Optimizing Trading Strategies with Genetic Algorithms

Design and Evaluation of the EvoTrade System

Abstract

The EvoTrade system employs genetic algorithms (GAs) to automate the development and validation of trading strategies, addressing the inefficiencies of manual trading and limitations of existing algorithmic platforms. This paper presents EvoTrade’s modular architecture, its GA-based optimization framework, and its performance in simulated and real-world trading environments. By integrating historical market data and advanced GA techniques, including island models, fitness sharing, and adaptive mutation, EvoTrade minimizes human intervention and enhances strategy adaptability in volatile financial markets. Evaluations across versions 1.6.2 to 1.7.1 show a fitness improvement from 0.30 to 2.50, with a TradingView simulation achieving a 68.29% win rate and a net profit of 4,139.12 USD. Challenges in convergence stability, real-world discrepancies, and overfitting risks are identified, highlighting areas for future refinement. EvoTrade’s open design and dataset generation capabilities position it as a significant contribution to fintech and automated trading research, offering a scalable solution for traders and AI developers.