Kase StatWare®, a library of mathematically sound, statistically based technical indicators available on eSignal, is great for trading single markets such as Forex, energy, softs, and the like.

After becoming a self-taught market technician, in January 1990, I left the major oil company I had been with for 10 years and joined a large money-center bank to manage its commodity derivates book. Back then, energy derivatives were in their infancy, and, so, in the face of the looming Gulf crisis and war that followed, I was faced with trading using only a screen and my wits, exiting the active ebb and flow of information one gets through constant telephone contact with counterparties. In the absence of support from immediate fundamental information and the weaknesses of traditional technical indicators, the areas in which innovation was called for became obvious.

I decided to develop an approach to trading that was suitable for specific markets, such as energy, foreign exchange, metals, grains, as well as other, similar markets. Those of us who trade individual markets cannot scan a wide range of issues or instruments looking for the special patterns that some technicians base their core success on. We need to be in a particular market day in and day out.

Also, because many single-market traders trade in a corporate or institutional setting, such as Forex for a bank or oil for a refiner, you must have a high degree of accuracy and low loss results. As it happens, the Kase StatWare® methodology exhibits a high degree of accuracy and produces low loss results.

The first area of technical analysis on which I focused was “stops”. Previously, I had used a fixed-value trailing stop. This is a stop of a fixed amount, such as $1.00. When you hold a long trade, the $1.00 is deducted from the highest high made since trade entry, commencing with the bar of entry, and the stop never decreases. If the market were to turn lower, the stop would be hit.

The opposite is true for stops in declining markets. In the last five years of the ‘80s, the average daily range on crude was $0.45 with a standard deviation of $0.10. Thus, using a stop around $0.45 to $0.55 was almost always reasonable. 

With the onset of the Gulf Crisis, it was not uncommon to see days higher than $3.00. Because of this, I began to think about the need for a stop that adjusted for volatility.

I came across a couple of trading techniques that used average TrueRange values for stop and reverse systems that used fixed multipliers. For example, a stop might be set at 3*ATR, where

While TrueRange is proportional to volatility and using such an approach is an improvement over a fixed-value stop, the use of the average is insufficient to truly capture market behavior. 

Let’s say, for example, that you wish to design a doorway that will allow 95 percent of a given population to enter without ducking or hitting their heads, but no more.

Population one has an average height of 5’9” and is made up of the Dallas Cowboys cheerleaders, the shortest of whom is 5’8” tall and the tallest 5’10”. 

Population two also has an average height of 5’9” and is made up of the Dallas Cowboys team, and their elementary-school-aged children. The shortest member of population two is 3 feet tall and the tallest 6’10”. The door for population one must be smaller than that for population two, despite the average height being the same.

The same is true for a market. A stop that is going to perform optimally must consider the variability of range, not just the average. So, I developed the DevStops, where “Dev” stands for standard deviation.

The value of the stop amount is calculated as follows.

Notice the stop uses a double true range, which I found to work better than the single after extensive research. For the actual stops, I use a default of 30 for n and calculate the average (0), 1, 2.2 and 3.6 standard deviations over the mean. The reason I add 10 percent to the two standard deviation level and 20 percent to the three standard deviation level is to account for skew.  Again, these adjustments are based on extensive research. 

The Kase DevStops automatically adjust for volatility and variability of volatility and also are placed at predictable levels. For example, a stop placed at the one standard deviation level has statistically 84 percent odds of being hit -- or a chance that, if hit, the odds are 84 percent that it was statistically significant. In addition, we know, if the average is hit, what the odds will be for the remaining stops to be hit and how closes in the direction of or against stops will affect market behavior.

These stops can be used to fine-tune both exit and reversal strategies, determine support and resistance levels and predict short-term market behavior.

The stops are placed above or below the data, assuming long or short, based on a simple moving average crossover system built into the indicator. The “front end” could be programmed for any entry system. As shown in the subsequent chart, stops are placed above or below the data automatically.

For the entire run to the downside, the Eurodollar remained mostly below Dev1 until the turn was imminent; at which point, it broke Dev1 and hit Dev2. A few bars later, the stops flipped and the market again held stops on the move up.

Forex, Eurodollar, November / December 2005

When I first started trading, I used the Stochastic, RSI and / or MACD to identify market turns and to exit when such a turn was imminent. I found that, if one of these indicators generated a divergence, the market turned most of the time. However, there were a lot of market turns that took place in the absence of divergence -- our recent research shows around 55 percent.

The problem is that all the traditional momentum indicators use very simple underlying measures of trend, such as moving averages or rates of change of closes. Because of this, I developed the Kase PeakOscillator (KPO) and KaseCD (KCD), which work as well as the traditional indicators in predicting turns, approximately 80 percent, but also catch approximately 80 percent (versus 55 percent for traditional) of the turns.

So, the Kase indicators are “bi-directional”. This means they work in the direction of predicting turns and resulting in-turns, where the traditional indicators are “uni-directional”, working only in the direction of resulting in-turns.

At approximately the same time, I learned about serial dependency in time series. A random event can set off series of predictable, or serially dependent, events in turn. For example, in late April 2005, after a protracted relaxation in prices in the energy market, there was a reversal back to the upside, following forecasts calling for a series of freak snowstorms to hit the U.S. Midwest.

The storms were random, but, even in the early stages of the rally, the upside run became predictable and perpetuated for more than a week. So, in thinking about improving momentum indicators, I came up with the idea of using serial dependency. Thus, I came up with the Kase Serial Dependency Index, or KSDI. What the index does, when looking at “up” markets, is divide the natural logarithm (log to base e -- an “existential” number) of the high n days ago to the low today, by the volatility, which is also logarithmically based.

The opposite calculation is made for the “down”, as shown subsequently.  Because volatility is a one standard deviation logarithmic rate of change, we can think of it as a standard of measurement for the market. The higher the ratio of the actual market movement to this measure, the more it is exhibiting serial dependency -- or “trendiness”. If prices move approximately two standard deviations, there is less than a 3 percent chance that the “trend” will continue, and prices will then be expected to revert.

There are two indicators that are derived from the KSDI, the PeakOscillator and the KaseCD. The Kase PeakOscillator is a simple oscillator that takes the difference between two moving averages. But, in the case of the PeakOscillator, the indicator takes the difference between the up and down indices. The actual index value is based on an “n”, where the number of periods in question, over a range of lookback lengths, gives the highest returned value. 

The subsequent chart shows approximately nine months of daily live cattle data to May 2005. Points “1”, “2”, “4” and “5” show “Peak Outs”. These are discrete pinpoint overbought or oversold signals that, unlike such signals with traditional indicators, work very well to catch market turns.

Traditional indicator overbought and oversold signals often perpetuate for some time in trending markets, making it difficult to choose a point at which to identify the exact turn. The PeakOscillator’s lookback length optimization focuses the signal usually to one bar, making the use of overbought or oversold indications practical and highly accurate.

For example, it has better than 75 percent accuracy in catching turns of approximately one standard deviation against the prevailing trend. In these cases, the Stochastic was technically oversold or oversold, but, if you had waited for either a K-versus-D crossover or a break of the 20 percent threshold line, the signal would have come too late to be as useful. In addition, the Stochastic generated numerous false signals, where the KPO did not.

Live Cattle Futures, October 2004 to May 2005

Points “3” and “6” generated classic bearish divergence signals on the KPO, where the Stochastic was non-divergent. The default of “14” was used for the Stochastic in this case, and it could be argued that constantly adjusting the periodicity of this indicator could compensate for some of its weaknesses. However, the KPO does not require such manipulation due to its internal self-optimization.

The KaseCD is to the PeakOscillator as the MACD histogram is to a moving average oscillator. Each is an oscillator of an oscillator, with the KCD = PeakOscillator -- average (PeakOscillator, n). As illustrated in the subsequent chart, the KaseCD has similar advantages to the PeakOscillator, in that it often catches turns that traditional indicators miss and develops much clearer, rounder and less choppy formations than the MACD. Also, unlike the MACD, it generates discrete overbought and oversold signals (like the PeakOscillator) called KCDpeaks. 

The Soybean futures chart shown subsequently, for the date range March through September 2005, compares the MACD histogram with the KCD histogram. For the first turn “1”, the KCD was divergent and the MACD failed, generating only a degrading pattern instead of clear, double peaks. Both indicators caught the downside turn “2”.  At point “3”, there was a major correction. 

The KCD’s oversold indication, called a “KCDpeak” (indicated by red “KCD” on the histogram) takes place when the histogram turns pink, which shows a potential oversold condition, in combination with the peak in the histogram itself, marked by the red +.  At this point, the KCDpeak caught the turn, as did the small KCDpeak that took place early on the chart, marked by the number “4” that indicated the small correction that followed.

Soybean Futures, March to September 2005

It is often said that entering the market is the easy part. All you have to do is choose an indicator or group of indicators and get in on relatively simple signals. While that is true, one of the keys to successful entries is using multiple time frames to trade -- any given chart should contain the actual time frame upon which the trade is focused and a higher time frame as a filter. If the longer term “trend” is up, long trades will be more successful and vice versa. 

When I first heard about the idea of using a longer-time-frame filter, it appealed to me, but I never had the patience to wait for the filter. For example, if on the open there was a signal generated after the first 15-minute bar, I didn’t want to wait another 45 minutes for an hourly bar confirmation. So, I developed the Kase Permission Stochastic and, in turn, the Kase Permission Screen. 

The Permission Stochastic updates every bar and has a variable time frame. It uses the normal Stochastic math, but instead of filtering, say, a 15-minute bar with a 60-minute bar, it filters a 15-minute bar with an "n" (four in the case of a 60-minute bar) ending with the current 15 minutes. So, a 15-minute bar generating a signal at 0915 would be filtered by an hourly bar from 0815 to 0915 and so forth every 15 minutes.

If you prefer using Fibonacci numbers, you could change “n” from four to three, and use a 45-minute bar, or five and use a 75-minute bar. Either way, any bar you used to filter would generate a higher time frame bar and, thus, a signal at the end of the smaller-time-frame bar. That way, any signal gets accelerated as much as possible. 

For those who like to interpret the fine points of indicators, the accelerated Kase Permission Stochastic may be sufficient as a higher-time-frame filter. However, I have added two additional layers of automation. The first layer is to place a rule set against the Permission Stochastic and simply display it as blue (permission to go long) and pink (permission to go short). (Details on the rules may be found in other articles I have published.) 

Forex, Japanese Yen, September to December 2005 with Kase Permission Stochastic and Screen

A further refinement is to color-code bars, which has resulted in the Kase Easy Entry System (KEES). It works like this. Blue-shaded (+) signs occur on buy bars; red or pink are sell bars. The underlying signals that generate the blue or red plus signs are simple Stochastic, MACD and RSI crossovers. (For those interested in the “fine points” of what the different colors mean, a complete description is available in the manual at www.eSignalCentral.com/university/.) All a trader has to look for are the letters “L” and “S”. A valid long bar has an “L” and a valid short bar has an “S”. In order to go long or short, we usually recommend waiting for a second, confirming signal, such as the long signal shown to the bottom left of the subsequent Yen chart.

The one exception to this rule that allows for taking first signals is the time when turns take place with a sharp downturn off a high, such as the one circled to the upper right. Notice that, for the entire move up, only one of the bars containing red + signs was a valid sell, and that one was just a first sell-off of a “non-spike” high and would have been ignored.

Forex, Japanese Yen, September to December 2005, with Kase Easy Entry System

Kase StatWare uses innovative concepts and techniques applicable in today’s hi-tech world of computing and trading. The Kase indicators are available on eSignal under Chart Options->Add-on Studies -> StatWare.