Book Guide

This map connects selected Machine Learning for Trading, Third Edition companion notebooks to the library workflows that implement the corresponding backtest operations. For the library task guide, start at the User Guide. The tutorials work with bundled synthetic data, so the book and its datasets are optional. All book links below point to one checked companion revision.

Each notebook description distinguishes direct library examples from research that supplies inputs or interprets outputs. A case study that uses a book-specific helper is labeled as such; its helper and datasets are not part of the installed ml4t-backtest package. Start with the linked library workflow for a standalone example.

Chapters and workflows

Book section and notebook	Role and learning task	Library workflow
Futures backtesting	Calls DataFeed and Engine. Prepare futures bars and contract specifications.	Example data
16.3 Vectorized and event-driven backtesting	Calls DataFeed and Engine. Run one strategy, reconcile fills and trades.	First backtest
16.3 Stateful strategies	Calls DataFeed and Engine. Carry realized state into later decisions.	Risk and state
16.3 Vectorized and event-driven backtesting	Calls Engine with controlled settings. Change one execution assumption at a time.	Profiles and parity; Zipline migration
16.5 Understanding performance metrics	Calls Engine and ml4t-diagnostic. Read returns and drawdowns.	Result exports
Case-study LEAN parity	Reads retained comparison evidence. Interpret the bounded framework audit.	Profiles and parity
Portfolio performance analysis	Calls ml4t-diagnostic for downstream analysis. Analyze returns and drawdowns.	Diagnostic handoff
17.4 Defining Baseline Allocators	Teaches sizing from prediction uncertainty. Turn uncertainty into position sizes.	Accounts and constraints
17.7 Comparing Allocator Performance	Calls Engine for allocator comparison. Compare allocators with matched inputs.	Multi-asset rebalancing
Market impact scenarios	Teaches calibration from market panels. Assess size and capacity assumptions.	Market impact
18.7 Transaction Cost Analysis and Model Validation	Teaches cost analysis from return series. Reconcile gross and net performance.	Costs and funding
19.4 Drawdowns, Path Risk, and Time-to-Recovery	Uses library risk and trade types in a research comparison. Compare fixed and trailing exits.	Risk and state
19.4 Drawdowns, Path Risk, and Time-to-Recovery	Calls ml4t.backtest.risk directly. Use library position rules and portfolio limits.	Risk management

The book develops research questions, statistical interpretation, and larger datasets. The library pages specify feed contracts, order timing, account behavior, executable examples, and result schemas. Follow the library reference when a notebook and the current API differ.

Case studies

Companion example	Role and learning task	Library workflow
ETF backtest	Calls the book helper backtest_runner, which uses Engine. Weight targets from a prediction stream.	Multi-asset rebalancing
CME futures backtest	Uses a book research workflow. Futures prediction selection and equal-weight baseline.	Example data
FX pairs backtest	Uses a book research workflow. FX prediction population and strategy grid.	Data Feed
Crypto perpetual funding	Uses a book research workflow. Funding and transaction-cost assumptions.	Costs and funding
ETF risk controls	Calls the book helper backtest_runner, which uses Engine. Position exits in a full strategy.	Risk management

Move from a notebook to a reusable run

1. Start with the first backtest and the bundled price panels.
2. Reproduce the notebook's decision and fill timing with the orders tutorial.
3. Specify capital, share precision, and exposure limits with the account tutorial.
4. Add costs or risk rules only after checking the baseline fills.
5. Export the result and join it to the original timestamps with the result tutorial.

The API reference gives current signatures for each step.