Methodology For Elliott Waves Pattern Recognition

The article is focused on an analysis and pattern recognition in time series, which are fractal in nature. The proposal methodology is based on an interdisciplinary approach that combines artificial neural networks, analytic programming, Elliott wave theory and knowledge modelling. The heart of the methodology are a methods, which is able to recognize Elliott waves structures including their deformation in the charts and helps to more efficient prediction of its trend. The functionality of the proposed methodology was validated in experimental simulations, for whose implementation was designed and created an application environment. Experimental simulations have shown that the method is usable to a wider class of problems than the theory itself allows only Elliott waves.This paper introduces a methodology that allows analysis of Elliot wave’s patterns in time series for the purpose of a trend prediction. INTRODUCTION ELLIOTT WAVE PERSONALITY AND CHARACTERISTICS Elliott wave analysts hold that each individual wave has its own signature or characteristic, which typically reflects the psychology of the moment (Poser2003). Understanding those personalities is a key to the application of the Wave Principle; they are defined as follows (Frost and Prechter 2001): Five wave pattern dominant trend (see Fig. 1) • Wave 1: Wave one is rarely obvious at its inception. When the first wave of a new bull market begins, the fundamental news is almost universally negative. The previous trend is considered still strongly in force. Fundamental analyses continue to revise their earnings estimates lower; the economy probably does not look strong. Sentiment surveys are decidedly bearish, put options are in vogue, and implied volatility in the options market is high. Volume might increase a bit as prices rise, but not by enough to alert many technical analysts. • Wave 2: Wave two corrects wave one, but can never extend beyond the starting point of wave one. Typically, the news is still bad. As prices retest the prior low, bearish sentiment quickly builds, and "the crowd" haughtily reminds all that the bear market is still deeply ensconced. Still, some positive signs appear for those who are looking: volume should be lower during wave two than during wave one, prices usually do not retrace more than 61.8% (see Fibonacci relationship) of the wave one gains, and prices should fall in a three wave pattern. • Wave 3: Wave three is usually the largest and most powerful wave in a trend (although some research suggests that in commodity markets, wave five is the largest). The news is now positive and fundamental analysts start to raise earnings estimates. Prices rise quickly, corrections are short-lived and shallow. Anyone looking to "get in on a pullback" will likely miss the boat. As wave three starts, the news is probably still bearish, and most market players remain negative; but by wave three's midpoint, "the crowd" will often join the new bullish trend. Wave three often extends wave one by a ratio of 1.618:1. • Wave 4: Wave four is typically clearly corrective. Prices may meander sideways for an extended period, and wave four typically retraces less than 38.2% of wave three (see Fibonacci relationships). Volume is well below than that of wave three. This is a good place to buy a pull back if you Proceedings 27th European Conference on Modelling and Simulation ©ECMS Webjørn Rekdalsbakken, Robin T. Bye, Houxiang Zhang (Editors) ISBN: 978-0-9564944-6-7 / ISBN: 978-0-9564944-7-4 (CD) understand the potential ahead for wave 5. Still, fourth waves are often frustrating because of their lack of progress in the larger trend. • Wave 5: Wave five is the final leg in the direction of the dominant trend. The news is almost universally positive and everyone is bullish. Unfortunately, this is when many average investors finally buy in, right before the top. Volume is often lower in wave five than in wave three, and many momentum indicators start to show divergences (prices reach a new high but the indicators do not reach a new peak). At the end of a major bull market, bears may very well be ridiculed (recall how forecasts for a top in the stock market during 2000 were received). Figure 1: The basic pattern of Elliott wave Three wave pattern corrective trend (see Fig. 1) • Wave A: Corrections are typically harder to identify than impulse moves. In wave A of a bear market, the fundamental news is usually still positive. Most analysts see the drop as a correction in a still-active bull market. Some technical indicators that accompany wave A include increased volume, rising implied volatility in the options markets and possibly a turn higher in open interest in related futures markets. • Wave B: Prices reverse higher, which many see as a resumption of the now long-gone bull market. Those familiar with classical technical analysis may see the peak as the right shoulder of a head and shoulders reversal pattern. The volume during wave B should be lower than in wave A. By this point, fundamentals are probably no longer improving, but they most likely have not yet turned negative. • Wave C: Prices move impulsively lower in five waves. Volume picks up, and by the third leg of wave C, almost everyone realizes that a bear market is firmly entrenched. Wave C is typically at least as large as wave A and often extends to 1.618 times wave A or beyond (Frost and Prechter 2001). FIBONACCI ANALYSIS AND ELLIOTT WAVE THEORY Fibonacci numbers provide the mathematical foundation for the Elliott Wave Theory. While the Fibonacci ratios have been adapted to various technical indicators, their utmost use in technical analysis remains the measurement of correction waves (Frost and Prechter 2001). The Fibonacci number sequence 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89,...is made by simply starting at 1 and adding the previous number to arrive at the new number: 0+1=1, 1+1=2, 2+1=3, 3+2=5, 5+3=8, 8+5=13, 13+8=21, 21+13=34, 34+21=55, 55+34=89,... This series has very numerous interesting properties: • The ratio of any number to the next number in the series approaches 0.618 or 61.8% (the golden ratio) after the first 4 numbers. For example: 34/55 = 0.618 • The ratio of any number to the number that is found two places to the right approaches 0.382 or 38.2%. For example: 34/89 = 0.382 • The ratio of any number to the number that is found three places to the right approaches 0.236 or 23.6%. For example: 21/89 = 0.236 These relationships between every number in the series are the foundation of the common ratios used to determine price retracements and price extensions during a trend (see Fig. 2). Figure 2: Fibonacci price retracements and price extensions (adapted fromhttp://www.markets.com/education/technicalanalysis/fibonacci-elliot-wave.html) A retracement is a move in price that "retraces" a portion of the previous move. Usually a stock will retrace at one of 3 common Fibonacci levels38.2%, 50%, and 61.8%. Fibonacci price retracements are determined from a prior low-to high swing to identify possible support levels as the market pulls back from a high.Retracements are also run from a prior high-to-low swing using the same ratios, looking for possible resistance levels as the market bounces from a low (Frost and Prechter 2001). Fibonacci price extensions are used by traders to determine areas where they will wish to take profits in the next leg of an up-or downtrend. Percentage extension levels are plotted as horizontal lines above/below the previous trend move. The most popular extension levels are 61.8%, 100.0%, 138.2% and 161.8%. In reality it is not always so easy to spot the correct Elliott wave pattern, nor do prices always behave exactly according to this pattern. Therefore it is advisable for a trader not to rely solely on Fibonacci ratios, but rather to use them in conjunction with other technical tools. ELLIOT WAVES DETECTION Elliott waves are characterized by wide and numerous descriptions of their distinctive phases, thus they are difficult to detect in time series. Detection according to the rules The first eventuality is the classification which gradually runs from smallest to largest parts of Elliott waves. This method is described in (Dostál and Sojka 2008). The process starts with finding a scale and separate monowaves marking. There are completed patterns according to particular ratios. These patterns are a base for other patterns. This approach is often used for manual evaluation with their subsequent processing. The method uses seven rules, which classify waves into groups depending on a ratio of heights of neighbouring waves. The rules use Fibonacci ratios with a deviation of 5%. The only possibility of searching is to check each monowave through the conditions and some experience of a researcher is expected as well. Here, the aim is not to deal with the evaluated segment, but to respect single figures as complex units. This method is accurate, but it is computationally very time consuming and it is limited to the detection of mono-waves according to the predetermined number of specific rules. Detection units and their progressive separation The second eventuality is classification of big parts of Elliott waves and their subsequent decomposition into smaller parts. Patterns of impulsive character can be detected clearly thanks to more accurate conditions than patterns of correction phase. Therefore it is possible to detect patterns proposed in input data. Here, the aim is to find a figure and then to classify its smaller units. A disadvantage is that impulse phases are only detected directly, while correction phases must be derived. Another disadvantage during detection of large parts is that their internal structure is unknown as long as other pulses are not found in these parts. Detection according to characteristic figures The third eventuality is to restrict detection to some significant figures, which are significant with respect to parts of patterns accordi