Chapter 21: AI for Finance¶

Apply ML to markets responsibly: returns and risk metrics (Sharpe, drawdown), a moving-average strategy, honest backtesting, and the biases that wreck naive results.

Metadata¶

Learning Objectives¶

• Compute returns (simple and log) and explain when to use each
• Measure risk-adjusted return with the annualized Sharpe ratio
• Quantify downside with maximum drawdown
• Build a signal from a moving-average crossover
• Backtest a strategy and produce an equity curve
• Avoid look-ahead bias by lagging signals correctly
• Account for costs and understand why they dominate high-turnover strategies
• Reason about overfitting and the dangers of multiple-testing on price data

What's Included¶

Notebooks¶

Scripts¶

• config.py — Chapter config: seeds, annualization factor, cost assumptions
• metrics.py — Returns, Sharpe ratio, maximum drawdown, volatility
• backtest.py — Moving-average crossover signal and a lagged backtester

Exercises¶

• Problem Set 1: Returns & Risk — Compute Sharpe and maximum drawdown and check their signs
• Problem Set 2: Backtesting — Run a backtest and prove the look-ahead lag matters
• Solutions — in exercises/solutions/ (notebooks and solutions.py for CI)

Diagrams (Mermaid)¶

• backtest_flow.mermaid, risk_metrics.mermaid

Read Online¶

• 21.1 Introduction — Simple vs log returns, Sharpe ratio, and maximum drawdown
• 21.2 Intermediate — Moving-average crossover signal and a look-ahead-safe backtest
• 21.3 Advanced — Look-ahead bias, overfitting, and where ML fits in markets

Or try the code in the Playground.

How to Use This Chapter¶

git clone https://github.com/luigipascal/berta-chapters.git
cd berta-chapters/chapters/chapter-21-ai-for-finance
pip install -r requirements.txt
jupyter notebook notebooks/01_returns_and_risk.ipynb

Created by Luigi Pascal Rondanini | Generated by Berta AI