Kase Dev Stop:
Thanks for posting the Dev Stop Nicolas, nice touch with the 10% and 20% volatility skew added to the double and triple std deviations calculation (p96 Cynthia Kase – Trading with the Odds).
It’s something I have been looking at (entries and Std Dev) and saw it a few days ago here when searching for “Kase’s Peak Oscillator.”
One thing though, I noticed when I applied the Dev Stop to my charts:
(1) There is no way of renaming the lines? eg Stop4/KDS4 is the “Warning Line” closest to price, isn’t it?
(2) Why doesn’t the “warning line” KDS4 cross underneath prices when the trend changes from long to short so that the “warning line” stays closest to the price (compared to the other 3 bands) and so the std deviation bands 2, 3, and 4 can then be further below the “warning line” as the price heads south? Pls see Kase chart and my chart.
Kase uses a simple double-moving average defaulted to 10 (red on my chart)- and 21-periods (blue on my chart) in the code to default the stops to either a long or short position.
Is that hard coded in the Dev Stop because I don’t see where that condition is within your code and why KD 1,2,3,4 all use “Highest[20](high-ATR)” ?
On my chart the thick black dotted KaseStop4 is now further away when the £/$ turns short whereas on the long part of the graph it was hugging close below the price i.e. as the first “warning line.”
Kase Peak Oscillator:
There aren’t a lot of ‘clues’ in her book as to how exactly it’s coded but it does look very interesting for the reasons mentioned below which are extracted from her remarkable book, Trading with the Odds:
On p77 – “Instead of using moving averages, one can determine mathematically whether a price series is serially dependent or independent. So our problem is solved by substituting a statistical measure of trend or serial dependency (in which each value in a series is either caused or affected by the preceding value), for the moving averages normally used in traditional oscillators. I call the resulting new indicator the Kase PeakOscillator. The mathematics is derived from a subset of probability theory called stochastic processes, which is separate and apart from the Stochastic indicator. Statisticians use the term stochastic to describe random models. (For the mathematical calculations behind this indicator, see below, “Stochastic Processes, Monte Carlo Simulations, and Random Walk Mathematics.“)
By substituting a statistical measure of trend for empirical measures, the indicator can be evaluated on a distribution curve and the 90th percentile can be chosen to determine overbought and over- sold regions without relying on local conditions. This allows identification of many market turns, which traditional indicators miss.”
and..
“The PeakOscillator also catches divergences often missed by other indicators because it is based on a measure of serial dependency over a look-back length that adapts automatically to the most significant cycle length. This indicator performs a more thorough analysis of the market and, therefore, generates superior results.”
and..
On p78 – “Currently, with the substitution of good math, there is an indicator that can be tested statisti- cally on an isolated, stand-alone basis. In fact, the indicator basically tests itself in that the value of the PeakOscillator in soybeans in 1985, for example, can be compared with the value of crude oil in 1995. This allows a trader to evaluate the indicator over many years of commodity history and determine a probability distribution for the PeakOscillator. (I tested my PeakOscillator over a history of 80 years.1 ‘The level which constitutes the 90th percentile of momen- tum (plus or minus) over this history provides a true measure of overbought and oversold conditions for this indicator.
In addition to plotting the PeakOscillator, a PeakOut line can be displayed. The PeakOut line is drawn at the 90th percentile (determined historically) or at a default of two standard deviations above the mean local data (about the 98th percentile of the local data), whichever is greater. Thus, by definition, if momentum peaks through the PeakOut line and then pulls back, there is a 90 per- cent chance of either a turn or a penultimate peak preceding a divergence.
A principal value of the PeakOscillator is the ability to identify non-divergent turns that all other momentum indicators miss, using the PeakOut signal, or the point at which the histogram penetrates the PeakOut line and then pulls back. I indicate valid PeakOuts automatically with a darker line in the histogram (shown in Figure 5-l). The computer determines when PeakOuts occur.
The PeakOscillator also catches divergences often missed by other indicators because it is based on a measure of serial dependency over a look-back length that adapts automatically to the most significant cycle length. This indicator performs a more thorough analysis of the market and, therefore, generates superior results.”
Kase then goes on to talk about the Random Walk Index on p87
and..
On p90 – “Thus, a minimum of 30 periods is employed when evaluating the range since the triple-smoothing uses computer power for little improvement.
The RWI, with our correction, forms the basis for the mathematics used in calculating the Kase PeakOscillator.” You may wish to delete part of this for copyright reasons.
Thanks Nicolas, I hope this and the Choppiness Index don’t take up too much of your time, both look superior to the standard indicators in trading s/ware packages and should help traders on this forum.
p.s. Will PRT be adding Displaced Moving Averages to the Backtest?
Best
Brad
.
.
