MetaQuotes Software Corp. -------------------------------------------------------------------------------- Genetic Algorithms By July 1, 2006, MetaQuotes Software Corp. is planning to have concluded the development of a new mechanism for trading strategies optimization based on genetic algorithms. As of this date, the corporate website will feature an updated version of the client terminal with this technology already embedded in it. Optimization of the expert parameters is selection of the best input values, at which the efficiency of the expert would be the highest. This is the most important stage before the expert is launched in the real trading. A complex and efficient strategy can become unprofitable if the inputs used are not the best ones. There are several ways to determine input values. The simplest way is a step-by-step search. This method was previously implemented in MetaTrader 4. However, such search can take too much time when there are very many variables available. In this case, optimization of only one expert can sometimes take a month or even more. Implementation of genetic algorithms allows significant reduction of the time needed, though it does not provide absolute accuracy. Genetic optimization makes it possible to significantly narrow the range of the most profitable values. It helps to see within what ranges the profitable combinations can be found. The narrowed, and the most profitable, range of values can be re-optimized, if necessary. Thus, the most profitable combination can be found. The genetic algorithm optimizer is based on the selection of the best results of the inputs and their subsequent crossover with an automatic cutoff of the least successful combinations. The next "generation" of the parameter crossover uses only the most successful combinations. Thus, the optimization results in the "survival" of the most successful parameters only, while the unsuccessful ones are cut off at the early stage. The new technology enables its users to sort huge numbers of all possible combinations and find the best results at a very high speed. Thanks to the cutoff of the inefficient parameters and exclusive focusing on the profitable ones, the time savings may reach 100 to 1000 times (if the amount of combinations exceeds 5000). In effect, most parameters which are considered unsuccessful do not simply reach the stage of testing. The new optimization system based on the genetic algorithms allows for the determination of the best options by five criteria: Balance the highest profit in the testing period; Profitability ratio of the profitable trades to loss-making trades; Absolute Drawdown minimal absolute drawdown; Relative Drawdown, % minimal relative drawdown; Expected Payoff statistical average value of profit on one trade