fix(notebook,#18124): research_asset_class_momentum -- local research path + real 12y top-3 backtest replace placeholder by jsboige · Pull Request #18162 · jsboige/CoursIA

jsboige · 2026-09-28T00:37:48Z

Grain: MED/notebook-python -- lane myia-po-2026:CoursIA -- prev: #18160

Summary

SOTA repair of Research-Executor/research_asset_class_momentum.ipynb (1st of
the placeholder set in the #18124 audit).

Was: QuantBook-only cells (unrunnable locally) ending in a fake
"BACKTEST RESULTS" placeholder printing unverified "expected
characteristics" (CAGR ~8-12%, MaxDD ~15-25%, Sharpe ~0.6-0.9).
Now: honest local research path (same 5 asset-class ETFs
SPY/EFA/BND/VNQ/GSG via yfinance, tz-normalized per env-gotchas, unified
close frame on both QC/local paths) and a real vectorized backtest of
the strategy itself: at each month-end rank by 252d momentum, hold top 3
equal-weight next month, 132 months measured.

Honest measured findings (SOTA-OK)

Strategy CAGR 10.25% | Vol 13.45% | Sharpe 0.76 | MaxDD -30.05% vs on
the same window: SPY buy&hold 15.38%, equal-weight-5 8.51% (final
equity 2.92x / 4.81x / 2.45x).
Momentum top-3 beats the naive equal-weight but trails SPY buy&hold
over this 12y window -- stated as measured, no spin.
The measured MaxDD (-30%) exceeds the old placeholder's "expected
15-25%": the previous claims were design expectations, not results.
Current selection measured live: GSG (+54.97% 252d momentum), SPY, EFA.
No costs/slippage modeled; production engine on QC Cloud remains the
reference -- stated in the closing note.

Validation

Papermill: 9/9 cells, 0 errors, all execution_count set (C.2)
C.1: no raise NotImplementedError / assert False / 1/0 (verified)
Catalogue byte-identical to main (single .ipynb changed)
SOTA verdict: SOTA-OK -- real ETF data, real strategy backtest,
committed outputs are the real outputs

See #18124 (partial: 4 of 10; remaining from this lane:
defensive_etf_rotation, long_short_harvest).

🤖 Generated with Claude Code

… path + real 12y top-3 backtest replace placeholder SOTA repair (1st of the placeholder set in the #18124 audit): - The notebook was QuantBook-only (unrunnable locally) and ended in a fake "BACKTEST RESULTS" placeholder with unverified "expected characteristics". - Added the honest local research path used across the family: the same 5 asset-class ETFs (SPY/EFA/BND/VNQ/GSG) via yfinance, tz-normalized, same `close` frame downstream on both paths. - Replaced the placeholder with a REAL vectorized backtest of the strategy itself: monthly top-3 by 252d momentum, equal weight, next-month holding, 132 months measured -- CAGR 10.25%, Sharpe 0.76, MaxDD -30.05% vs SPY 15.38% and equal-weight-5 8.51% on the same window. - Honest findings written in: momentum beats the naive equal-weight but trails SPY buy&hold over this window, and the measured MaxDD (-30%) exceeds the placeholder's optimistic "expected 15-25%" -- the old claims were design expectations, not results. Executed 9/9 cells, 0 errors, all execution_count set (C.1/C.2). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

github-actions · 2026-09-28T00:38:44Z

Notebook PR Validation: PASS

Notebooks checked: 1
Code cells validated: 5
Result: All passed

Checks: H.1 (no errors), H.3 (execution_count), C.1 (no banned patterns)
Non-Python kernels (.NET/Lean): C.1 + errors only (execution_count advisory)
QuantConnect notebooks: C.1 + errors only (require QC Cloud for execution)

github-actions · 2026-09-28T00:39:02Z

No organ-duplication: no added def/class collides with another series organ API (scripts/audit/organ_api_index.yaml).

Detector: python scripts/audit/detect_organ_duplication.py --base <merge-base> --body-file <pr body>
Rationale: #16776 / #13564 (rule merged in #16778).

github-actions · 2026-09-28T00:39:02Z

Golden-Set Execution (H.7 P3)

✅ 8/8 notebooks passed (certified reproducible)

Notebook	Status	Time
2.1-Workflow-ML.ipynb	✅ SUCCESS	3.3s
2.2-Descente-de-gradient.ipynb	✅ SUCCESS	3.4s
2.3-Regression-lineaire-logistique.ipynb	✅ SUCCESS	4.1s
2.4-Arbres-Forets-Ensembles.ipynb	✅ SUCCESS	4.2s
Search-01-StateSpace.ipynb	✅ SUCCESS	3.2s
SL-1-LogicalLearning.ipynb	✅ SUCCESS	2.1s
rl_4_multi_armed_bandits.ipynb	✅ SUCCESS	16.2s
GameTheory-04c-NashExistence-Python.ipynb	✅ SUCCESS	2.8s

Pinned lockfile: scripts/notebook_tools/golden_set.lock.txt (H.7 P3, axe A #4208)

github-actions · 2026-09-28T00:40:49Z

Notebook outputs-required (H.4 schema): PASS (every code cell carries an `outputs: list`)

github-actions · 2026-09-28T01:05:53Z

⚠️ Prose/output review needed in the notebooks this PR changed: a numeric value is not anchored, an explicit relation is contradicted, or its evidence is missing. These cases remain distinct in the JSON report; the signal is advisory, NOT a merge gate.

Scope = notebooks CHANGED in this PR, not the whole corpus. Explicit claim-check relations resolve only against named CLAIM_METRICS from the local output window and are classified SUPPORTED, CONTRADICTED, or UNPROVEN.
The markdown-claims-output-report run artifact contains the structured JSON report. See python scripts/check_markdown_claims_output.py --help for re-running locally.
Detector rationale: c.290 / c.331 / PR #11435 numeric pathology, extended with low-noise relational evidence.

…, 3.11 -> 3.13 to match series env)

…python3 (3.11) kernel to match base stamp

github-actions · 2026-09-28T13:52:10Z

Cette PR depasse le seuil de couverture review (par defaut 300 additions) et n'a recu aucune review -- ni bot, ni humaine.

Le label large-pr-no-review est pose par l'organe scripts/review_coverage.py porte par l'issue #11232. Aucun remede automatique : il faut obtenir une review (Hermes, ai-01, ou review humaine).

Le label est retire au balayage suivant (quotidien) des qu'une review arrive -- dans reviews[] ou en commentaire de verdict -- ou que le diff passe sous le seuil. Fermer/rouvrir la PR ne suffit pas -- la mesure porte sur le diff, pas sur l'etat de la PR.

Seuil, historique et exceptions : cf. docs/reference/review-coverage-threshold.md.

jsboige · 2026-09-29T02:34:58Z

[ADJOINT PREFLIGHT]
schema: 1
lane: myia-po-2023:CoursIA
pr: 18162
head: c18632d
complete: true
body: read
comments-reviewed: 6
reviews-reviewed: 0
threads-reviewed: 0
threads-unresolved: 0
surfaces-sha256: f23188c857003f4b9a052fdb2923099c0ada19ad4603a8e60e722d6321f6d302
diff-files: 1
diff-additions: 303
diff-deletions: 143
checks: latest-wins-green
b0: clear
scope: pass
domain: not-applicable
verdict: READY
[/ADJOINT PREFLIGHT]

Note de lecture (informatif, ne change pas le verdict) : le commentaire sticky markdown-claims-output-advisory signale une valeur numerique / relation non ancree dans le notebook change -- advisory par design, le detail vit dans l'artifact markdown-claims-output-report du run. A tracer dans la lecture finale (meme classe de prose que le realignement livre sur #18314 ce jour).

jsboige mentioned this pull request Sep 28, 2026

audit(sorties): 3 notebooks dont la prose décrit un chemin que les sorties montrent sauté + 30 chemins dégradés déclarés (audit du 27/09) #18124

Open

github-actions Bot added the variation-tag-prev-absent Tag Grain sans 'prev: <TIER>/<GENRE> #<PR>' (adjacence G-VAR-3 inevaluable) label Sep 28, 2026

jsboige mentioned this pull request Sep 28, 2026

fix(notebook,#18124): research_defensive_etf_rotation -- measured signal table + real rotation backtest replace placeholder #18163

Merged

jsboige added 3 commits September 28, 2026 03:16

fix(notebook,#18124): re-execute under python313 (kernel-drift repair…

6e73346

…, 3.11 -> 3.13 to match series env)

merge: refresh base from origin/main

8b5fcfb

fix(notebook,#18124): re-execute research_asset_class_momentum under …

c18632d

…python3 (3.11) kernel to match base stamp

github-actions Bot added the large-pr-no-review PR > seuil sans review (ni bot ni humaine) -- retire quand une review arrive (#11232) label Sep 28, 2026

jsboige mentioned this pull request Sep 29, 2026

fix(notebook,#18124): research_long_short_harvest -- real VIX quintiles + measured Hurst + trained ML core replace placeholder #18164

Merged

myia-ai-01 merged commit c68be11 into main Sep 29, 2026
90 of 93 checks passed

Navigation Menu

Repository navigation

fix(notebook,#18124): research_asset_class_momentum -- local research path + real 12y top-3 backtest replace placeholder - #18162

fix(notebook,#18124): research_asset_class_momentum -- local research path + real 12y top-3 backtest replace placeholder#18162
myia-ai-01 merged 4 commits into
mainfrom
feature/18124-researchexec-acm

jsboige commented Sep 28, 2026

Uh oh!

github-actions Bot commented Sep 28, 2026

Uh oh!

github-actions Bot commented Sep 28, 2026

Uh oh!

github-actions Bot commented Sep 28, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Sep 28, 2026

Uh oh!

github-actions Bot commented Sep 28, 2026

Uh oh!

github-actions Bot commented Sep 28, 2026

Uh oh!

jsboige commented Sep 29, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

jsboige commented Sep 28, 2026

Summary

Honest measured findings (SOTA-OK)

Validation

Uh oh!

github-actions Bot commented Sep 28, 2026

Notebook PR Validation: PASS

Uh oh!

github-actions Bot commented Sep 28, 2026

Uh oh!

github-actions Bot commented Sep 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Golden-Set Execution (H.7 P3)

Uh oh!

github-actions Bot commented Sep 28, 2026

Notebook outputs-required (H.4 schema): PASS (every code cell carries an outputs: list)

Uh oh!

github-actions Bot commented Sep 28, 2026

Uh oh!

github-actions Bot commented Sep 28, 2026

Uh oh!

jsboige commented Sep 29, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

github-actions Bot commented Sep 28, 2026 •

edited

Loading

Notebook outputs-required (H.4 schema): PASS (every code cell carries an `outputs: list`)