A market technicians’ studies involve observing and testing for persistence of conditions in price data. Persistence can manifest itself as a strongly correlated relationship between markets or a repeated chart pattern or cycles of low and high volatility. Persistence is information and information is profit, therefore, persistence is a mantra. This article describes weaknesses in common methods used to seek out persistence and suggests an improvement.
Probably the most common form of persistence is in price change. If the quantity or quality of up days outnumber down, then we’re likely in an uptrend, or vice versa. The trend line, or its statistical equivalent, the moving average is used to identify trends. With these tools we can judge the condition of the trend and see turning points. Ideally, money flows steadily in to or out of a stock over long periods of time and trends are clear. More realistically, news, market volatility and the whims of traders create uncertain trends. In the case of the moving average, we can extend the look-back window, or period, to reduce the effect of so-called noise. There is a trade-off though: the longer the period of the moving average, the greater the signal lag. Moving averages rely on long term trends. Unfortunately, the markets do not allow long, consistent trends.
One way to find shorter trends or trade trends more profitably is to look at the movements of a stock in the context of the markets, known as the relative strength method. Many theorists (i.e. not traders) contend each stock can be linearly related to the market to provide additional information. They assume the market is this homogeneous entity, with the constituents having normally distributed returns and a high degree of correlation of price movements over time. With this simplifying assumption in place, dividing stock prices by an index value should uncover cleaner trends by differentiating between money flows in the markets and the stock. In practice, the market is a set of fundamentally diverse companies, and should not be generalized. A comparative examination of technology and utility stock charts would clarify this point.
A refinement measures strengths of stocks within the same sector. This is an improvement because stocks may be more highly correlated to their sectors than the over-all markets. However, this form of relative strength still adds volatility to low volatile stocks leading to misinterpretation of trends. For example, sectors created by industrial groupings contain small and large cap stocks. Different levels of the company’s development and leverage translate into different levels of volatility. A suggestion is to normalize price movements by volatility before assessing persistence or aggregating stocks into sectors.
Volatility is measured in different ways. Beta is a typical candidate. It relies on the homogenous markets assumption and requires too many data points to be effective. An observably better measure is the price range. Either one or two day range or average true range suffices. These values are more consistent over time and require less data to be significant. It is also an easier calculation. In volatility normalization, there is the assumption that long term volatility is related to short term volatility. There is an expectation of a volatile stock to maintain its volatility over different time periods and on different time scales. With this assumption in hand, the process can begin.
The first step is to calculate the typical historical volatility of each stock and rank them. The median two day range of the past 100 days can be used. The next step compares the recent price change of each stock to its neighboring stocks within the volatility ranking and assigns a price change score. Daily price change is recommended. If the price change of a highly volatile stock is perfectly average for the group of highly volatile stocks, then the score should be zero. If it is a strong, greater than average movement, then the score can be +1, or less than average, then -1. Scoring systems can vary but there should be a symmetrical upper and lower limit with zero mean. The final step cumulatively adds the scores over time for each stock. This creates a plot similar to a price chart. If the stock is quite average then the plot should be an almost flat line, independent of volatility.
To determine sector strength, the scores of stocks forming the sector can be averaged, accumulated and plotted. The side benefit of this calculation is it quantifies intra-sector correlation. If there is a wide dispersion of scores within a sector, the average score only achieves moderate values. If there is cohesion of scores, as in the stocks of a sector are consistently moving in the same direction, the average scores add up to a substantial value over time.
Volatility normalization may sound familiar. It is incorporated into many standard indicators and charting tools. RSI, Bollinger Bands, channels, adaptive moving averages, and Cynthia Kase’s Dev-Stop are just a few examples. The method described above is different because it adds market data to the calculation, and it does not average price change over time. Persistence in price trends is clearer helping traders and analysts make more timely decisions.
Technical Analysis Terms
RSI (Figure 1) refers to the Relative Strength Index an overbought/oversold indicator introduced by Welles Wilder in his 1978 classic New Concepts in Technical Trading Systems. It is calculated as an oscillator measuring the internal strength of the price action over a certain time period, usually 9-days or 14-days. A market is considered to be overbought when RSI rises above 70 and traders expect the market to turn lower after this level is reached. An overbought reading below 30 alerts traders to expect a reversal to the upside. RSI is calculated with the following formula:
RSI = 100 – (100/(1+U/D))
Where
- U = Average up price change over the last n days (n is usually 9 or 14)
- D = Average down price change over the last n days (n is usually 9 or 14)
FIGURE 1: AN EXAMPLE OF RSI 
Bollinger Bands, as defined in Bollinger on Bollinger Bands by John Bollinger (2002), “are bands constructed around a moving average that define in relative terms what is high and what is low. The band width is a multiple of the standard deviation of price. Bollinger Band defaults are a 20-day moving average with 2 standard deviations. An example is shown in Figure 2. Defined as formulas:
- Middle Bollinger Band = m-day simple moving average
- Upper Bollinger Band = Middle Bollinger Band + n * m-day moving standard deviation
- Lower Bollinger Band = Middle Bollinger Band – n * m-day moving standard deviation
FIGURE 2: S&P 500 INDEX WITH BOLLINGER BANDS 
Moving Averages are a basic tool of technical analysis. The moving average is found by calculating the average price using data from the most recent time period. If we want a 200 day moving average we add up the closing price (usually closing prices are used) for the last 200 days for the series we want to follow then divide by 200 and we have a starting point. On the next day, the closing price from Day 201 is added and the closing price from Day 1 is subtracted. This smoothing of price makes it easier to follow the trend. Simple buy or sell signals are generated when price crosses above the moving average (buys) or crosses below (sells). This is shown in Figure 3.
FIGURE 3: S&P 500 WITH 200-DAY MOVING AVERAGE 
Adaptive moving averages involve a more complex calculation which introduces a factor to change the amount of data used to account for recent market volatility.
The Kase DevStop was developed by Cynthia Kase (www.kaseco.com). The stop is calculated to account for volatility (which is directly proportional to risk), and also for the variance of volatility (how much risk changes from bar to bar) and volatility skew (the propensity for volatility to spike higher from time to time). Specifically, the DevStop places exit points at 1, 2 and 3 standard deviations over the mean two bar true range, corrected for skew. This makes it possible to take profit or cut losses at levels where the probability of a trade remaining profitable is low, without taking more of a loss or cutting profits any sooner than necessary.
SOURCE: BLOOMBERG L.P.
SOURCE: BLOOMBERG L.P.
SOURCE: BASELINE.
SOURCE: BLOOMBERG L.P. 
SOURCE: BLOOMBERG L.P.
SOURCE: BLOOMBERG L.P.
SOURCE: BASELINE
