Editing Technical analysis (section)

==Scientific technical analysis==

Caginalp and Balenovich in 1994<ref>{{cite journal |author1=Gunduz Caginalp |author2=Donald Balenovich |year=2003 |title=A theoretical foundation for technical analysis |journal=Journal of Technical Analysis |volume=59 |pages=5–22 |url=http://www.pitt.edu/~caginalp/TechAn90.pdf |access-date=11 May 2015 |archive-date=24 September 2015 |archive-url=https://web.archive.org/web/20150924074513/http://www.pitt.edu/~caginalp/TechAn90.pdf |url-status=dead }}</ref> used their asset-flow differential equations model to show that the major patterns of technical analysis could be generated with some basic assumptions. Some of the patterns such as a triangle continuation or reversal pattern can be generated with the assumption of two distinct groups of investors with different assessments of valuation. The major assumptions of the models are the finiteness of assets and the use of trend as well as valuation in decision making. Many of the patterns follow as mathematically logical consequences of these assumptions.

One of the problems with conventional technical analysis has been the difficulty of specifying the patterns in a manner that permits objective testing.

Japanese candlestick patterns involve patterns of a few days that are within an uptrend or downtrend. Caginalp and Laurent<ref>{{cite journal | last1 = Caginalp | first1 = G. | last2 = Laurent | first2 = H. | year = 1998 | title = The Predictive Power of Price Patterns | journal = Applied Mathematical Finance | volume = 5 | issue = 3–4 | pages = 181–206 | doi = 10.1080/135048698334637 | s2cid = 44237914 }}</ref> were the first to perform a successful large scale test of patterns. A mathematically precise set of criteria were tested by first using a definition of a short-term trend by smoothing the data and allowing for one deviation in the smoothed trend. They then considered eight major three-day candlestick reversal patterns in a non-parametric manner and defined the patterns as a set of inequalities. The results were positive with an overwhelming statistical confidence for each of the patterns using the data set of all S&P 500 stocks daily for the five-year period 1992–1996.

Among the most basic ideas of conventional technical analysis is that a trend, once established, tends to continue.  However, testing for this trend has often led researchers to conclude that stocks are a random walk.  One study, performed by Poterba and Summers,<ref>{{cite journal | last1 = Poterba | first1 = J.M. | last2 = Summers | first2 = L.H. | year = 1988 | title = Mean reversion in stock prices: Evidence and Implications | journal = Journal of Financial Economics | volume = 22 | pages = 27–59 | doi = 10.1016/0304-405x(88)90021-9 | s2cid = 18901605 }}</ref> found a small trend effect that was too small to be of trading value.  As Fisher Black
noted,<ref>{{cite journal | last1 = Black | first1 = F | year = 1986 | title = Noise | journal = Journal of Finance | volume = 41 | issue = 3 | pages = 529–43 | doi = 10.1111/j.1540-6261.1986.tb04513.x | doi-access = free }}</ref> "noise" in trading price data makes it difficult to test hypotheses.

One method for avoiding this noise was discovered in 1995 by Caginalp and Constantine<ref>{{cite journal | last1 = Caginalp | first1 = G. | last2 = Constantine | first2 = G. | year = 1995 | title = Statistical inference and modeling of momentum in stock prices | journal = Applied Mathematical Finance | volume = 2 | issue = 4 | pages = 225–242 | doi = 10.1080/13504869500000012 | s2cid = 154176805 }}</ref> who used a ratio of two essentially identical closed-end funds to eliminate any changes in valuation. A closed-end fund (unlike an open-end fund) trades independently of its net asset value and its shares cannot be redeemed, but only traded among investors as any other stock on the exchanges. In this study, the authors found that the best estimate of tomorrow's price is not yesterday's price (as the efficient-market hypothesis would indicate), nor is it the pure momentum price (namely, the same relative price change from yesterday to today continues from today to tomorrow). But rather it is almost exactly halfway between the two.

Starting from the characterization of the past time evolution of market prices in terms of price velocity and price acceleration, an attempt towards a general framework for technical analysis has been developed, with the goal of establishing a principled classification of the possible patterns characterizing the deviation or defects from the random walk market state and its time translational invariant properties.<ref>J. V. Andersen, S. Gluzman and D. Sornette, Fundamental Framework for Technical Analysis, European Physical Journal B 14, 579–601 (2000)</ref> The classification relies on two dimensionless parameters, the [[Froude number]] characterizing the relative strength of the acceleration with respect to the velocity and the time horizon forecast dimensionalized to the training period. Trend-following and contrarian patterns are found to coexist and depend on the dimensionless time horizon. Using a [[renormalisation group]] approach, the probabilistic based scenario approach exhibits statistically significant predictive power in essentially all tested market phases.

A survey of modern studies by Park and Irwin<ref>C-H Park and S.H. Irwin, "The Profitability of Technical Analysis: A Review" AgMAS Project Research Report No. 2004-04</ref> showed that most found a positive result from technical analysis.

In 2011, Caginalp and DeSantis<ref>G. Caginalp and M. DeSantis, "Nonlinearity in the dynamics of financial markets," Nonlinear Analysis: Real World Applications, 12(2), 1140–1151, 2011.</ref> have used large data sets of closed-end funds, where comparison with valuation is possible, in order to determine quantitatively whether key aspects of technical analysis such as trend and resistance have scientific validity. Using data sets of over 100,000 points they demonstrate that trend has an effect that is at least half as important as valuation. The effects of volume and volatility, which are smaller, are also evident and statistically significant. An important aspect of their work involves the nonlinear effect of trend. Positive trends that occur within approximately 3.7 standard deviations have a positive effect. For stronger uptrends, there is a negative effect on returns, suggesting that profit taking occurs as the magnitude of the uptrend increases. For downtrends the situation is similar except that the "buying on dips" does not take place until the downtrend is a 4.6 standard deviation event. These methods can be used to examine investor behavior and compare the underlying strategies among different asset classes.

In 2013, Kim Man Lui and T Chong pointed out that the past findings on technical analysis mostly reported the profitability of specific trading rules for a given set of historical data. These past studies had not taken the human trader into consideration as no real-world trader would mechanically adopt signals from any technical analysis method. Therefore, to unveil the truth of technical analysis, we should get back to understand the performance between experienced and novice traders. If the market really walks randomly, there will be no difference between these two kinds of traders. However, it is found by experiment that traders who are more knowledgeable on technical analysis significantly outperform those who are less knowledgeable.<ref>K.M. Lui and T.T.L Chong, "Do Technical Analysts Outperform Novice Traders: Experimental Evidence" Economics Bulletin. 33(4), 3080–3087, 2013.</ref>