Do You Want To Beat The Market?

My thesis in this paper is simple and clear. To consider the relative positions of high, low, open and closing prices with respect to each other is important in stock trading. Furthermore; a correct assessment of future volatility (correctly assessing if the future volatility of a stock will increase or decrease) increases the chance of identifying a profitable trade.
In the age of computers; technical analysis became accessible for all the small investors. There are a number of data vendors and technical analysis programs (e.g. Metastock) around. I observe, however, an unjustifiable importance being ascribed to closing prices in most of the common applications of technical analysis. The moving averages, MACD, RSI and ROC are the well known examples of widely utilized indicators whose formulas are essentially based on closing prices. This insistence on using closing prices may not be the correct approach in todays volatile markets. It neglects the two essential dimensions for correctly and quickly identifying a change in the market sentiment. Neither the change in the volatility of the stock nor the daily relative positions of open, high, low and close prices are adequately addressed by these indicators. As a result; these indicators give late buy and sell signals and frequently miss the early profitable phase of a bullish or bearish move.
There are also indicators whose constructions are based on high, low, opening and closing prices. The best example is the Stochastic Indicator and its various variants (e.g. Stochastic Momentum). Stochastic indicator is a better indicator in identifying the turning points of stock prices, a fact easily verifiable by a simple historical test. A better method of reading the markets the hidden messages, however, is to utilize the Japanese candlestick patterns. The whole philosophy of these patterns rests on comparisons of the changing relative positions of open, high, low and closing prices in subsequent days. The book written by Gregory L. Morris1 gives convincing proof of the superiority of Japanese candlesticks for U.S stocks and indexes over other traditional indicators on the basis of historical simulations. A more vivid demonstration of this superiority can also be seen in sites that use Candlestick Patterns to generate buy and sell signals. A site called www.americanbulls.com, for example, generates daily and weekly buy and sell signals for most of the stocks in NYSE, AMEX and NASDAQ using computer based Japanese candlestick pattern recognition. The site uses the same algorithm to analyze the stocks of 34 other countries as well. The history tables provided for each stock which tracks the performance of a trader who supposedly followed the sites buy or sell signals in the last two years, proves the success Japanese candlestick pattern recognition. For example; Fairborne Energy Trust (FEL) of Toronto Stock exchange was trading at $5.06 two years ago and currently trades at $4.20. An investor, who is assumed to be following sites signals was able to increase his/her initial investment of $100 to $808 according to the history table provided by the site for Fairborne Energy Trust.
It may, of course, be argued that the evidence put forward so far is circumstantial and lacks rigorous scientific evidence. Let me then remind the academic finance worlds recent focus on the possible use of high, low, closing and opening prices to effectively forecast the future volatility. Correct estimation of the volatility of a security is important in pricing options and measuring portfolio risks.2 The importance of volatility is also acknowledged recently in technical analysis. Adding volatility bands to RSI, Stochastic or other indicators is quite fashionable nowadays. However; the major problem here is the correct definition of the volatility.
Volatility, historically, is measured by standard deviation or variance. This definition depends on squared deviations of past prices from their mean value in a relevant observation period and generally uses closing prices in the calculation. The new volatility which is based on technical indicators mentioned above use this classical definition of volatility. The academic literature, however, is now shifting its focus to alternative definitions of volatility. Parkinson estimator3 , Garman and Klass estimator4 , Rogers and Satchell estimator5 and Yang-Zhang estimator6 are some of the new range estimators proposed. All of these four estimators use high, low, opening and closing prices (daily range) in order to obtain a more accurate estimation of the future volatility. Rogers and Satchell adds a drift term (a trend term) to their estimator in order to account for the observed drop in volatility during the trending periods. Yang-Zhang estimator is a multi-period range estimator which allows an adjustment for opening jumps.
The Monte-Carlo simulations and empirical tests with S&P 500 index show that these range estimators are able to capture the short run dynamics of volatility variation much better than the classical standard deviation7. The maximum relative difference from the true volatility is not more than 6.5%. There is a very significant improvement over classical close to close standard deviation in terms of efficiency (efficiency measures the degree of bias in an estimation). Parkinson is five times more efficient while Yang-Zhang is 7.3 times more efficient than the classical one. Though these range estimators assume that the logarithm of stock prices follow continuous Brownian motion (while the data is discrete in real life), these results still indicate that a very definite improvement is possible when daily ranges are considered. A recent study investigated the contribution of range estimators to the short and long-term forecasts of the Nifty Index.8 Several forecasting methods including asymmetric GARCH, exponentially weighted moving average, simple moving average and autoregressive methods were used to generate the daily, weekly and monthly forecasts of Nifty. Almost all of these forecasting methods showed a markedly better performance when they incorporated range estimators rather than classical standard deviation. The bias of forecasts (measured by root mean squared error) also declined with the use of range estimators for weekly and monthly periods. The forecasting performance of GARCH improves significantly by the inclusion of range estimators. The Rogers and Satchell estimator which uses the exponentially weighted moving average method, gives the best results in daily and weekly forecasts. The overall results indicate a possibility of a substantial improvement in volatility and in price forecasting just by using the readily available price range information.
Given all this circumstantial and scientific evidence; it is apparently a folly to disregard the precious information conveyed by the open, high, opening and closing prices.

References
1.Morris G.L, 2006,Candlestick Charting Explained ; Third Edition, McGraw Hill
2.Merton R.C., 1990, Continuous-Time Finance, Cambridge, Mass; Blackwell
3.Parkinson,M; 1980, The extreme value method for estimating the variance of the rate of return, Journal of Business, 53, 61-68
4.Garman,M and Klass,M;1980, On the estimation of security price volatilities from historical data, Journal of Business, 53,67-78
5.Rogers,L.C. amd Satchell S.E; 1991, Estimating variance from high, close and closing prices, Annals of Applied Probability,1, 504-512
6.Yang,D and Zhang,Q; 2000, Drift independent volatility estimation based on high, low, open and close prices; Journal of Business,73, 477-491
7.Shu,J and Zhang J.E., 2006, Testing Range Estimators of Historical Volatility, The Journal of Futures Markets, 26, 297-313