Many of the indicators described in Cynthia A. Kase’s Trading with the Odds: Using the Power of Probability to Profit in the Futures Market (McGraw-Hill, 1996) are now available in the public domain for popular charting programs such as MetaStock, TradeStation, eSignal, Ninja Trader, CQG, and Aspen Graphics. The most intriguing indicator to my mind is what she calls the Kase Dev-Stop, and today I want to spend some time looking at its rationale.
Kase set out to improve on the standard volatility-based Chandelier trailing stop. This stop was normally calculated by multiplying a stock’s average true range (ATR) over a given period by some constant. For example, let’s say the 10-period ATR is 0.5 at the time of entry and the multiplier is 3. The trader is long XYZ with an entry at 20. The stock moves up to a closing high of 22 but then starts to pull back. If volatility remained constant during the course of the trade, the trader would be stopped out when the stock closed below 20.5 [22 – (0.5*3)]. If, however, volatility increased to 0.75, the trade would have more breathing room; in this case the stop would move down to 19.75.
One problem with this trailing stop method, Kase argues, is that “the level of noise is variable. This variability is not captured by an average, but by the standard deviation around the mean.” (p. 95) For the statistically challenged, Kase compares the height profiles of two different populations. The first group is made up of chorus line dancers; the second is an assortment of preschool children and basketball players. Let’s assume that we want to figure out how high a doorway would have to be so that 97.5% of each population could pass through without ducking. If the average height of the chorus line group is 5’7” and the standard deviation of the population is one inch, the door would have to be a little higher than two standard deviations above the mean—that is, a little higher than 5’9”—for 97.5% to pass through. In the second case the average height of the group is the same 5’7”, but the standard deviation is five inches. So the door would have to be 6’5” to accommodate the same 97.5%.
Another problem is that since volatility is skewed to the right (“bounded by zero on the downside and infinity on the upside”), “the distribution of range is not normal.” (p. 96) It too is skewed to the right. So instead of having “stop bands” placed at equidistant standard deviations, Kase suggests that there should be “a correction of about 10 percent on the second standard deviation and about 20 percent on the third standard deviation.” (p. 96)
To use the Dev-Stop Kase first draws a warning line, which “reflects the average two-bar reversal against the trend. The second, third, and fourth lines reflect one, two, and three standard deviation moves against the trend, corrected for skew.” (p. 97) Kase herself normally concentrates on the line that represents a three standard deviation retracement, though during highly volatile periods or if “in a profit-taking mode,” she will exit at the one standard deviation line.
The Aspen Graphics site, by the way, has a chart that displays Kase Dev Stops; it also describes possible uses for these stops. When I have some time I’ll test out Kase’s stop against the standard Chandelier stop to see whether it makes a positive difference to the bottom line. Readers can, of course, undertake this task themselves.