Skip to content

v0.2.4 - Metric Clarity and Contributor-Led Review

Choose a tag to compare

View all tags
@initial-d initial-d released this 06 Aug 13:03
v0.2.4
58e3727
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

v0.2.4 makes the time basis of backtest trading costs explicit. An external
contributor showed that the old cost_drag label sat beside annualized metrics
even though it represented a cumulative total over the whole run.

The calculation was correct; the reporting contract was too easy to misread.

Highlights

  • New output uses cost_drag_cumulative in result tables and reports.
  • The legacy cost_drag key and property remain available as compatibility
    aliases.
  • Aggregation and audit tools continue to accept archived reports.
  • A metric glossary now records the unit and time basis of every validation
    column.
  • Regression tests cover duration scaling, fee scaling, and legacy access.
  • The README now leads with the runnable path and community-reviewed evidence.
  • Package metadata and mlquant.__version__ are synchronized at 0.2.4.

Community Credit

Thank you to @sergio12S for identifying the
ambiguity, supplying controlled evidence, designing the compatibility path,
and adding the implementation, documentation, and tests in
PR #47.

Read the full technical story:
The Backtest Metric Was Correct — and Still Easy to Misread.

Migration

Prefer the explicit field in new code:

result.cost_drag_cumulative
result.metrics["cost_drag_cumulative"]

Existing reads continue to work:

result.cost_drag
result.metrics["cost_drag"]

Verification

  • Full local test suite: 92 passed.
  • Ruff: passed.
  • GitHub CI: Python 3.9, 3.10, and 3.11 passed, including the CLI smoke test.

Research Boundary

This release improves reporting clarity. It does not add a new trading signal,
change historical validation results, or claim live-trading profitability.

Assets 2
Loading