DoyStratOptimizer is a project within the DoyVestment ecosystem and is designed for the automated optimization of trading strategies based on technical analysis.
The goal of the project is to identify optimal parameter values and thresholds for rule-based trading strategies by systematically backtesting them on historical market data and evaluating their performance.
DoyStratOptimizer uses an evolutionary optimization approach to efficiently evaluate and iteratively improve large numbers of strategy variants.
Core concept:
- Multiple variants of a strategy are backtested in parallel
- Each variant receives a score based on predefined metrics
- The best strategies of a generation are selected
- New generations are created through controlled mutations of these top strategies
- A fixed number of strategies with different parameter sets are backtested simultaneously per generation
- All strategies share the same core logic but differ in their parameter values
Each strategy is evaluated using several key performance indicators, including:
- PnL (Profit and Loss)
- Maximum Profit
- Positive-to-Negative Trade Ratio
- Time in Drawdown to Time in Profit Ratio
- Maximum Drawdown
These metrics are combined into a total score, representing the overall quality of the strategy.
- After a generation completes, all strategies are ranked by their score
- The top 10 strategies are automatically transferred to the next generation
- All other strategies are discarded
- The next generation is created by mutating the parameters of the top 10 strategies
- Mutation strength is:
- determined randomly
- constrained within a range defined at program startup
Benefits of this approach:
- Fine-tuning through small mutations
- Exploration of new parameter spaces through larger mutations
- Prevents strategies from getting stuck in local optima when the global optimum lies outside narrow mutation bounds
- An optional target score can be defined; once reached, the optimization stops automatically
- Alternatively, during computation the user can:
- save the currently best candidate via a keyboard shortcut
- manually abort the evaluation
- Programming Language: C#
- Framework: .NET 10
- Current State: Console application
- Platform: Cross-platform (depending on .NET runtime support)
Planned enhancements and improvements:
- Unit & integration testing
- Stronger abstraction via interfaces
- MAUI-based GUI application
- Alternative for the console application
- Visualization of scores, generations, and metrics
- Interactive control of the optimization process
DoyStratOptimizer is intended for:
- Quantitative traders
- Algorithmic traders
- Developers of rule-based trading strategies
- Users of the DoyVestment ecosystem seeking data-driven strategy optimization
The project is under active development. Feedback, ideas, and improvement suggestions are welcome and will be considered for the ongoing roadmap.