Optimal Step-wise Parameter Optimization of a FOREX Trading Strategy

The goal of trading simply consists in gaining profit by buying/selling a security: the difference between the entry and the exit price in a position determines the profit or loss of that trade. A trading strategy is used to identify proper conditions to trade a security. The role of optimization consists in finding the best conditions to start a trading maximizing the profit. In this general scenario, the strategy is trained on a chosen batch of data (training set) and applied on the next batch of data (trading set). Given a strategy, there are different issues to deal with, to obtain the best performances from the optimization. First of all, among all the parameters that define the strategy, it is important to identify and select the most relevant ones that become the optimization problem variables. In this way the problem complexity is reduced and the overfitting on the training set is avoided. Once the variables are chosen, the focus is on the time period used for the training and the trading sets. Accordingly, for any parameter, a proper box constraint is fixed taking into account the frequency of the given trading strategy (time scale, reactivity, etc.). Since the objective function is not defined in closed form but through an algorithm, the problem lies within the framework of black-box optimization.