Strategy Filtering by Paul Cottrell

How to pick the best strategy?  I believe that the best way to answer this question is to utilize a complexity science framework.  This can be accomplished by letting agents provide a solution.  Agents can be programmed with simple trading rules, which are the following: (a) random selection of risk behavior at the beginning of the simulation, (b) start with equal equity in trading account, (c) agents select a particular strategy from a defined strategy landscape, (d) agents evaluate if there trading has created a profit.  During the simulation the agents store their strategies and profit performance into a data array with the parameters that the respective agents used. 

Through many simulations a global optimum should be reached. Theses simulations for strategy filtering can also implement artificial neural networks, genetic algorithms, and other agent–based methods for the agent’s trading strategy.  It is important to note that the simulation’s price action can be a stochastic function or based on historic time series, but verification of strategy should be conducted with out-of-sample testing to establish external validity—assuring that stylized facts within the time series did not affect the overall validity of a trading strategy.