Tuesday, August 27, 2002 3:22:44 PM
Using oscillators to spot tops and bottoms
Hello all. First I want to apologize for the wide format of this page - there's a big image below and the text auto-flows to the width of the image.
I've been exploring momentum oscillators to help identify intermediate peaks and troughs on the NYSE Index, thought some of you could provide feedback on my exploits.
All of you are probably familiar with the problems associated with using banded oscillators to identify overbought and oversold extremes. To beat a dead horse, oscillators don't work very well in trending markets where momentum is sustained over a long period of time - they tend to generate many invalid signals. Most technicians look for divergences or other indicators to filter out invalid signals.
I'm not a big fan of using divergences as a filter for three reasons:
1)Spotting divergences requires subjectivity and interpretation. For example, how close or far apart do peaks need to be? What exactly qualifies as a peak vs. a minor blip? What degree of slope is needed to truly qualify as a divergence? What is the exact rule that triggers the signal?
2)False divergences occur, especially in strongly trending markets.
3)Excellent trading opportunities are missed when the market fails to form a double-top or double-bottom - when no divergence occurs.
I began exploring filtering techniques based on a very basic concept - don't trade against a strong trend. Although I knew this approach would result in a few missed opportunities, my goal was to improve the overall percentage of excellent entry/exit points.
This study consisted of two indicators:
1. A Full Stochastic with %K at (60,12) and %D at (12). I didn't dork with the parameters at all - no curve fitting, data torturing, etc. Just wanted parameters that fit an intermediate term market cycle.
2. The Aroon Oscillator, set at 200 days. Again, I didn't fiddle with the parameters at all. All I wanted was a stable indicator that could identify the direction and strength of long-term trends. The Aroon Oscillator would be used as my trend strength filter.
The buy/sell rules I established for this test were pretty simple:
You buy when %K crosses above %D if the oscillator is oversold (or near oversold) AND the Aroon Oscillator isn't below -50.
You sell when %K crosses above %D if the oscillator is overbought (or near overbought) AND the Aroon Oscillator above +50.
I've included an annotated chart below. Please excuse the large chart, but I wanted to test the $NYA data set between Jan 1, 1990 through August 2002. A larger chart was needed for clarity.
Key:
Black arrows = Valid signals that resulted in excellent entry or exit points (17).
Blue arrows = Valid signals that resulted in less than ideal entry or exit points (11)
Green arrows = Invalid signals that would have resulted in an excellent entry or exit point (8).
Red Boxes = Stochastic signals filtered out by the Aroon Oscillator. I counted (39) invalid signals, the vast majority would have been less than ideal intermediate entry/exit points.
Things to note:
1.Most of the ill-timed sell signals (blue arrows) were a result of the Aroon Oscillator failing to identify strong up-trends early enough. Note the signals generated in late '90, early '95, and early '99. This leads me to consider using a faster Aroon Oscillator in subsequent studies.
2.There were four excellent sell signals (green arrows) missed at end of strong up-trends. There were also two instances where good, but invalid sell signals occurred in the midst of strong up-trends (mid '96 and early '98). In almost every case, these signals occurred after strong negative divergences. Peaks were spaced 3 to 6 months apart. This made me re-think my above stated opinion on divergences, except for the fact that I noticed seven other sell signals based on what could have been interpreted as negative divergences. These signals would have resulted in less than ideal results (Aug '91, Oct '94, July '95, Apr '96, Oct '97, Mar or Jun '99, Aug '00).
3. For the entire period, there were only two double-bottom positive divergences. Note there were also seven instances where bottoms were formed on single downward spike by the Stochastic Oscillator.
I would welcome any thoughts on this subject.
Hello all. First I want to apologize for the wide format of this page - there's a big image below and the text auto-flows to the width of the image.
I've been exploring momentum oscillators to help identify intermediate peaks and troughs on the NYSE Index, thought some of you could provide feedback on my exploits.
All of you are probably familiar with the problems associated with using banded oscillators to identify overbought and oversold extremes. To beat a dead horse, oscillators don't work very well in trending markets where momentum is sustained over a long period of time - they tend to generate many invalid signals. Most technicians look for divergences or other indicators to filter out invalid signals.
I'm not a big fan of using divergences as a filter for three reasons:
1)Spotting divergences requires subjectivity and interpretation. For example, how close or far apart do peaks need to be? What exactly qualifies as a peak vs. a minor blip? What degree of slope is needed to truly qualify as a divergence? What is the exact rule that triggers the signal?
2)False divergences occur, especially in strongly trending markets.
3)Excellent trading opportunities are missed when the market fails to form a double-top or double-bottom - when no divergence occurs.
I began exploring filtering techniques based on a very basic concept - don't trade against a strong trend. Although I knew this approach would result in a few missed opportunities, my goal was to improve the overall percentage of excellent entry/exit points.
This study consisted of two indicators:
1. A Full Stochastic with %K at (60,12) and %D at (12). I didn't dork with the parameters at all - no curve fitting, data torturing, etc. Just wanted parameters that fit an intermediate term market cycle.
2. The Aroon Oscillator, set at 200 days. Again, I didn't fiddle with the parameters at all. All I wanted was a stable indicator that could identify the direction and strength of long-term trends. The Aroon Oscillator would be used as my trend strength filter.
The buy/sell rules I established for this test were pretty simple:
You buy when %K crosses above %D if the oscillator is oversold (or near oversold) AND the Aroon Oscillator isn't below -50.
You sell when %K crosses above %D if the oscillator is overbought (or near overbought) AND the Aroon Oscillator above +50.
I've included an annotated chart below. Please excuse the large chart, but I wanted to test the $NYA data set between Jan 1, 1990 through August 2002. A larger chart was needed for clarity.
Key:
Black arrows = Valid signals that resulted in excellent entry or exit points (17).
Blue arrows = Valid signals that resulted in less than ideal entry or exit points (11)
Green arrows = Invalid signals that would have resulted in an excellent entry or exit point (8).
Red Boxes = Stochastic signals filtered out by the Aroon Oscillator. I counted (39) invalid signals, the vast majority would have been less than ideal intermediate entry/exit points.
Things to note:
1.Most of the ill-timed sell signals (blue arrows) were a result of the Aroon Oscillator failing to identify strong up-trends early enough. Note the signals generated in late '90, early '95, and early '99. This leads me to consider using a faster Aroon Oscillator in subsequent studies.
2.There were four excellent sell signals (green arrows) missed at end of strong up-trends. There were also two instances where good, but invalid sell signals occurred in the midst of strong up-trends (mid '96 and early '98). In almost every case, these signals occurred after strong negative divergences. Peaks were spaced 3 to 6 months apart. This made me re-think my above stated opinion on divergences, except for the fact that I noticed seven other sell signals based on what could have been interpreted as negative divergences. These signals would have resulted in less than ideal results (Aug '91, Oct '94, July '95, Apr '96, Oct '97, Mar or Jun '99, Aug '00).
3. For the entire period, there were only two double-bottom positive divergences. Note there were also seven instances where bottoms were formed on single downward spike by the Stochastic Oscillator.
I would welcome any thoughts on this subject.
Join the InvestorsHub Community
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
