Chapter 19: Method II: Trend Following 159

5/00 6/00 7/00 6/00 9/00 10/00 11/00 12/00 1/01 2/01

Figure 19.3 Method II as an alert, PerkinElmer, 200 days. Near misses on the part of our methodology should be seen as alerts.

for desirable fundamental candidates or problematic stocks is sure to improve your results.

Another approach to filtering signals is to look at the EquityTrader.com performance ratings and take buys on stocks rated 1 or 2 and sell on stocks rated 4 or 5. These are front-weighted, risk-adjusted performance ratings that can be thought of as relative strength compensated for downside volatility.

KEY POINTS TO REMEMBER

Method II buys strength and sells weakness.

Buy when %b is greater than 0.8 and MFI is greater than 80.

Sell when %b is less than 0.2 and MFI is less than 20.

Use a Parabolic stop.

May anticipate Method I.

Explore the variations.

Use Rational Analysis.

Lists of Method II candidates are available on www.Bollin-geronBollingerBands.com.

# R

METHOD III: REVERSALS

Method Til anticipates reversals by comparing tags of the bands to indicator action. Initially we look at multiple upper-band tags accompanied by deteriorating indicators and multiple lower-band tags accompanied by strengthening indicators. Then we look at tags in isolation where the indicators are in the opposite state-a lower band tag with a positive indicator or an upper band tag with a negative indicator.

Somewhere in the early 1970s the idea of shifting a moving average up and down by a fixed percentage to form an envelope around the price structure caught on. All you had to do was multiply the average by 1 plus the desired percent to get the upper band, or divide by 1 plus the desired percent to get the lower band, which was a computationally easy idea at a time when computation was either time-consuming or costly. This was the day of

Chapter 20: Method III: Reversals

columnar pads, adding machines and pencils, and, for the lucky, mechanical calculators.

Naturally market timers and stock pickers quickly took up the idea, as it gave them access to definitions of high and low they could use in their timing operations. Oscillators were very much in vogue at the time, and this led to a number of systems comparing the action of price within percentage bands with oscillator action. Perhaps the best known at the time-and still widely used today- was a system that compared the action of the Dow Jones Industrial Average within bands created by shifting its 21-day moving average up and down 4 percent with one of two oscillators based on broad market trading statistics. The first was a 21-day sum of advancing minus declining issues on the NYSE. The second, also from the NYSE, was a 21-day sum of up volume minus down volume. Tags of the upper band accompanied by negative oscillator readings from either oscillator were taken as sell signals. Buy signals were generated by tags of the lower band accompanied by positive oscillator readings from either oscillator. Coincident readings from both oscillators served to increase confidence. For stocks for which broad market data wasnt available, a volume indicator such as a 21-day version Bostians Intraday Intensity was used. This approach and a myriad of variants remain in use today as useful timing guides.

Many modifications to this approach are possible, and many have been made. My own contribution was to substitute a departure graph for the 21-day summing technique used for the oscillators. A departure graph is a graph of the difference of two averages, a short-term average and a long-term average. In this case, the averages are of daily advances minus declines and daily up volume minus down volume, and the periods to use for the averages are 21 and 100. The plot is of the short-term average minus the long-term average.

The prime benefit of using the departure technique to create the oscillators is that the use of the long-term moving average has the effect of adjusting (normalizing) for long-term biases in market structure.1 Without this adjustment, a simple advance-decline (A-D) oscillator (Figure 20.1) or up-volume-down-volume oscillator will likely fool you from time to time. However, using the difference between averages very nicely adjusts for the bullish or bearish biases that cause the problem.