Live Chat Software by Kayako 
Automatic Optimization
Posted by Alexander Makhnitsky on 30 October 2012 06:12 PM


Once your Trading system is tested on the History and you've also defined all settings needed to set up it, then you can perform Automatic Optimization. Optimizer tool is integrated in AlgoStudio. You will see Optimization option in the top righthand corner of the Studio. When run backtesting in Optimization mode, system will take the settings one by one and run them on the History in different combinations, trying to achieve better perfomance. Run your Trading system in Backtester Mode several times with different parameters before start optimization to make sure there is no serious logic errors left. Setup Optimization Settings1. Type  optimization algorithm. There are two optimization algorithms are avaible:
2. Optimization Target  By default Profit. Optimization algorithm will try to improve a candidate solution to achive best Target. 3. CycleLimit  maximum number of cycles for optimization algoritm. Optimization will stop after reaching this Limit. 4. Use gradient (GA). If checked, then using the best solution found with the GA as initial guess, a gradientbased optimization method is implemented to quickly converge towards the optimum. 5. Use elitism (GA). If checked, best candidates from previous generation will be used in next. 6. Optimize factor  select optimize creteria for Optimization Target. 7. Chromosomes count (GA). Setsnumbers of chromosomes to test. 8. Use sharing (GA). If Use sharing is checked, the GA use collective sharing which successfully allow to find robust solutions after evaluating only a few percent of the full parameter space. 9. Crossover probability (GA). Each new generation is created from a combination of randomly generated offspring and offspring created from combining (crossing over) parent parameters. Crossover Probability determines the percentage of the new generation that is generated from the crossover process. 10. Mutation probability (GA). Sets the probability that a crossover offspring will contain some mutated parameters. 12. Local gravity factor, Global gravity factor (PSO)  used for Gravitational Search Algorithm which is improved Partical swarm algoritm. 13. Inertia (PSO) Inertia Weighted coefficient. Responsible for keeping the particle moving in the same directionit was originally heading. 14. Maximal velocity (PSO) In order to keep the particles from moving too far beyond the search space, we use a value clamping to limit the maximum velocity of each particle. Strategy variablesIn Strategy variables window select parameters to optimize. Choose really needed parameters. Choose smaller range for fast searching. Optimization LimitsHere you may specify custom condition for better search. Click on Add Item add new condition. In Right column select parameter, in Command column select sign (>=, ==, <=) and in Left column type value. Press Del to remove created condition. Results will be filtered to show you only variants that are met your criteria.  
