This is a build-along workshop, not a lecture: attendees go from generating to validating and refining an end-to-end ML trading strategy with real ETF market data — using AI coding agents to build the feature engineering, model training, and backtesting pipeline, and AI research agents to generate, validate, and refine the trading ideas themselves.
How a single-value performance metric can distort the ranking of trading strategies — and how to avoid falling for it and gain deeper insight into strategy comparison.
Gifted review copy from Packt Publishing — honest, unpaid opinion below.
Just finished reviewing Stefan Jansen’s Machine Learning for Trading, 3rd Edition (Packt, 2026), several days after publication.
I wrote my newest paper mainly to challenge the conventional way of judging an investment or trading strategy through a single observational point of its performance metric — an approach that discards precious information about…
Gifted review copy from Packt Publishing — honest, unpaid opinion below. This is a genuinely useful reference for anyone building trading systems in Python. The recipe format is the book’s greatest strength — plenty of self-contained recipes spanning data acquisition, data wrangling, and visualization; building fundamental data screeners; factor trading; backtesting (both vector-based and event-based) and its evaluation; and using the IB API for historical and streaming data in portfolio building and management.
XAI methods enable the use of tools such as AI/ML models or backtesting in finance while gaining explainability of their outputs.
Introduction The Capital Asset Pricing Model (CAPM) represents an asset’s market sensitivity using a single scalar (beta), estimated by ordinary least squares on the full return distribution.
I am pleased to announce that my new paper is now available:
“Cross-Sectional Topological Anomaly Scores and Intraday Return Predictability: BallMapper Descriptors, Decoder-Conditional VAE, and Function-on-Function Regression on S&P 500 Intraday Data”
The chart visualizes rolling risk–return–volume dynamics for ten highly liquid S&P 500 constituents — AAPL, AMZN, AVGO, GOOG, GOOGL, META, MSFT, NVDA, PLTR, and TSLA — selected, over the April 2025–March 2026 period, by ranking all index members on the Amihud (2002) illiquidity ratio — a widely used measure of price impact per unit of trading volume — and retaining the ten most liquid stocks. Intraday OHLCV data for that period were sourced from EODHD (eodhd.com) at 5‑minute resolution and then aggregated to the 15‑, 30‑, and 60‑minute frequencies.
When evaluating trading algorithms, it is common to compare their average performance across the entire dataset. But if algorithms are used differently across market regimes, aggregated performance can produce a completely false ranking. The example below shows how an algorithm that is worse in every regime can appear superior overall.
