Ship v0.3 real-data quantitative engineering showcase

[MateoBodon]

Outcome

• adds a deterministic 2017–2025 public-factor volatility-targeting case with strict t+1 chronology, target-position deltas, explicit commission/spread/impact costs, annual folds, uncertainty, risk-matched baselines, and corrected selection inference
• adds machine-verifiable schemas, source/output receipts, a generic artifact verifier, active engineering benchmarks, and a portable python -m microalpha surface
• leads README and Pages with real empirical evidence while keeping the conclusion honest: useful risk shaping, no statistically established alpha
• adds the ranked gap audit, before/after evaluator, resume bullets, interview explanation, visuals, and remaining weaknesses in PORTFOLIO_IMPACT_BRIEF.md

Local evidence

• 152 passed, 1 provider-only test deselected; 78.32% coverage
• Ruff, Black, isort, mypy, strict MkDocs, data-policy and secret gates pass
• clean-clone wheel install reports microalpha 0.3.0
• canonical market receipt: ee8fd20013c3469809c0b4054d61fa85db5eafe62de0551548c5595427bca1d1
• Audit Lab receipt remains 6e36c2397696d7e9eecbd058cbfc1ba522c8ffba7e5798224de86b20457b6575

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: b09945a62e

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Shift decision columns before publishing ledger

For each OOS row, daily is copied from primary, so decision_weight and realized_vol_at_decision are still calculated with the execution day's return; only executed_weight is lagged. The new signal_available_date then labels those same-day decision columns as if they were available the previous session, so daily.csv/the verifier present a look-ahead decision ledger even though the return calculation uses the lagged weight. Shift or relabel the decision columns before publishing them.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Allow target-weight orders that reduce capped exposure

When a target_weight order reduces an already-over-cap position but leaves it above the configured cap (for example 100% long -> 40% target with max_exposure=0.3), this emits a SELL delta, but the later exposure/single-name/gross cap checks still compare only the post-trade exposure to the cap and drop the order. That blocks de-risking unless the target is below every cap; the cap gates need to allow orders that reduce absolute/gross exposure from the current state.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Use the exported verifier name in API docs

The new module exports verify_artifact_dir, not verify_artifacts, so this documented snippet raises ImportError for users following the API reference. Update both the import and call to the actual function name.

Useful? React with 👍 / 👎.