Optimization is the process of changing the system parameters in order tomaximize or minimize a given utility function. Generally we see situationswe want to maximize profitability and minimize risk or volatility.
The most popular optimization by the market today is therisk versus return model (Markowitz) of a portfolio of assets (stocks, forexample). The goal of this optimization is to allocate capital in a basketof assets in a proportion that seeks to maximize return and minimize risk ordispersion. The model is particularly interesting when the objective is toreduce risk, but must be used with caution if the objective is to use as astrategy for asset management, because the estimated profitability presents avery high degree of inaccuracy (the model uses the arithmetic mean of
historical returns toestimate future returns). It can be used, for example, a second process,where the pre-selected asset were chosen as another key strategy.There are other applicationsof tools in corporate finance, we can mention the optimization of the value ofcash balance. The working capital of a company has the primary function ofgenerating liquidity for routine operations and in any institution. A lowworking capital may incur a significant cost in loans and management fees, aswell as excess working capital incur in a opportunity cost, incurring on losingtheir real value over time. This value will determine precisely theoptimal balance between these two demands.
The development of tradingsystems makes extensive use of optimization methods, and sometimes a level ofcomplexity that would make the algorithmic description impracticable inthis post. One of them is the optimization of the parameters of a systemwhen using a window of historical data and performed various backtestings. Can be used in the management of multiple systems. Thetopic of optimization of trading systems will be addressed again and need a moredetail teaching methodology.
The optimization process,however should not be used indiscriminately. The result of an optimizationis only one step in a process that often needs to be restarted. It ispossible to enumerate many factors for a good optimization, and even more forthe validation of the optimization solution generated. You can, forexample, achieve a certain solution and actually be very good theoretically,but it can have a high sensitivity to other factors not taken intoconsideration as a possible variation in the spread, changes in the brokeragefee used in the simulation, etc.