and copyright © 2009-2013 by Thomas N. Bulkowski. All rights reserved.
If you are not familiar with the high and tight flag, then the link provides a good overview.
High & Tight Flag Study Summary
When I wrote the book,
Encyclopedia of Chart Patterns,
now in its second edition and pictured on the left, the high and tight flag was the best performing chart pattern. According to research for the book,
the high and tight flag has a 69% average rise in a bull market and none of the 307 patterns that I looked at failed to rise at least 5% after the breakout.
This study takes a closer look at the high and tight flag, to see what attributes result in the best performance and to update the statistics.
High & Tight Flag Study Methodology
When I wrote my Encyclopedia of Chart Patterns book, I looked at each pattern multiple times to not only catalog its features, but to make sure it was a valid pattern. This study
uses an automated approach. I consider it inferior to manually qualifying each pattern, but you can get substantially more samples. With careful programming, the results can be equally
Here is how I conducted the study.
- I used 1,018 stocks from January 1, 1995 to May 13, 2009. Some of them no longer trade or their price data has not been updated since I archived them. Of those stocks, 552 had
high and tight flags, filling a spreadsheet with 2,588 non-overlapping patterns. In other words, I only logged a new high and tight flag after the prior one completed.
- Price had to rise at least 90% (from lowest low to highest high) and do so in less than 42 price bars (2 calendar months).
- The lowest low was at least $1. Stocks priced below $1 often are headed toward bankruptcy and are in danger of being delisted from the major exchanges. I did not want the pump-and-dump
artists moving a stock from a penny to two and having that qualify as a high and tight flag. The average price at the start of the high and tight flag was $12.13, with the highest being
JDS Uniphase at $596, and a median price of $6.11.
- No stock was allowed to qualify in one day. That excluded all stocks experiencing a large one-day price spike.
- To gauge the slope of the price data leading to the start of a high and tight flag, I used the slope of a line found with linear regression on the closing prices for one month (21 price bars)
and two months (42 price bars) before the start of the high and tight flag.
- After the stock moved up by at least 90%, I allowed it to continue rising if it made a higher high. Once the highest high was found, I termed that the top of the flagpole. Then I
counted the time it took to rise at least a penny above the top of the flag pole (a breakout, in other words). The time between the flagpole and breakout was the flag.
- Price tumbling below the low price at the start of the high and tight flag was deemed a failure.
- The ultimate high price is the highest high after the breakout and before price dropped by more than 20%, measured highest high to lowest low.
- After the breakout, a stock with a closing price below the low posted in the flag meant the search for the ultimate high had completed.
- The results were not split into bull and bear markets.
All of that may sound complicated, but if you want to reproduce these results, it is best to know the complete story.
High & Tight Flag Study Quick Results
The results are manifold, but here are a few quick ones.
- It took an average of 36 calendar days to climb at least 90% (that is, the flagpole of the high and tight flag).
- The median rise was 102% with the average rise being 111% (this is the height of the flagpole)
- (1) In searching for the ultimate high, 18% of the patterns closed below the flag low.
- (2) 61% of the patterns found the ultimate high.
- 14% had price drop below the start of the high and tight flag (meaning they broke out downward).
- 7% of the patterns ran out of data (the high and tight flag occurred too recently) and were not used.
- The average rise was 27% for those in categories (1) and (2) above.
- The height of the flag averaged 26% of the breakout price (assuming a penny below the flag low as a stop price). In other words, using a penny below the flag low as a stop price meant
taking an average 26% loss.
- For those 18% of patterns that hit the flag low on their way to the ultimate high, the loss averaged 10%.
High & Tight Flag Study More Results
The reason I started this study was to determine if the slope of price trending into the start of the high and tight flag had any bearing on the outcome. I contended that steep
downtrends (like during the 2008 bear market) produced high and tight flags that failed. Here are the performance results.
|Slope||1 month slope||2 month slope|
In words, price that trended upward leading to the start of the high and tight flag, whether over one or two months,
resulted in better performance than price trending downward.
Patterns with shallow rises leading to the start of a high and tight flag resulted in better post breakout performance.
I used a slope range between -0.1 and +0.1 and found that patterns with those slope values had gains averaging 33%. Patterns outside that range had gains averaging 24%.
In other words, a flat base (or nearly so) leading to the start of the high and tight flag is a good thing.
The chart makes the slope discussion clear. The red line is the high and tight flag, where price zips upward at least 90% in two months or less.
The blue line is the inbound trend leading to the start of the hight and tight flag.
In the upper left pane, we see a shallow inbound price trend leading to the start of the high and tight flag. If the slope lasts 1 month, then the high and tight flag tends to
perform the best of the four combinations (an average rise of 35%, from the above table). If the slope lasts more than a month, performance is not as good (27% post breakout rise).
The pane in the lower left quadrant shows the most likely trend, that of price moving upward but not too steeply before increasing in slope during the high and tight flag.
The shallow uptrend gives the second best performance if the trend is one-month long (34%) and it gives the best performance if it is two months long (36%). It is the preferred
The two panes on the right side of the figure show the worst combinations, that of a steep inbound up or down trend. You should avoid trading high and tight flags
with a steep drop or steep rise leading to the start of the high and tight flag. Of the two, the steep downtrend gives the worst performance (an average rise of 22%
after the high and tight flag).
High & Tight Flag Study Failure Rates
Failure rate is another term for awful performance. I mentioned in the quick results that 14% of patterns failed to find an upward breakout. For those that did
find an upward breakout, 19% failed to climb at least 5%. That gives a combined failure rate of 33%. Clearly waiting for a breakout cuts your chances of having a failure in half.
The following table shows a frequency distribution of failures.
For example, 19% of the patterns failed to show price climbing at least 5% after the breakout. At the right end of the table, 22% of the patterns had gains over 45%, but
just 6% more than doubled in price after the high and tight flag completed. One more example: 4% of the patterns had price climb over 40% but less than or equal to 45%.
That describes the third box on the right, lower row.
High & Tight Flag
The following analysis pertains to the flag portion of the high and tight flag. The flag occurs after the flagpole and before the breakout. The flag drop referenced in the table
is how far price dropped below the breakout price and the resulting rise of patterns qualifying. The days is the width of the flag and the resulting performance.
For example, there were 105 flags that showed price drop from 0% to 5% below the breakout price. They resulted in gains averaging 17%. There were 488 patterns that had a flag
width between 0 and 5 days, and they showed post breakout gains averaging 18%.
What this table suggests is that high and tight flags with flags that retrace between 10% and 34% show the most promise. Flags that are between 10 and 29 calendar days wide
also perform well. Ignore sample counts below 30 because they may be unreliable.
-- Thomas Bulkowski
Written and copyright © 2009-2013 by Thomas N. Bulkowski. All rights reserved. Under capitalism, man exploits man. Under Communism, it's just the opposite.