Developing a Single Indicator or Multiple Indicator Market Timing System

12.1 Introduction

This chapter sets the stage on how we could interface Genetic Algorithms (GAs) or apply on a trading system in order to improve our accuracy and profits (Bauer, 1994). So the question here comes into our mind that why we use Genetic Algorithms (GAs). Before the answer to this question, I would like to ask one simpler question i.e why we are involving the use of computers to develop our trading system? Answer to this question is that human brain has limited processing power or cognitive ability. Though we imagine and do many things which are of superior in nature, but we have our limits. Our brain has limited intelligence. Human mind can imagine and find a small number of combinations at a given time. If we imagine a certain thing, our imagination has limitation. Consider a game of chess; even best player will anticipate seven to ten moves in advance. Computers which have a huge processing power and large memory, can make a large numbers of combinations among different parameters which we haven’t imagine and dreamt of. Consider a 7-bit binary string. Each position can have 0 or 1 value. Then total number of combination we can have is 27 i.e 128 total combinations. To find all 128 combinations will require a lot of imagination and thought process. But with the help of computers it will be a easy task and can be completed in a seconds time. Thus computers are helpful in solving complex problems such as stock trading indicator system.

Another advantage in using computers as a decision making is the consistency in its working, when in use. We humans are not always same as making good decision. It depends upon our mood. So the rules or parameters of the rules given by the computers are reliable and they are of great help as compared to the rule found by human intervention.

There are no less trading techniques which we can use. A lot of qualitative models are discovered by the researchers and are popular amongst the traders and stock market analysts. Genetic Algorithms (GAs) based models have an edge over other traditional models, as they are easy to use, less time consuming etc. We are going to discuss all these features in the coming section.

There is a lot of flexibility in the implementation of Genetic Algorithms (GAs) (Deb, 2002), (Rajasekaran et. Al, 2007). We can implement Genetic Algorithms (GAs) with different permutation and combinations, with different sets of operator’s probabilities and so on. Since our imagination can go to the farthest end. Genetic Algorithms (GAs) can be applied until that end also. There is no limitation in applying Genetic Algorithms (GAs) as solution provider (Goldberg, 2002). Here in the coming section of this chapter we are going to explain in detail how Genetic Algorithms (GAs) is implemented to optimize the parameters of a stock trading system. The way by which we are applying Genetic Algorithms (GAs) to the technical trading system is going to be just one of the way. There are n number of ways by which Genetic Algorithms (GAs) can be applied. Genetic Algorithms (GAs) can be applied to optimized trading strategies based on fundamental analysis or technical indicators. Since technical indicators in use are simple mathematical calculation. Genetic Algorithms (GAs) in addition can be implemented as per user requirement and first choice. After the thought process to the implementation of Genetic Algorithms (GAs) i.e string mapping and representation, fitness function calculation is done. There are almost lot of ways in which Genetic Algorithms (GAs) can be implemented. The basic difference between the Genetic Algorithms (GAs) implemented for two different problems is string representation and mapping, fitness function formula. Basic working of Genetic Algorithms (GAs) is on strings and the internal procedure is a black box technique, as it does not care what these strings are mapped to or represent.

Let us consider that we are applying Genetic Algorithms (GAs) to optimize stock market trading system. Here our fitness formula is such that it should calculate annual rate of return on the invested amount. Since we had artificially applied Genetic Algorithms (GAs) on past year of stock closing price data, which is publically available. There are two choices, one if we take average of year wise returns. But it would be more advisable that recent year rate of return is more applicable. So instead of averaging, weighted average must be done i.e recent year return must be given more weights as compare to the distant year rate of return. So in case of Genetic Algorithms (GAs), we have to make changes in the fitness function formula, not on the whole Genetic Algorithms (GAs) procedure.