Problems of forecasting financial markets

Thanks to the rapid development of information technology, it became possible to analyze a large amount of information in a matter of seconds, build complex mathematical models, and solve multi-objective optimization problems. Scientists involved in the cyclical development of the economy began to develop theories, believing that tracking the trends of a number of economic variables would clarify and predict periods of recovery and decline. One of the objects to study was selected stock market. Many attempts have been made to build a mathematical model that would successfully solve the problem of predicting the increment of the price of shares. In particular, “technical analysis” has become widespread.

Technical analysis (technical analysis) is a combination of methods for studying market dynamics, most often through graphs, in order to predict the future direction of price movement. Today, this analytical method is one of the most popular. But can we assume those. analysis suitable for generating profits? To begin, consider the theory of pricing in the stock market.

One of the basic concepts since the 1960s. the efficient market hypothesis (EMH) hypothesis is considered, according to which, information on prices and sales volumes for the past period is publicly available. Consequently, any data that could ever be extracted from the analysis of past quotes have already been reflected in the stock price. When traders compete with each other for more successful use of this publicly available knowledge, they necessarily bring prices to levels at which the expected rates of return fully correspond to the risk. At these levels, it is impossible to say whether the purchase of shares is a good or bad transaction, i.e. the current price is objective, which means that it is not necessary to expect over-market returns. Thus, in an efficient market, the prices of assets reflect their true values, and the holding of those. analysis loses all meaning.
')
But it should be noted that today, none of the existing stock markets in the world can be called fully information efficient. Moreover, taking into account modern empirical research, we can conclude that the theory of an efficient market is rather a utopia, since unable to fully rationally explain the real processes occurring in the financial markets.



In particular, a professor at Yale University, Robert Schiller, discovered a phenomenon that he later called excessive price volatility of stock assets. The essence of the phenomenon lies in the frequent change of quotations, which cannot be rationally explained, namely, there is no possibility to interpret this phenomenon with the corresponding changes in the fundamental factors .

In the late 1980s. The first steps were taken to create a model that, unlike the concept of an efficient market, would more accurately explain the real behavior of stock markets. In 1986, Fisher Black in his publication introduces a new term - "noise trade".



“ Noise trading is trading in noise, perceived as if the noise were information. People who trade in noise will trade even when, objectively, they would have to refrain from this. Perhaps they believe that the noise on the basis of which they trade is information. Or maybe they just like to trade . ” Although F. Black does not indicate which operators should be categorized as “noise traders,” a description of such market participants can be found in the work of De Long, Schleifer, Summers and Waldman. Noise traders mistakenly believe that they have unique information about future asset prices. Sources of such information may be false signals about non-existent trends, given by indicators of those. analysis, rumors, recommendations of financial "gurus". Noise traders greatly overestimate the value of the available information and are willing to take on unreasonably high risk. Empirical studies also indicate that, first of all, individual investors should be classified as noise traders, i.e. individuals. Moreover, this group of traders incurs systematic losses from trading due to the irrationality of their actions. For Western stock markets, empirical evidence of this phenomenon can be found in the studies of Barber and Odin, and for operators of the Russian stock market - in the work of I.S. Nilov. The theory of noise trading allows us to explain the phenomenon of R. Schiller. It is the irrational actions of traders that cause excessive price volatility.

Summarizing modern research in the field of theories of pricing in the stock market, we can conclude about the ineffectiveness of using technical analysis for profit. Moreover, traders using those. Analysis tries to highlight repetitive graphic patterns (from the English. Pattern - model, sample). The desire to find different patterns of price behavior is very strong, and the ability of the human eye to highlight obvious trends is amazing. However, the selected patterns may not exist at all. The graph shows the simulated and actual data of the Dow Jones Industrial Average for 1956, taken from a study by Harry Roberts.



Graph (B) is a classic head-and-shoulder model. Chart (A) also looks like a “typical” pattern of market behavior. Which of the two graphs is based on the actual values ​​of the stock index, and which one - using the simulated data? Graph (A) is based on actual data. Plot (B) is created using values ​​given by a random number generator. The problem with identifying models where they really do not exist is the lack of necessary data. Analyzing the previous dynamics, you can always identify the schemes and trading methods that could give a profit. In other words, there is a collection of an infinite number of strategies based on those. analysis. Some of the strategies from the general population show a positive result on historical data, others are negative. But in the future, we cannot know which group of systems will allow us to consistently make a profit.

Also, one of the ways to determine the presence of patterns in the time series is to measure the serial correlation . The existence of a serial correlation in quotes may indicate a certain relationship between past and current stock returns. A positive serial correlation means that positive rates of return are usually accompanied by positive rates (the inertia property). A negative serial correlation means that positive rates of return are accompanied by negative rates (the reversion property or the "correction" property). Applying this method to stock quotes, Kendall and Roberts (Kendall and Roberts, 1959), proved that it is impossible to detect patterns.

Along with technical analysis, fundamental analysis has become quite widespread. His goal is to analyze the value of shares, based on such factors as the prospects for profit and dividends, expectations of future interest rates and the risk of the company. But, as in the case of technical analysis, if all analysts rely on publicly available information about the company's profits and its position in the industry, it is difficult to expect that the assessment of prospects obtained by any one analyst is much more accurate than the estimates of other specialists. Such market research is carried out by many well-informed and generously funded firms. Given such stiff competition, it is difficult to find data that other analysts do not yet have. Therefore, if information about a particular company is publicly available, then the rate of return that an investor can count on will be the most common.

In addition to the methods described above, they try to use neural networks, genetic algorithms, etc. to forecast the market. But an attempt to use prognostic methods in relation to financial markets turns them into self-destructing models . For example, suppose that one of the methods predicts a basic market growth trend. If the theory is widely accepted, many investors will immediately start buying shares in anticipation of rising prices. As a result, growth will be much sharper and faster than it was predicted. Or growth may not take place at all due to the fact that a large institutional participant, having discovered excessive liquidity, will begin to sell off its assets.

Self-liquidation of prognostic models arises from their use in a competitive environment, namely in an environment in which each agent tries to extract its own benefit, in a certain way influencing the system as a whole. The influence of an individual agent on the entire system is not significant (in a sufficiently developed market), but the presence of a superposition effect provokes the self-destruction of a particular model. Those. if prognostic methods underlie the trading algorithm, the strategy acquires a property of instability, and in the long term, the model self-destructs. If the strategy is parametric and prognostically neutral, then this provides a competitive advantage compared to trading systems in which a forecast is used to make a decision. But it should be borne in mind that the search for strategies that satisfy such parameters as, for example, profit / risk occurs simultaneously with the search for similar systems by other traders and large financial companies on the basis of the same historical data and practically by the same criteria. This implies the need to use systems based not only on generally accepted basic parameters, but also on indicators such as reliability, stability, vitality, heteroscedasticity, etc. Of particular interest are trading strategies based on the so-called “additional information dimensions” . They manifest themselves in other, usually related areas of activity and for various reasons are rarely used by a wide circle of people in the stock market.

The above reasoning allows to draw the following conclusions:

1. The theory of noise trading, in contrast to the concept of an efficient market, makes it possible to more accurately explain the real behavior of stock assets.
2. There is no regularity in the changes in quotations of trading instruments, i.e. the market is impossible to predict.
3. The use of prognostic methods, in particular technical analysis, leads to the inevitable ruin of a trader in the medium term.
4. For successful trading in the stock market, it is necessary to apply prognostically neutral strategies based on “additional informational measurements”.

---

References:

1. Shiller R. Irrational Exuberance. Princeton: Princeton University Press, 2000.
2. Black F. Noise // Journal of Finance. 1986. Vol. 41. R. 529-543.
3. De Long JB, Shleifer AM, Summers LH, Waldmann RJ Noise Trader Risk in Financial Markets // Journal of Political Economy. 1990. Vol. 98. P. 703-738.
4. Barber BM, Odean T. Trading. 2000. Vol. 55. No. 2. P. 773-806.
5. Barber BM, Odean T. Boys will be boys: Gender, overconfidence, and common stock investment // Quarterly Journal of Economics. 2001. Vol. 116. R. 261-292.
6. Odean T. Do investors trade too much? // American Economic Review. 1999. Vol. 89. R. 1279-1298.
7. Nilov I.S. Who loses his money when trading on the stock market? // Financial management. 2006. № 4.
8. Nilov I. S. Noise trade. Modern empirical studies // RCB. 2006. No. 24.
9. Harry Roberts. Stock Market Patterns and Financial Analysis: Methodological Suggestions // Journal of Finance. Marth 1959. P. 5-6.