fix(main): pe_ttm in raw_metrics still used eps_diluted (audit #6 final loose end)#51
Merged
Merged
Conversation
…d consumer) Final audit of post-PR-#49 production output revealed PE distribution median was 77.8 — virtually unchanged from the pre-fix 77.5. PR #49 fixed `pe_ratio` (Value pillar), `_build_universe_metrics` (peer median), `multiples_pe_fair_price` (fair-price method), and `graham_number` — but missed THIS one: `_build_raw_metrics` in compute/main.py builds the RawMetrics object that surfaces in the StockSummary / StockDetail JSON output, and it was still computing: pe_ttm = current_price / snapshot.eps_diluted # single-period bug This is the field that shows up as `raw_metrics.pe_ratio_ttm` in the public JSON the frontend renders. With this missed consumer, the production output kept shipping inflated PEs even after PR #49 merged (AAPL 61.6 vs correct 35.7, TSLA 3425 vs correct 432, ABBV 534 vs correct 105, ... 88.6% of universe). Fix: route through `NI_TTM / shares_outstanding`, matching the same formula in the other 4 already-fixed consumers. Validation pending workflow_dispatch re-run + audit. Expected: production median PE drops 77.8 → ~26 (target). Other corruption fixes already taking effect post-PR-#49 confirmed: - NVDA $215.9B revenue ✓ - AVB $3.1B (was $7M) ✓ - WMT 7.97B shares post-split ✓ - META 2.564B shares (was None) ✓ - BKNG NI $6.2B (was None) ✓ - WFC $85B / GS $60B / DUK $33B revenue ✓ - Beneish coverage 6.8% → 31.9% ✓ (5x improvement) - |NI|>|rev| corruption: 17 → 4 tickers ✓ This PR closes the last loose end of the audit-#6 deep-clean cycle. Full offline suite: 646 / 646 pass. https://claude.ai/code/session_015649aRyi2bvciQYZVNACd2
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
3 tasks
dackclup
added a commit
that referenced
this pull request
May 14, 2026
#55) Workflow run #30 (2026-05-14, 52m 59s) failed at the "Commit JSON outputs" step: [main f409f51] chore: update rankings 2026-05-14 1006 files changed, 23767 insertions(+), 23767 deletions(-) To https://github.com/dackclup/quantrank ! [rejected] main -> main (fetch first) error: failed to push some refs to ... hint: Updates were rejected because the remote contains work that you do not have locally. ... Root cause: workflow checked out commit 48855aa (PR #51's tip) and ran for 52 minutes. During that window two docs PRs landed (#52 recommendation-badge, #54 loss-chance + price-chart), advancing main ahead of the workflow's local HEAD. Push rejected with "fetch first". Fix: rebase the bot commit on top of remote main before pushing, with a 3-attempt retry loop. The bot commit only touches `frontend/public/data/` (502 stock JSONs + rankings.json + metadata), so rebase against ANY docs/code change is conflict-free — `--autostash` covers the edge case where the runner's pre-checkout state had any uncommitted noise. Without this fix, every workflow run is fragile to "main moved while compute was running" which becomes more likely as PR velocity increases (we had 7 PRs land in 24h yesterday). The compute itself SUCCEEDED — it's purely the commit-and-push step that was racy. https://claude.ai/code/session_015649aRyi2bvciQYZVNACd2 Co-authored-by: Claude <noreply@anthropic.com>
This was referenced May 14, 2026
dackclup
added a commit
that referenced
this pull request
May 17, 2026
… UI spot-check requirement (#101) PHASE_STATUS.md: row 4.5 → ✅ DONE (PRs #89/#90/#91 + #93 + #95 + #97 + #100). 4.5f ✅ DONE block added with production verified snapshot (run #51, commit e57f09c, warm-cache 5m14s), 5 schema field inventory, rank-source contract restatement, and live UI Section I spot-check results. Phase 4.5 timeline summary table 4.5f row flipped ⚪ → ✅; 4.5e marked deferred-until-after-tag. SKILL.md: schema versions table appended with 0.8.0-phase4.5f row covering the minor-bump rationale (5 new fields + new UI surface + semver-coupled tag), the FLAG_WEIGHTS additive design, and the Rule-16 rank-source contract. WORKFLOW.md: 4.5f task list [ ] → [x] for all sub-tasks. Captures module-location decision (manipulation_index.py kept separate from composite.py to keep compute_composite pure-pillar), 4.5e reserved-slot weights, and the new Section I spot-check result. Tag v1.2.0-phase4.5 is the only [ ] remaining — cut after this docs PR merges. CLAUDE.md: Phase status section refreshed for 4.5f complete. Schema 0.7.1-phase4g → 0.8.0-phase4.5f; defense layer 16 → 17; next deliverable: tag v1.2.0-phase4.5 → then 4.5e. Verification ladder amended with the new "after workflow_dispatch green" rule pointing at the Section I spot-check (REQUIRED 2026-05-17). .claude/skills/verify-production-output/SKILL.md: new Section I (Live UI visual spot-check via Playwright) added to the A-H table + a long-form section covering the 4-ticker matrix (worst- case stack, mid-band, boundary, top-clean), per-ticker capture template (card-present, headline, band-label, components, penalty, screenshot), the design-system regression mode this catches (the PR #70 invisible-dark-text-on-forced-light-bg class), and the sandbox / browser-version caveats (executable_path override + ignore_https_errors=True). Production verified: commit e57f09c / run #51 (workflow 25983422610) / 502 stocks / 856 offline tests / warm-cache 5m14s. Phase 4.5 cluster (4.5a-4.5f) complete; 6 sub-PRs landed in 1 calendar day after the cluster kickoff on 2026-05-16. Tag v1.2.0-phase4.5 is the next action after this merges. https://claude.ai/code/session_015649aRyi2bvciQYZVNACd2 Co-authored-by: Claude <noreply@anthropic.com>
5 tasks
2 tasks
dackclup
added a commit
that referenced
this pull request
May 28, 2026
… open-issues list (#291) Closes deferred substance items from PR #286 + PR #290 that opted to keep scope tight under the `pointer-to-CLAUDE.md` delegation pattern. Changes (1 file, AGENTS.md): - Line 377 — Production-verified run pointer bumped from cron #51 (`b1588b2a`, 2026-05-14) → cron #69 (`233117ac`, 13m 16s warm-cache, 2026-05-28). Cross-tool agents now have the current known-good baseline; verified by defense-layer-auditor + stock-detail-auditor on schema `0.10.7-phase4.6`. - Lines 380-382 — "Open Phase 4+ issues" list refreshed from 4 → 11 entries. Removed stale qualifier on #67 (PR #204 already landed data-collection; #67 is Phase 4 follow-up not Phase 5+). Added 7 missing issues: #115 (JKP license) · #130 (Q3 cohort) · #137 (9arm-skills license deadline 2026-06-17) · #150 (Phase 2-3 epic) · #287 (FORM4 revert + durable timeout) · #288 (GOOG/GOOGL XBRL) · #289 (NVR DQIC false positive). Added zero-exploitability context to #41. Doc-only PR. No compute / schema / scoring / valuation / frontend / Python / TS code change. CLAUDE.md substance untouched this PR (PR #286 + PR #290 already bumped it); this PR closes the AGENTS.md side of the lockstep. PHASE_STATUS_INFLIGHT.md side-file satisfies §Conventions lockstep per PR #237 convention. Co-authored-by: Claude <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Final-audit follow-up to PR #49. Production median PE was still 77.8 post-PR-#49 (essentially unchanged from 77.5). Root cause: PR #49 fixed 4 consumers of
snap.eps_dilutedbut missedcompute/main.py::_build_raw_metrics— the helper that builds the RawMetrics object that surfaces in the public StockSummary / StockDetail JSON. So the inflated PE values kept flowing into the frontend display + downstream consumers.Fix: route
pe_ttmthroughNI_TTM / shares_outstanding, matching the formula in the other 4 already-fixed consumers (pe_ratio,_build_universe_metrics,multiples_pe_fair_price,graham_number).Audit findings (post-PR-#49 production)
✅ What WORKED post-PR-#49 + cache key v2 bump:
|NI| > |revenue|corruptiondata_quality_input_corruptionveto count❌ What was BROKEN:
_build_raw_metricsmissedFiles changed
compute/main.py_build_raw_metricsnow derives TTM EPS fromNI / shares(was usingsnapshot.eps_diluted)11 LOC change, no new dependencies, no schema change.
Verification
ruff check .— cleanpython -m pytest -m "not network"— 646 / 646 passeps_dilutedfield still stored in raw_metrics for display, only the derivedpe_ratio_ttmmath was fixedTest plan (post-merge)
Phase 3e roadmap
https://claude.ai/code/session_015649aRyi2bvciQYZVNACd2
Generated by Claude Code