Forex AI

Trading algoritmico con intelligenza artificiale, machine learning e reinforcement learning per progettare, validare e mantenere un sistema adattivo sul Forex

Luca Marinich

ENRead it in English: Forex AI

How do you go from a trading idea to a system that holds up to costs, tests, and the market? Forex AI is a complete path in Italian for designing, validating, and maintaining over time an algorithmic trading system on Forex. It starts from how the FX market really works and from real execution costs, moves through feature engineering, machine learning and reinforcement learning, and arrives at risk, validation, infrastructure, and Italian rules. Each chapter combines clear explanations, numerical examples, pseudocode, real cases, and review questions, to turn the idea of an adaptive system into a project that survives costs, tests, and production. Inside: - The FX market: OTC structure, sessions, spread, slippage, gaps and volume - Feature engineering: indicators turned into variables, correlations and currency strength - Multi-level AI stack: regime, direction, magnitude and meta-learner - Online learning and drift detection: ADWIN, Page-Hinkley, KS and PSI - Money management: fractional Kelly, volatility targeting, risk parity and Monte Carlo - Realistic cost model: spread, slippage, swap and protocols for extreme events - Scientific backtesting: walk-forward, purged cross-validation, Deflated Sharpe Ratio - Reinforcement learning: reward, PPO, curriculum learning and sim-to-real - MLOps and monitoring: data pipelines, versioning, dashboards and recovery - Explainability with SHAP and calibrated confidence - Italian compliance: ESMA, MiFID II, GDPR and taxation - Real cases: Swiss Franc 2015, LTCM 1998, Knight Capital 2012 This is not a signal manual. It is an architecture book: how to build a system, how to measure whether it works, and how to avoid fooling yourself. For developers and data scientists entering quantitative finance, experienced traders who want to understand what AI can add to their method, managers and investors evaluating automated systems, and students in finance and data science masters.