The biggest danger of strategies generated using any machine learning process is overfitting (or curve-fitting) strategy to the historical data …
The general flow of work when generating new strategies can be described as a series of the following steps:
1. Import or download data for backtesting
2. Configure Builder options
3. Run build
4. Evaluate generated strategies
5. Retest or optimize, perform more checks
StrategyQuant X allows you to choose from 3 different strategy “styles”. By style we mean how the strategy is constructed.
Every trading strategy consists of set of IF – THEN rules, managing IF something happens THEN do some action. There are however some differences in how exactly these rules are constructed.
There are two build modes to choose from:
StrategyQuant first generates initial population of random candidates (using the Random Generation mode) and then uses genetic evolution process to evolve the population and produce better and better candidates with each generation.
The process ends when predefined number of generations is reached or when there’s no further improvement.