Aqui está uma descrição geral do sistema completamente automático do site AmericanBulls.com. Segundo eles a % de acerto é de 72%-80%.
Alguém quer comentar a lógica que eles apresentam a seguir (já que o sistema propriamente dito é secreto claro!)?
Introduction
American Bulls is a fully automated and highly sophisticated technical analysis system developed to provide both short and long term perspective about the North American stock markets for the small investors. The system scans all the North American stock exchanges including U.S exchanges such as NYSE, AMEX, NASDAQ and OTC as well as the Canadian stock exchanges TSX and CDNX. The system scans more than 15000 stocks each day including even the penny stocks.
The generation of signals and daily scanning routine is a fully automatic process. A highly sophisticated algorithm without any human interference generates portfolio signals (such as buy, hold, etc.), inspects Japanese candlestick patterns and writes a short commentary for each and all of the stocks on a daily basis.
A distinctive feature of the system is the fact that it captures both the short-run and the long-run perspective of a common stock. The short-run perspective (possible outlook in the next few days) is captured through Japanese Candlestick patterns while the long-run perspective (outlook up to a few months) depends on a hierarchical technical analysis approach whose some details are given in the subsequent sections.
This hierarchical approach combined with candlestick pattern analysis creates substantial flexibility concerning the system signals and their interpretations. The lack of such flexibility in most other technical analysis based systems is, in our view, a serious deterrent for the small investor. The investor reads that the system has generated a buy signal for that day. However, he/she cannot devise a rational portfolio strategy if the only information input is restricted to a simple buy or sell signal. Apparently investors would like to know if buy signal means a temporary reaction upwards which will last just for a few days or if it is the beginning of a long-term up trend which may continue for the next few months. The lack of such information virtually eliminates the possibility of effective money management or intelligible portfolio adjustments.
Our major aim in creating this system was to overcome this important shortcoming, which unfortunately characterizes the majority of signal-based products in the market. American Bulls generates more than ten different type buy signals each with its specific content and as well warns the investor about the possible future risks. Hence an investor may read a commentary in American Bulls stating that the buy signal is possibly a short corrective move on an otherwise strong downtrend and any buying position inherently carries substantial risk.
The performance of the system has been more than satisfactory since its inception more than six months ago. The signal success ratio (hit ratio) of the system hovered around the range of 72% to 80%, which is high in technical analysis standards. More importantly, this success ratio is surprisingly stable and it was never below 70% during the aforementioned period. It is a remarkable achievement in our view given the extreme volatility of the markets after September 11.
In this paper; we would like to share our experience concerning the system given the proven performance of the system in demanding and rough circumstances as well as the system’s unique approach to signal generation process. The salient features of the system will be covered in section three. The next (second) chapter deals with our basic philosophical approach in creating this system. We will briefly conclude in section four.
Basic Philosophical Approach of the System
Technical analysis is the core of the system. We have not incorporated any sort of fundamental analysis into the system since it is virtually impossible to base a mainly daily system on fundamental information (such as earnings) only. We have discussed the possibility of incorporating some fundamental analysis into the basically technical framework however we have decided not to pursue this alternative for the following two reasons. First; incorporation of fundamental analysis (whose logic is somewhat different than that of technical analysis) into a framework basically characterized by technical analysis impairs the organic unity of the whole system according to our previous experience. Second; most of the information contained in fundamental analysis is already reflected in the prices and a thorough analysis of price patterns by technical analysis may already be sufficient to capture the effects of such information on the prices. Hence; the contribution of fundamental information will then be only marginal and mostly redundant.
However; the subject of technical analysis was and is a matter of intense debate. Though it is a widely used method in the professional circles and among the brokers, there is also a mostly academic and serious objection regarding the possible merits of technical analysis. There is, as well known, a voluminous literature and numerous articles showing the efficiency of financial markets. Apparently; if the markets are really efficient (even in the weak sense) and stock prices have a normal or log-normal distribution; any signal system based on technical analysis is doomed to failure from the very start. This fact compels us to clearly state our position regarding the issue since our system is based on technical analysis. In other words; we need to reveal our basic philosophical outlook, which was pivotal in designing the overall structure of the system.
We believe that the markets are stochastic in character. However a strict definition of ‘stochastic’ means the co-existence of random factors with systematic information. Technical analysis may not be able to extract any information from the random noise but it may clearly use the (mostly nonlinear) information in the series for predictive purposes. It is then quite obvious that we do not believe that the financial markets display a typical random walk process. Instead; they rather demonstrate a biased random walk process.
There is, by now, considerable evidence pointing out to the fact that the stock market prices in U.S are not distributed normally. This fact is also not a recent phenomenon. Studies back as far as 19651 support this view. The non-normality of stock price distributions is a general character of the markets rather than a specific event pertaining to a specific period. Obviously, evidence in the opposite direction (proving normality) would be a death sentence for the technical analysis. However the distributions are rather Paretian distributions with fat tails and non-negligible skewness.
We have, as well, conducted considerable research in the last 20 years ourselves regarding the behavior of financial markets in U.S some of which is published. Our studies show that the financial variables including the interest rates display short and long term cycles2. Existence of long-term cycles is particularly noteworthy since such long and big cycles means that certain up or downtrends are possible as part of these cycles (A big long cycle may be visualized as an aggregation of such trends). Moreover; cyclic structure shows that the markets have long memory while the opposite view (random walk hypothesis) allows at most Markovian-type short-term market memory. In short; markets display both trend behavior and long-term memory. Moreover, the information processing character of the markets is also cycle-dependent.
The market response to new information depends to some extent on the cyclic position of the market (if the market is in the ascending or descending phase of the cycle). The information processing is asymmetrical in nature and the response magnitudes (to what extent the prices respond to novel information) as well as response lags are variable and context-specific. This last point is particularly important since context sensitivity means the same information may have widely different effects on prices depending on the psychological context (or mood) of the market participants at the time of information arrival. A very positive piece of information about a company may hardly move the prices if the overall mood of the market is pessimistic however this information is nonetheless recorded in the collective memory of the market despite its failure to elucidate an immediate strong price move. However, if the overall mood (context) is favorable, even an otherwise marginal piece of information is sufficient to produce a strong response in prices. Apparently, all the favorable information received up to that point and not incorporated in the stock prices (but recorded in collective memory) is discharged suddenly causing a strong upward move (trend) when the overall context of the market starts changing. The marginal information only plays the role of a catalyst during the process.
These facts illustrate the fact that the markets may display occasional quantum jumps and the markets have an inherently non-linear character showing a strong resemblance to chaos theory models. It is a well known and mathematically proven fact that certain seemingly random processes may be generated by non-linear nonrandom processes. There is also additional evidence based on frequency analysis showing that the same piece of information may be evaluated differently in short and long run. This simply means that part of the information is utilized immediately (leading to short-term expectations). However, it also is inscribed in the collective long-term memory of the market contributing to the shaping of overall context (mood) of the market.
These observations should be no surprise for anyone acquainted with fractal literature. They simply show that the fractal dimension of the market is neither 2 (meaning purely normal distribution) nor 1 (purely deterministic process) but somewhere in between. There are studies indicating a fractal dimension in the vicinity of 1.3 to 1.4.
Such a fractal dimension allows enough scope for long-term memory. Moreover non-linear response patterns and Paretal distributions are compatible with such a fractal dimension. The inverse of fractal dimension (Hurst coefficient) shows the probability of an up or down move following a previous move. A normal distribution with a fractal dimension of 2 and Hurst coefficient of 0.5 simply says the objective probability of an up or down move is 0.5 (up or down moves are equally likely) in a sequence just as in a head and tails experiment. A purely normal distribution view interprets a trend purely as a chance phenomenon (a sequence of consecutive heads may occasionally be generated by a pure chance mechanism). It is however different in a world of fractal dimension less than 2. In a world characterized by a fractal dimension of 1.4, there is more than 65% chance that an up move will be followed by an up move. Hence this is a more conducive environment for trend formation. However there is still a probability (not less than 30%) that an up move may be followed by a down move. This is indicative of the fact that the trends are not deterministic but stochastic. Market has potential to generate trends but neither the starting point of the trends nor the ending points can be exactly determined in advance. A trend may randomly start forming and it may abruptly and randomly end. All we can do is to ride on the trend as quickly as possible when it randomly starts and exit the stage when there is enough evidence that the trend is over. We cannot, in advance, tell when the trend will start or end by any known method of technical analysis or fundamental analysis. The inherent structure of the market precludes such luxury. All we have is a world of randomly forming patterns (after all stock market provides us with some patterns) amidst an ocean of randomness.
Our fundamental approach is shaped on the basis of these premises. We do not try to predict anything in the future on the basis of past information since the market inherently does not have this type of determinism. We then shaped our approach by tacitly accepting the fact that we deal with a stochastic process here driven with an inevitable chance factor and randomness. However; it is a biased randomness and existence of such bias creates a scope of action for a careful analyst. For the same reason, we avoided the use of any forecasting technique despite the fact that we are quite seasoned and experienced in the application of time series techniques and a variety of different forecasting algorithms. Forecasting theoretically requires specification of nearly deterministic differential equations and stock market simply lacks this dimension.
Then the only viable alternative was to design a system that will maximize the probability of specifying the initial phases of a trend when such a trend randomly comes into existence. We do not impose anything on the market but let the market tell us if a trend is indeed starting. Certainly the same logic applies as well to the exits. System simply follow the market signals to decide if the existing trend is over. A number of technical signals are employed based on past market experience with a hierarchical order in order to maximize the probability of determining the onset or the termination of a trend. We simply deal with trends and nothing else while respecting the stochastic nature of such trends. We however also acknowledge the fact that the market has a long memory as well as a short memory (our context viewpoint) and consequently a trend at its initial phase may only be a short trend but it is also possible that this short trend have the potential of turning into a long term. Hence our signal system and the hierarchical approach address this possibility as well. The signals are constantly reevaluated in a hierarchical order to see if the currently developing short trend has a potential of turning into a long- term trend with an acceptable margin of probability.
At this point, the context specific interpretation of the markets provides a valuable clue to us. We acknowledge the fact that the markets have a hidden long-term memory that is essential in the formation of current context (mood) of the market. Our system respects this fundamental trait of the markets. The system just like the market itself has a long-term memory and it does not forget what happened in the past. The current status of the technical indicators is compared to their past performance in generating the signals. The system does not only process the oncoming information but retains this information also in order to use it in the future as a reference point.
A Synopsis of the System
The signal system is structured on the basis of the basic philosophy mentioned above. In fact; the system structure is fairly simple in terms of the technical tools utilized. It basically uses momentums and the slopes of moving averages. Appropriate filters are applied to momentums in order to prevent very short-term fluctuations. A set of different moving average lengths is used. The choice of proper moving average length depends on the current context and cycle of the market.
The unique feature of the system is its hierarchical structure. Certain levels of momentum are classified as critical. These levels are derived as a result of optimization using appropriate past data and are subject to periodic re-adjustment. System continuously checks if the current momentum values are below or above these critical levels. This phase constitutes the hierarchy level one. All the other following steps are taken according to the current status of this first hierarchy level. In this sense, hierarchy level one is the top hierarchy in the system controlling the decision process at the lower hierarchy levels. Hierarchy level one roughly corresponds to the long-term hidden memory of the market or to the overall context of the market. This stage is essential to determine if we are in a long-term trend or not.
The hierarchy level two checks the slopes of the momentums. Depending on the slope of the momentum, it decides either to buy or sell. However the type of decision (the type of buy or sell) generated in hierarchy two is conditional on the status of hierarchy one. The system also counts the number of buy signals and the maxima (or minima) of the momentums. This algorithm basically mimics the typical market behavior. As is well known; a bullish market reveals itself usually in an ascending waveform. We move to higher highs not in a linear fashion but in the form of an ascending wave with an attempt to a higher crest after a brief rest at a trough. However each attempt to a higher level also weakens somewhat the future gain potential of the stock since the stock becomes progressively overvalued at each higher high. The reverse logic applies in the case of a declining market.
The system takes this fact into account. The entry conditions (conditions for a buy) are tightened further following a previous buy. In this sense, the system remembers what it had done in the past and continuously evaluates its past decisions while dealing with the current situation. The number of previous buy signals and the status of the momentum slope correspond respectively to the hierarchy level two and the hierarchy level three. The type of decisions generated in hierarchy level three is dependent on the status of the hierarchy level two. The fourth level of the hierarchy is the slope of the moving averages. The length of moving average to be adopted depends on where we are in upper hierarchy levels. The lowest hierarchy is the status of the candlestick patterns. The system recognizes the candlestick patterns as a result of a unique quantification algorithm employed for this purpose. In fact, candlestick patterns are not part of the standard signal system. Their purpose is rather to describe the short run outlook of the stock prices.
The overall logic of the system based on hierarchy principle resembles a decision tree algorithm and in a sense it tries to mimic the way human brain works. At each stage it asks itself where it is and it may follow different paths according to the type of the answer. We believe this type of a structuring is more compatible with the fractal nature of the markets. As is well known, fractal patterns are generated by subsequent stages of branching and re-iteration. The system here similarly depends on a branching logic. Each new branching (the evaluation of a specific position at a certain level of hierarchy) depends on the status and type of the previous branching as it is in the case of fractal branching algorithms.
The weakest point of the system is the case of penny stocks with very limited volume and very abrupt price moves. System may be quite late in giving a proper signal in some of these stocks. We however retained these penny stocks in the system since system signals proved to be fairly timely in more than half of these stocks. Since the system relies heavily on momentums and moving averages, it apparently works much better when the price series are continuous (when there are no gaps in price charts).
Conclusion
The system is designed for the benefit of the small investor. The primary purpose is to minimize the effect of outside factors and emotion in the investment decision process and create a fairly reliable investment advisor. In order to create a reliable system; we have adopted a fractal outlook which we believe truly characterize the inherent logic of the market. The system is designed in a way similar to fractal branching algorithms and the way human brain thinks with tacit assumption of stochastic trends and hierarchy principle. It does an immense job of running through more than 15000 stocks each day and writing a distinct commentary for each one of these stocks. It also has the capacity to scan for specified conditions and it can easily find out a stock with a newly arrived buy signal and an accompanying particular Japanese candlestick pattern in any one of the North American markets. System has full transparency. We always report when and what type of signal was generated in the past and what then actually happened even if the outcome is embarrassing for us. The system does not hesitate to show its past mistakes and this allows the investor to make a more objective assessment about the weak and strong points of the system. We would also like to remind the fact that this product with all its strong and weak points is the result of brainpower only with very little financial support. It is, in this sense, a pure example of innovative drive if we understand by this term a product of mind with scanty financial backing.
DISCLAIMER
The opinions provided herein are intended to inform. They come with no warranty of any kind. If you should choose to interpret AmericanBulls.com information as investment advice, you do so at your own risk. Investing can be a very dangerous venture and it is you who must assume the entire cost and risk involved in all of your investment decisions, should you choose to follow this advice or use this information. AmericanBulls.com staff, members of the staff's families, and/or entities with which they are affiliated, may from time to time, buy, sell or hold stock in and have other financial dealings with the companies that appear on the AmericanBulls.com web site, mailings, or publications. The information contained on the AmericanBulls.com web site, mailings, and publications is drawn from sources believed to be factual and reliable, but in no way does AmericanBulls.com represent or guarantee the accuracy or completeness thereof, nor in providing it, does AmericanBulls.com assume any liability. This information is given as of the date appearing on the AmericanBulls.com web site report or mailing, and AmericanBulls.com assumes no obligation to update the information or advice on further developments relating to any named securities. The information found on the AmericanBulls.com web site, mailings, or publications is protected by the copyright laws of the United States and may not be copied, or reproduced in any way without the expressed, written consent of the editors.
Bonus a quem leu até aqui: lista dos sinais BUY para OTC BB e pinkies:
First the OTCBBs...
APPI .021
GTSM .022
NCVM .026
OSEE .015
POPN .018
O Clubeinvest.com informa que nenhuma da informação
aqui facultada deverá ser entendida como conselho ou recomendação
de qualquer tipo de transacção ou investimento.
