Machine Learning for Trading, 3rd Edition

Built on algoseek data

Stefan Jansen’s bestselling title on ML and systematic trading runs its intraday and derivatives case studies entirely on algoseek data. The same NASDAQ-100 minute bars and S&P 500 options live in the algoseek data sandbox.

27 chapters

/

9 case studies

/

7 asset classes

/

No credit card

ML for Trading

The ML4T workflow

Where algoseek data does the work

Nine case studies across seven asset classes. algoseek data powers the intraday and derivatives stages at the center of the book.

01

Foundations

Market microstructure on minute bars

CH 1–5

02

Features

Intraday features · algoseek data

CH 6–10

03

Models

Sequence models · algoseek data

CH 11–15

04

Strategy

Options short-straddle · algoseek data

CH 16–20

05

Advanced AI

RL, RAG and agents

CH 21–24

06

Production

Deploy and operate

CH 25–27

The data

The data behind the book’s hardest chapters

Both datasets are hosted on an algoseek bucket and linked from the companion code repository.

Dataset

Format

Download

NASDAQ-100 minute bars

1-minute bars · 2020 to 2021 · up to 90 fields per bar

CSV

NDX

S&P 500 options & analytics

Straddle and IV-surface analytics · 2017 to 2021 · ~630 symbols

CSV

SPX

NASDAQ-100 trade and quotes

Level I tick data · 2020-03-13 and 2020-03-16

CSV

SPX

Live access

Run it in the sandbox

Up to a year of historical data across all asset classes. Use Jupyter notebooks, SQL, download data, or Excel.

  • Jupyter notebooks and SQL
  • Download data, or Excel
  • No credit card required

Why this data

The same data institutions run on

Stefan’s case studies use the data algoseek built for desks that cannot afford to be wrong.

A security master, not raw tickers

Battle-hardened and built in-house: every security tracked through mergers, splits, and delistings, with ASID, FIGI, and ISIN to cross-reference.

The data two US regulators rely on

The same pipeline serves two US regulators, bulge bracket banks, and startup funds. The data built for oversight runs the book’s case studies.

Up to 90 fields to build features from

Up to 90 quantitative fields per bar, the order-flow and pressure signals the book’s intraday features are built on. Standard bars carry ten to fifteen.

What you research on is what you deploy

One schema and one security master from sandbox to real-time. The chapter 16 to 20 strategy deploys in chapters 25 to 27 without re-mapping a thing.

Point-in-time, survivorship-bias-free

Adjusted for splits and corporate actions, point-in-time mapping returns the universe as it stood on your query date: no look-ahead, no survivors inflating the result.

Equity history since 2007

The minute bars draw on an equity archive back to 2007, the continuity institutional clients build on.

Author

Written and maintained by

Stefan Jansen

Stefan Jansen is the founder of Applied AI and the author of Machine Learning for Trading. For over a decade, he has built production machine-learning systems across finance, insurance, and healthcare — from contract intelligence that turns legacy insurance documents into structured logic, to forecasting at scale, to live trading infrastructure, and on into LLMs and agents. He maintains the open-source ml4t-* libraries and the book’s companion repository, which has drawn more than 19,000 GitHub stars. He holds a Harvard master’s in economics, a Georgia Tech MS in computer science, and the CFA charter.

Explore the data. Build the strategy.

The same data behind the book, in your tools. No credit card required.