Размещено на http://www.allbest.ru/

Table of Contents

• Introduction
• The basic idea and key definitions
• An implementation of the model
• Current and expected results
• Conclusion
• References

Introduction

The object of this study is financial markets and trading strategies. The global trend to use modern means of information and communication technologies affects financial markets. This trend in the application to an exchange leads to a significant simplification of performing trading operations. The usage of the Internet reduces time costs of submitting bids and asks and executing clearing operations; as a result, there are more private traders in each exchange than a decade ago. These qualitative and quantitative changes determine a current growth of a percentage of short-term speculative traders in comparison with long-term investors. Speculative traders use trading strategies to select proper moments to open and to close their positions in these or those financial instruments. The most part of trading strategies is based on indicators calculated with the usage of volume and price values of the preceding time periods, so the most part of deals on financial markets are determined not by a value of a real economic object from which a financial instrument derives, but also by an abstract indicator. The usage of trading strategies and trading algorithms changes the behavior of financial markets.

Thus, the subject of this study is the behavior of traders on financial markets. Here, the behavior means a set of rules used by traders to determine their trading policies.

The primary aim of this study is to analyze the influence of different groups of trading strategies in financial markets with the help of the example of stock exchange. The secondary aim of this study is to discover the patterns of market behavior formed by combinations of particular trading strategies. The chief set of tasks of the study is connected with designing and programming a computer model of stock exchange. This set of tasks includes the choice of a platform, frameworks and technologies, development of requirements and limitations. Another set of tasks is connected with the usage of the developed model to analyze the market behavior. This set of tasks includes the task to perform a series of simulations with the model, to sort strategies into different groups according to the results of simulation and to formulate rules of detecting patterns in real stock exchange with the examples of real stock quotations.

The problem of market behavior has been analyzed many times and from different angles; however, this problem has not been analyzed from the particular angle of influence of trading strategies. Meanwhile, the volatility of markets is increasing; as a result, the risks of downfall are becoming greater. Therefore, it is important to understand the market behavior. The results of this study will not only disclose the problem of market collapses. Application of discovered patterns to the analyzes of real markets will also result in an increase of chances to predict future market behavior and future market collapses.

The logic of the study is the following: at the first stage a computer model of the stock exchange will be developed; at the second stage a program realization of that model will be performed; at the third stage a series of simulations will be conducted; then, the results will be processed; finally, the basic rules of how to apply patterns to real stock quotations will be formulated and all necessary conclusions will be made.

At the end of this work all the ideas expressed above will be summarized and some anticipated results will be presented.

The basic idea and key definitions

An idea of this study is based on the hypothesis that contemporary electronic stock exchange gets large groups of speculative traders involved in the trading process. The stock exchange is defined as an organized marketplace, where stocks, bonds and common stocks equivalents are traded by brokers and traders. Electronic stock exchange involves the usage of electronic terminals by traders to place orders on an exchange through their brokers. Speculative traders or speculators are market members who purchase or sell volatile stocks or other instruments and hold them for a short time in order to make a profit. Speculators perform deals on the market more frequently in comparison with investors. The investor in a wide meaning is defined as a market participant, who puts money for interest. In this study this term will mean only those investors who invest money in companies by means of buying their stocks. The shorter the periods between two deals on the market are, the greater fluctuations of prices and frequency of fluctuations will be. In this study only common stocks will be studied as a financial instrument; thus, the model will only simulate the operating of a stock exchange. The basic assumptions of this study are formulated as follows: in those cases when trading strategies of notable part of traders are similar, the market is stable and moves as if there is only one trader with consolidated sum of money; meanwhile, when trading strategies are opposing, their usage may cause a sharp drop in a financial market. The main goal of this study is to check whether these assumptions are right or wrong by means of studying an influence of various combinations of strategies on market behavior.

This paper gives a brief description of the model. To understand the principles of its design some financial terms should be given. Two basic types of orders exist: ask, or offer, and bid. Other complex types are derived from them. Ask is any offer to sell at a specific price. Analogically, bid is any offer to buy at a specific price. Term spread means the difference between the highest available bid price and the lowest available offer price.

financial market strategy trading

There are four types of orders possible for both ask and bid orders. They are market order, limited order, stop order and limited stop order. All these types will be presented in the model that is described in this draft. A market order is an order to buy or sell at the best available price or market price. Thus, if a number of shares set by a market bid exceeds a number of shares, available in the lowest ask, deficient shares will be acquired by the next ask with higher price. As a result, an average price of acquired shares will be high. Most orders executed on the exchanges are market orders. A limited order is an order to buy or sell at a specific price or at a better price. A stop order is an order to buy or sell at the market price once the instrument was traded at a specific price called the stop price. Finally, a stop-limit order is an order to buy or sell at a specified price or at a better price, called the stop-limit price, after a stop price has been reached. It is a combination of a limit order and a stop order.

Traders can form an order with special instructions and rules revealing in which cases positions should be closed automatically by their broker without their consent. Such instructions are called stop-losses and take profits. Stop-loss is defined as an order to a broker with the sell price of an instrument below the market price. Thus, a stop-loss order will save profits that have already been achieved or prevent further losses if the tickets decline. Take profit is symmetric to stop-loss order to a broker with the sell price of an instrument above the market price to close a position and take profit.

On the stock exchange there also exists a special type of a trader, who is called a market-maker. A market-maker is a dealer, who maintains a specified spread in a given instrument by means of buying or selling lots of shares at market prices, increasing liquidity. However, market-makers do not work with every stock and some quotations are traded without the influence of market-makers. Market-makers change a behavior of market significantly, yet in this study attention will not be concentrated on this influence.

Some stocks can be sold short, i. e. sold without being owned by the seller. A trader borrows stocks from his broker to sell them on the market. Later he must return these stocks. If the seller can buy these stocks later at a lower price he makes a profit; if the price rises, the trader suffers losses. When stocks were sold short it is said, that a short position was opened, When these stocks were bought, a position is closed.common trading deal, possible for every stock, is an opening and closing of long positions, that means owning financial instruments with the right to transfer ownership by means of a sale or a gift. An owner can sell these instruments or can hold them for a long period. In this study only long positions will be taken into account in spite of the fact that short sells alter the behavior of the market.

There are several basic approaches to stock trading. Long-term investors buy stocks when they are offered with the discount to fair price; i. e. the price that reflects the real value of the underlying company represented by that stock. Long-term investors sell, when market price reaches the fair price. This type of traders will be represented in the model, but only partially. To find the fair price of the stock one should obtain information about companies activities. Thus, the model should simulate the work of the company, and it is not possible in this study. The fair price will be set at the beginning and will be constant during the simulation. Another group of traders is news traders, whose strategy involves the usage of news factor in decision making. Just because it is too difficult to include such a factor in this model, this group of trading strategies will not be studied in this survey. Meanwhile, algorithmic (algo) trading strategies will be simulated in this study; for example, scalping strategies. Scalper is a trader who trades for small gains, usually opening and closing positions within a small time frame, such as one minute, five minutes, ten minutes or one hour. This type of trading is high frequency trading (HFT). Other groups of algorithmic traders will also be studied in this survey. In the most part of trading strategies indicators are involved. An indicator is a measurement used by traders to forecast the market's direction and as a signal to perform a deal. An example of an indicator is a moving average, which reflects an average longer frame market trend.

Another term, used in the core of the model is clearing. The term clearing means the set of operations to perform financial transactions, i. e. to transfer property rights by operating clients' accounts. In the model clearing will be simplified, because the complexity of property rights does not affect the studied problem, as well as banks and brokers interconnections.

Problems studied in this work are essential for the understanding of financial markets. They can be studied from different perspectives, such as financial perspective, general economic perspective, psychological view, system approach, probability and statistics perspectives. Most researches concerning these problems use general economic approach and statistical data. Those works study current situation, but they do not study the underlying factors. Many researches are provided by reserve banks and market regulators or sanctioned by them. For instance, The Reserve Bank of India is considering an idea to restrict high frequency trading and algorithmic (algo) trading. This suggestion is based strictly on statistical data. Several models, developed by different scientists, were dedicated to testing trading strategies. None significant results were presented considering the influence of algo trading on market behavior. Some studies paid attention to market and traders behavior, but their approaches used logical reasoning and statistical data. However, a combined simulation-based approach described in this draft possesses several advantages. First of all, it is possible to study market when different distributions of trading strategies are applied. Studying such an influence will favor better understanding of market itself, regardless of concrete trading strategies. Differences in market behavior caused by distributions of trading strategies can reveal patterns, applicable to real markets. Each simulation will generate data, which must be explained from general market perspective. The combination of simulation and reasoning results in detailed description of market behavior. That is the main aim of this study.

An implementation of the model

In order to show and to understand the influence of speculative traders on the quotations dynamics's a simulation model will be developed. The core of the simulation model will copy the main features of the stock exchange. Main features are types of orders, account management, order placement and clearing. Other part of the simulation model will contain trading strategies and algorithms, indicators and trading signals.

The model will be written in the programming language Java, using programming environment Eclipse Indigo. Development of the model requires a high-level easy-to-use language. Because of the simplicity of the underlying idea there are only few requirements to the language involved. It should be mentioned that the development platform must be free to use. Another requirement is that the paradigm involved should be object-oriented because an agent model is based on agents-objects. Modern object-oriented programming languages, such as C++, Java, Ruby, Perl, PHP and others - all satisfy these requirements. Java has recommended itself as a good tool for such a task. Experience with this language was also taken into consideration. That is why programming language Java was chosen to develop this simulation model.

This model is an agent model, i. e. a model, where intelligent agents are acting according to their own behavioral rules and instructions. This approach was chosen, because the stock exchange operating is defined by traders, or agents. In this case system dynamics or discrete models are inconvenient. The number of traders on the stock exchange and their individual behavior mean that differential equations for system dynamics will be non-linear and too complex in comparison with the simplicity of agent-based model. Meanwhile, an agent-based model reflects all effects studied here. That is why this type of simulation model was chosen.

An interaction with the model will be conducted by means of a simple but functional graphical user interface. All initial parameters will be installed before the simulation in special options. It will be possible to run the simulation by steps, i. e. after each iteration a user of the model will be able to study intermediate results and change model parameters or place an order in the market. This option is important because it allows the user to study an influence of a single order on non-volatile and stable market.

The model will download statistical trading data from simple TXT or CSV files. Initial parameters will be saved in configuration XML file.

Programming libraries for graphical output of information are inconvenient. That is why all generated information will be uploaded in output files. The format of the files should be common and widespread; thus, TXT and CSV file formats are suitable. Output information can be analyzed and modified with the help of appropriate software - statistical packages. They are developed strictly for such usage and can provide better analytical tools. That is why there will be no implementation of the work with graphs in the developed model.

It is important to mention the fact, that on the real stock exchange there is no true parallelism. All the orders are placed sequentially, as if they were placed one after another. That is why in the developed model agents will act sequentially. This approach simplifies the final model. To build a parallel model all agents-objects should run simultaneously. That means that all threads should interact according to event handlers; and all required interfaces, classes and methods should be used. The chosen sequential approach is much simpler. It is based on one process which in this model is called a planner. The planner is a long cycle, that invokes agents from the container one by one, sequentially. After each iteration a clearing method is called.

An iterative approach means that agents do not see the current situation on the market, but only a situation, as it was at the previous iteration. This phenomenon is also presented in the real stock exchange. But in the real market it is explained as a technical delay caused by software and hardware characteristics. That is why it is possible to simplify an implementation of the model by means of using iterative approach without significant distortion of the real market.

The process of trading is a sequence of actions performed at each iteration of the cycle. First, all values of indicators used in strategies are calculated. Agents one by one check their individual trading strategies. If the strategy sends a signal to buy or to sell, an order will be formed. The order is placed in the market. At the end of each iteration a process of clearing is initiated; all proper orders are executed in due order, and ownership of assets is transited. Thus, principles of the core of the model are simple. No implementation of information and bank security is necessary as well as logging and other operations and functions, that are necessary for real stock exchange. That is why in the realization of clearing operation only those functions are developed, that are necessary for the current research. Trading strategies are added separately. Thus, they are not included in the core of the model. New strategies can be included and tested when they are prepared.

Parameters of trading strategies for agents can be specified in two different ways. It is possible to download them from separate file, but in this case individual parameters for each agent should be stored. They should be manually written to this file. That is why time costs will be enormous. Meanwhile, there is no actual benefit to simulation from such an explicit specification of parameters. A specification of distributions of parameters is more convenient. Thus, each individual agent receives random value of parameters. Random generators with possible distributions will be included in the model. A user will be able to set in the user interface different distributions for different parameters.

Current and expected results

The core of the model is developed. Simple trading strategies are implemented, but more complex and realistic trading strategies are in the process of development. Conducted test simulations allow insisting that the model works correctly. With this important result it is possible to plan further progress of the work.

By this moment two groups of strategies are implemented. The first group is based on moving averages indicators. A moving average is an average value for the specified period, which calculation base moves with each new time step. In the first strategy in this group lower and upper limits are set for moving average. They are expressed in percentage of the price value and form a channel. According to this strategy buy order is placed when the price crosses above the lower limit. Sell order is placed when the price crosses above the upper limit. For every strategy in the model it is possible to specify rules of placement of stop-losses and take-profits.

The second strategy from this group uses several moving average indicators. It is evident that the moving average period represents the length of a previous trend. In this strategy bid order is placed under the following conditions: all moving averages were growing in the last two periods and at least one of them is crossed above be price value. Take profit is placed at the level of a previous local maximum, which is also called a line of resistance.

Finally, an intelligent investor strategy is implemented in the model. According to this strategy, an intelligent investor buys stocks, when they are offered with a sufficient discount to fair price. The stocks are sold when the price in the market reaches fair price or exceeds it by fixed percentage. The fair price is based on the real operational results of the underlying company. In this project there will be no modeling of company activities due to an another orientation of the project. But it is possible to set before a simulation real or realistic performance values of the company. Each intelligent investor will evaluate stocks to find the fair price. This function is not yet implemented in the intelligent investor strategy, but it will be implemented according to the plan. At this moment fair prices are set using random generator with uniform distribution.

These strategies were implemented first because they were easy and fast to implement. They were used to test the core of the model and they were sufficient.

By this moment some results were obtained. These results are trivial. However, the results which can be obtained by means of usage of logical reasoning and argumentation with the basic financial and economics knowledge have been verified with the help of the model. Verification and confirmation of hypothetical assumptions are important in scientific researches. For instance, an influence of the strategies of the same type or the same basic idea on the behavior of the market was tested with the help of this model. As it turned out, the domination of similar strategies in the market results in the stabilization of the market. This verification was conducted on the example of moving averages and intelligent investor strategies. It is too early to confirm the validity of the hypothesis, that similar strategies can improve market stability. More strategies should be involved to make some substantiated allegations. However, the obtained results are not in conflict with logical arguments. Finally, more complex and solid results will be obtained in further simulations.

In the conducted simulations the market was very stable, returning to some value after brief and small-amplitude oscillations. Hypothetically this phenomenon can be explained as a result of a similarity of strategies. By this moment it can be hypothesized that only opposing behavior of traders can destabilize the market. Influential fluctuations should be a result of a situation when opposing strategies are equally represented in the market. This assumption is a result of logical reasoning. For instance, deep drop of the price is possible only in the situation when in the market there are enough sellers and buyers. If the number of buyers is not enough to satisfy growing supply, no order can be executed. Prices will freeze and stay fixed for several periods of time. After several periods of time some strategies will force traders to cancel their order and price will begin new but calm movement.

Another assumption can be suggested. Even if there are equally represented opposing strategies in the market, the decline is possible only if in the orders of one side of the deal, i. e. either sellers or buyers, the price is not specified and orders are market orders. Market orders as opposed to limit orders are triggered not in favor of the traders who posted them. That is why if the number of market orders or their aggregate purchasing power is significant prices can be quickly moved in one direction. Articulated hypothesis are only theoretical and should be verified. Set of simulations will be conducted in further planned development of the project to verify these hypothesis.

An activity of investors and insiders can not be taken into account in the developed model. Similarly, news factor and activity of news traders can not be considered. Moreover, a psychological factor can intervene in the decision making process in the real stock exchange. For example, in several cases traders may sell or buy under the influence of excitement and, vice versa, under the influence of fear of losses. These factors alter the behavior of a real stock exchange. Therefore, in this study it is essential to realize that the real market can be unlike the market, created by the model. All anticipated results, obtained in the process of simulations, and all conclusion, based on these results should include some correction for the real stock exchange. Only permanent and thorough comparison with the real stock exchange can provide results, that are valid for real market, but not only for this limited model.

Using this model it will be possible to study the effects of groups of strategies by means of altering distributions of strategies between agents and parameters of these strategies. Different groups of strategies are expected to produce different visual effects on the graphs. That is why it is expected, that behavioral patterns will be found among those groups. By means of identifying the groups it will be possible to formulate sets of instructions, describing how to find these patterns in the application to real markets. This path of research is not the main, because it seems to be more important to explain the behavior of the market influenced by algo trading, than to describe how to trade with profit.

Conclusion

In conclusion, the model, presented in this paper will be able to describe hidden peculiarities of financial markets, caused by different distributions of trading strategies. These effects will be checked by means of simulations with the help of the model. The core of the model will simulate an exchange and it will be possible to test various trading strategies simply by including them in the model.

The core of the model has already been developed. Further modifications of the model and implementation of various strategies should be done. After the model and all practically interesting strategies are prepared, a set of simulation tests can be run. These tests will show the peculiarities of influence of different distributions of strategies among market participants. This knowledge is required for better understanding of market fluctuations. An approach used in this study uses idealized market situation. However, it is possible to compare a real market and this model and find out the difference between them. Finally, it will be possible to explain complex market situations with the knowledge obtained by means of simulating. Ideas involved in the development of the model are based on economics, logic, maths, basic principles of exchange and human behavior. Patterns discovered in the study would probably be applicable to real market. The half part of the work has been done, the other is being worked at.

References

1. Downs J., Goodman J. E. (1998). Dictionary of Finance and Investment Terms. Fifth Edition. New York: Barron's?.

2. Graham B. (2003). The Intelligent Investor. A Book of Practical Counsel. Revised Edition. New York: Harper.

3. Horstmann C. S., Cornell G. (2011). Core Java. Volume II - Advanced Features. Eighth Edition. Prentice Hall: Sun Microsystems Press.

4. Lynch P. (2008). One Up on Wall Street. How to Use What You Already Know to Make Money in the Market. New York: Simon and Schuster Paperbacks.

5. Teweles R. J., Bradley E. S., Teweles T. M. (1999) The Stock Market. Sixth Edition. New York: John Wiley and Sons, Inc.

6. Sera C. M., Sera C. E., CMT. Market and Investor Behavior. The 98,1% Solution.

http://financialtales.com/white-papers/market-and-investor-behavior/ (visited Feb 04, 2013).

7. Shah P., Modak S., Mumbai. RBI Sounds a Cautious Note on Algo Trading (Jun 29, 2012).

http://www.business-standard.com/india/news/rbi-soundscautiousnotealgo-trading/478830/ (visited Feb 04, 2013).

8. ZEW. Press Release. Algo-Trading Involves Risks for Stability on Financial Markets (Apr 21, 2011).

http://www.zew. de/en/press/1429/algo-trading-involves-risks-for-stability-on-financial-markets- (visited Feb 04, 2013).

9. Gencay R., Dacorogna M., Olsen R., Pictet O. Foreign Exchange Trading Models and Market Behavior (2002).

http://www.sfu. ca/~rgencay/jarticles/jedc-rtt. pdf (visited Feb 04, 2013).

Размещено на Allbest.ru