Our project presents an approach to optimize automated trading strategies using genetic algorithms (GA) to enhance the performance of the Freqtrade open-source crypto trading bot. The methodology involves representing strategies as Python classes with user-defined parameters, initial population generation through random sampling, fitness evaluation via backtesting profitability using Freqtrade, and genetic operators including mutation and crossover. Through iterative evolution of candidate strategies, we aim to demonstrate the effectiveness of GA approach. Developed Genetic Algorithm was compared to random initialization of a strategy and to Freqtrade's built-in ML optimizer. Additionally, our GA optimizer is evaluated on three strategies: "SampleStrategy", "Diamond", and "Strategy005" from Freqtrade GitHub examples via backtesting, a modeling technique for simulating real behavior and evaluating performance on previousely collected data with cross validation.
A guide for launching and testing the project is in LAUNCH.md