Once you have the trading system structured and defined its parameters, you have set the period of analysis in the historical, to choose among the alternatives in the best behaves.
First is sets the year X as a game. In this year gets to run the investment strategy, looking for the best optimization for block year, months, etc.
The second stage when you have the results of the best parameters of the strategy of year X, is extends in successive years to view their behavior, without any variation, this so truly it sees the operation. Subsequently, it will be completed with larger periods and with the initial base timeframes and classifying data that yields.
You can not take a system and optimize the entire historical, as always the best combination would not provide any future stability is sought.
If the results are positive and rising s is the time to activate the system. Later the track above, the data will be made to consolidate.
With automatic trading systems to be programmed strategy can be used platforms created for this purpose, with the possibility of making both linear optimization and genetic algorithms. Lineal or “brute force” is a sequential search, which will increase the variables step by step depending on the interval. It usually takes a long time, but is more accurate, because not leave any unanalyzed data.
Optimization of genetic algorithms, consists of random combinations, based on the best previous results as a base, discarding the rest. It is much faster and is essential if many variables are analyzed. It is used in engineering, medicine, astrophysics, etc.
Problem fall into the sobreoptimización or “curve fitting” to focus on the best combination of the past, discarding strategies that may be more stable with less performance.
In analyzing the results of the optimizations, the count 3D maps , helps to see the results in a graphical manner and avoid spikes that are optimal premises of the past, with very little chance of repeat. We must find the point that is repeated with greater benefits.[/fusion_text]