Complete Backtest task routes and verified book guide

[stefan-jansen]

Summary

• Add a task-level Zipline migration guide with a clean-wheel target-weight example and explicit differences in data input, callbacks, scheduling, orders, costs, and results.
• Give each principal User Guide task a direct API destination and checked book notebook. The Book Guide now labels all 18 linked notebooks by how they use the library or a related research workflow.
• Pin all 22 companion book paths to the checked public book commit 2d6e8f95eeccaee66906245606471f570b5807e5. Correct two case-study descriptions that overstated what the notebooks show.
• Remove homepage claims that all execution details are configurable or that event-driven operation alone prevents look-ahead.

Capability inventory

Principal task	Task route	API reference	Wheel example	Book example
Prepared bars, signals, timestamps	Data Feed	DataFeed	five asset-panel smoke examples	futures feed notebook
Strategy decisions and state	Strategies, Stateful Strategies	Strategy.on_data	quickstart, risk/state	stateful strategies notebook
Orders and fill timing	Orders, Execution Semantics	Broker.submit_order	orders/timing	single-asset notebook
Multi-asset allocation	Rebalancing	Broker.rebalance_to_weights	equity/ETF, futures, FX examples	allocator comparison notebook
Cash, margin, sizing	Accounts, Configuration	BacktestConfig	accounts/constraints	conformal sizing notebook (method, not engine API)
Costs and funding	Market Impact	BacktestConfig	equity costs, funding, flip	gross/net notebook (manual analysis)
Position and portfolio risk	Risk Management	StopLoss	risk/state and eight risk checks	library risk demo
Execution profiles	Profiles	BacktestConfig.from_preset	profile inspection and one-setting comparison	engine divergence notebook
Fill, trade, equity exports	Results	BacktestResult	result export	performance reporting notebook
Zipline migration	Migrate from Zipline	Engine.run, Broker.order_target_percent	new target-weight round trip	engine divergence notebook

The book notebooks are learning material, not clean-wheel fixtures. Book-specific helpers and datasets remain in the companion repository. The migration page states that bundle ingestion, data.history, record, and scheduling have different interfaces, and that the zipline profile covers a documented comparison protocol rather than arbitrary native Zipline behavior. Same-bar timing, costs, corporate actions, and unsupported parity pairs retain their existing limits.

Verification

• Current public book main was 2d6e8f95eeccaee66906245606471f570b5807e5; inspected the linked notebook cells and book helper where a role assertion depends on it. Generated book_link_manifest.json from a shallow Git checkout; check_book_links.py verified 22 paths and their pinned revision.
• Built a candidate wheel, installed it into a clean Python 3.12 environment, and passed all 34 required documentation examples with recorded output. The checker bound 146 direct public API calls. Installed ml4t-diagnostic[viz]==0.1.4 into that environment and passed the optional diagnostic handoff and tearsheet examples.
• mkdocs build --strict passed. The rendered link check passed 8,422 content links across the complete site and 33 external guide destinations, including section anchors and pinned book URLs. A new contract test covers homepage-to-guide anchors.
• Documentation contract tests: 18 passed. Full local non-benchmark suite: 2,212 passed, 9 skipped, 5 deselected; 87.25% coverage against an 84% gate. Ruff lint/format and all repository pre-commit hooks passed; git diff --check passed.

Related to #120. Keep the issue open until this follow-up is reviewed and the deployed guide is checked. No book or other library repository was changed.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.