Written and copyright © 2009-2013 by Thomas N. Bulkowski. All rights reserved.
This page is the second installment on stop placement and it discusses new research on how well stops work. See part 1 for an introduction to stop placement,
including volatility stops and what I prefer to use.
Stop Placement: Summary
When buying a stock, how close should you put the stop? If you place it 2 cents below the current low, you will be stopped out almost two-thirds
of the time. If you use 2 cents below yesterday's low, the stop rate drops to 41%. Use the day before that, and you'll be stopped out once every three trades, on average.
On a percentage basis, you will be stopped out sometime during the coming month almost half the time if the stop is closer than 4% below the breakout price. Place the stop
no closer than 8% away and it will protect you 75% of the time.
Half the time it will take 8 days before the stock returns to the breakout price. The study also reveals that using a trailing stop changes a potential loss into a profit if you
trail it properly (raise it as price climbs and never lower it).
Stop Placement: Methodology
I used stocks from three databases covering the period from July 1991 to July 2008, including the 2000 to 2002 bear market. Not all stocks covered the entire range,
and the number of stocks searched was 1,551. However, only 1,099 different stocks provided samples for the tests.
Having selected the stocks, I used several types of chart patterns on which to conduct the research: ascending triangles, double bottoms (all
combinations of Adam and Eve), rectangle tops and bottoms. These were chosen because of their known breakout points (meaning
a breakout occurs when price closes above the top of the chart pattern). It was also easier to detect and remove downward breakouts from those patterns. The tests used 3,068 samples.
I scoured the databases looking for the patterns and logged when price climbed above the highest high in the chart pattern. That constituted an upward breakout. Then I monitored how price moved
during the next month and to the end of data. The following section describes what I found, but the results apply to stops after upward breakouts and not downward ones.
Stop Placement: Results
Time to Reach the Low
During the month immediately after the breakout, the lowest low occurred an average of 11 days after the breakout, with a median of 8 days. The median means that
half of the 3,068 samples had the lowest low occur within 8 days of the breakout and half did not. The results sort like the following.
|Days After breakout||1||2||3||4||5||6||7||8||9||10|
The table shows a frequency distribution of time to reach the lowest low within the coming month (only the first 10 trading days are shown. The remainder of the
bins had values of 3% or less).
For example, the first column of numbers shows that 21% of the time, price reaches the lowest low the day after the breakout. That means price either continued rising or reversed immediately.
Another example shows that 6% of the chart patterns found the lowest low on day 3 and 32% (21% + 5% + 6%) found the low within 3 days of the breakout.
The chart shows an example of price breaking out upward (with the horizontal red line signifying the breakout price), curling back but not dropping below
the low the day after the breakout before moving higher. The day with the lowest low is marked in blue and it occurs the day after the breakout.
|Drop Below Breakout||1%||2%||3%||4%||5%||6%||7%||8%||9%||10%|
|Drop Below Breakout||11%||12%||13%||14%||15%||16%||17%||18%||19%||>20%|
During the month, price can drop below the breakout price and it does a massive 80% of the time. How far does it drop? The average is 5.3% with a median decline of 4.2%.
The above table shows a frequency distribution and it indicates that 48% drop less than 4%, 75% drop less than 8%, and 89% drop less than 12%. If you place a stop less than 4%
below the breakout price, you will be stopped out almost half the time within a month. Place it 8% away and it will be hit only once every four trades, on average.
I placed an initial stop using 2 cents below the lowest low from 0 to 10 days before the breakout and found how often price dropped to the stop within the coming month.
The following table shows the results.
|Breakout||of Time Hit||Loss|
To help explain this method, the chart shows price movement before the day of breakout. Days 2 to 4 would use the stop location as marked because it is the lowest low
among that period. Using a look back of 5 days would include point A as the new stop price because it is the lowest of the 6 days (including the breakout day). Using 1 day
before the breakout would show point B as the stop price.
This table is easy to read and it is best explained with examples. On the day of the breakout, using a stop 2 cents below the day's low, price hit the stop sometime during
the coming month 61.5% of the time. Measured from the breakout price (the highest peak in the chart pattern), the stop represented a loss of 2.19%.
Another example: placing a stop 2 cents
below lowest low from the breakout day to 3 days before the breakout, shows price hitting the stop 28.2% of the time and that represents a loss of 5.77%. One last example: From 10
trading days before the breakout through and including the breakout day, price dropped to the lowest low within that 11 day range 14.9% of the time within a month of the breakout. The
loss measured an average of 7.53%.
In other words, the table shows how likely your stop will be hit within a month after the breakout if you place it using 2 cents below the lowest valley preceding the breakout as the stop
location. If you use the breakout day's low, you will likely be stopped out almost two-thirds of the time (61.5%). Using the lowest low from the breakout to 4 days before,
you stand to remain in the trade over 75% of the time (you will be stopped out 24.7% of the time). The more days you include in the "lowest low" computation, the larger the loss.
In another test, I used a trailing volatility stop instead of a fixed stop and found that you would have been stopped out 92.1% of the time but would have made,
on average, 1.82%. Since a trailing stop follows price higher and is never lowered, it tends to reduce the loss size but when price turns, you will be stopped out if the drop is
For the final test, I used a stop look back from 0 to 10 days and followed price higher and continued following it until the end of data or until the trade was stopped out. The table
shows the results.
|Stop look back||0||1||2||3||4||5||6||7||8||9||10|
If you use 2 cents below the current day's low as the stop location and raise it each day that price rises (and never lower it), you will make an average of 2.42%. If you use the
lower of the current day's low and the prior day, you will make an average of 2.68%. At the far end of the table, using the lowest low from the current day to 10 days before,
you will make an average of 5.66% before being stopped out. And let me mention that the testing period included the 2000 to 2002 bear market.
How do the above "Trailing Stops" results compare to a volatility stop (which is a variety of trailing stop)? Without the 30-day restriction that the prior volatility stop used,
you would have made an average of 3.73%. This corresponds to
a stop look back of 5 days. In other words, you can calculate a volatility stop or you can find the lowest low (minus 2 cents) of the prior 5 days and the current day to set the stop location.
With either method, the average rise is the same. In short, a volatility stop gives the same profit as placing a stop 2 cents below the lowest low of the prior week.
-- Thomas Bulkowski
Written and copyright © 2009-2013 by Thomas N. Bulkowski. All rights reserved. Q: What did God say after He created man? A: I can do better!