feat(schema): Epic #150 Phase 2.1 — explicit valuation_methods_applicable field by dackclup · Pull Request #161 · dackclup/quantrank

dackclup · 2026-05-20T19:05:34Z

Summary

Epic #150 Phase 2.1 — surfaces the positive-framed count of valuation methods that produced a non-outlier applicable estimate as StockDetail.valuation_methods_applicable (top-level + mirrored inside the fair_price dict).

Schema bump 0.9.3-phase4h.3 → 0.9.4-phase4h.4
New helper: compute.valuation.ensemble._count_applicable_non_outliers
EnsembleResult dataclass gains valuation_methods_applicable: int = 0
compute_fair_price_ensemble computes it from extreme_*_estimate warnings via the new helper (zero recomputation cost — re-uses the already-emitted warning list)
ensemble_result_to_dict emits it (inside fair_price.valuation_methods_applicable)
compute/main.py writer passes it to StockDetail at top level
frontend/lib/types.ts mirrors in both FairPriceEnsemble (nested) and StockDetail (top level)
frontend/lib/schema-snapshot.json regenerated

Why

Per Phase 0 of epic #150 (Known Limitations): extreme_*_estimate warnings are method-applicability signals, not manipulation signals. Today consumers have to count those warnings to derive the applicable-method count. This PR exposes the count directly at the schema-snapshot level so filtering / audits / future UI can read it without unpacking the warning list.

Scope: additive-only

This PR does NOT migrate the existing consumers:

compute/scoring/loss_chance.py still reads extreme_*_estimate flags for the per-flag penalty
frontend/components/FairPriceBarChart.tsx still uses extreme_*_estimate warnings for outlier-bar shading

Migrating those is a follow-up PR (Phase 2.1b). Keeping this PR additive avoids a breaking change to valuation_warnings semantics and preserves the 0.10.0 minor bump for the structural Phase 3 work.

Tests

7 new cases in tests/test_valuation/test_ensemble.py Section J:

J1 helper excludes outliers + skips
J2 all-skipped yields 0
J3 all-outliers yields 0
J4 EnsembleResult default is 0 (covers hard-stale + data-quality null paths)
J5 ensemble_result_to_dict emits the field
J6 full-ensemble integration on synthetic IT stock (3 methods compute, no outliers → 3)
J7 hard-stale short-circuit path (0)

Existing test_I1_ensemble_result_to_dict_shape_matches_ts_type updated for the new key. test_schema_version_is_phase4h_4 updated to match the bump.

Full offline suite: 945 passed, 7 skipped (ipca / qlib not installed), 18 deselected (network).

Test plan

ruff check . clean
python -m compute.output.schema_check in sync
python -m pytest tests/ -m "not network" — 945 passed (excluding 3 pre-existing collection errors from missing openassetpricing in this CI env)
cd frontend && npx --no -- tsc --noEmit — no new errors related to this PR
Section A-J + Section I Playwright spot-check on next workflow_dispatch green (post-merge)

Notes

Part of epic Foundation reconciliation roadmap — process + analytical scrutinize findings (2 rounds) #150 (Phase 2.1)
Defense surface unchanged — purely metadata + observability
Field is int | None on the Pydantic / TS side so legacy snapshots cast cleanly; the writer always emits a non-null int from 0.9.4+
Severity: low (additive observability; no behavior change)

https://claude.ai/code/session_01Nj5sMzisnqDmF46g5ckEJn

Generated by Claude Code

…able field Surfaces the positive-framed count of valuation methods that produced a non-outlier applicable estimate as `StockDetail.valuation_methods_applicable` (top-level + mirrored inside `fair_price` dict). The field is the inverse of the count of `extreme_*_estimate` warnings emitted by `compute.valuation.ensemble._classify_outliers`. Why: per Phase 0 of epic #150, the `extreme_*_estimate` warnings are method-applicability signals, NOT manipulation signals. Today consumers have to count those warnings to derive the applicable-method count; this PR exposes the count directly at the schema-snapshot level so filtering / audits / future UI can read it without unpacking the warning list. Additive only: `extreme_*_estimate` flags stay in `valuation_warnings` for back-compat with `loss_chance.py` (penalty input) and `FairPriceBarChart.tsx` (UI outlier shading). A future PR can migrate those consumers to read the new field; that's out of scope here. Schema bump 0.9.3-phase4h.3 → 0.9.4-phase4h.4. The 0.10.0 minor bump is preserved for the structural Phase 3 work (pillar correlation + top-decile veto → joint-gate). Tests: 7 new cases in `tests/test_valuation/test_ensemble.py` Section J covering helper unit + dict round-trip + integration via full ensemble path + hard-stale default. Existing `to_dict_shape` test updated for the new key. https://claude.ai/code/session_01Nj5sMzisnqDmF46g5ckEJn

vercel · 2026-05-20T19:05:41Z

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
quantrank	Ready	Preview, Comment	May 20, 2026 7:06pm

…#164) Adds the Phase 3 correlation analysis: a reproducible one-shot script (`scripts/phase3_flag_correlation.py`) plus baseline outputs under `docs/phase3-correlation/` containing firing rates, pairwise φ-coefficients, redundancy candidates, diversity-confirmed pairs, a heatmap PNG, and an interpretive findings doc. Methodology: - φ-coefficient (Matthews correlation for boolean variables) — the proper analog of Pearson for the {0,1} case. - Sources merged: `risk_flags` + `valuation_warnings` + boolean keys in `tier2_events`. - Universe: 502 S&P 500 constituents (production output @5dfe6287, pre-Phase-2.4 so `loss_avoidance_pattern` is still 0% — re-run after the next weekly cron to include it). Headline findings (full version in `findings.md`): - Defense layer is mostly orthogonal — 35 diversity-confirmed pairs (|φ| ≤ 0.05, both base-rates ≥ 5%) vs 15 redundancy candidates (|φ| ≥ 0.30). `altman_distress` is orthogonal to nearly every other flag — its financial-distress signal is genuinely distinct. - `restatement_history` is independent of the Sloan/Beneish/accruals manipulation cluster (φ ≈ 0). Phase 2.2 recalibration can proceed without redundancy concerns. - Warning-band ↔ active-veto pairs (Dechow + Beneish) are correlated by design (φ ≈ +0.6-0.7) — confirms the half-PPV Phase 2.5 calibration of `BENEISH_HIGH_WEIGHT` and `DECHOW_HIGH_WEIGHT`. - `manipulation_triple_flag` is φ-locked to `dechow_high` at current sample size (n=2 each). Watch in Q3 cohort audit — if persistent, downgrade `TRIPLE_FLAG_WEIGHT = 10.0` (currently labeled gut-feel in Phase 2.5 provenance docstring). - `extreme_*_estimate` family clusters (φ ≈ +0.4-0.5) — already surfaced positively as `valuation_methods_applicable` (PR #161). No additional work needed. Decision matrix for downstream Phase 2.x PRs in `findings.md`. Doc + script only — no compute / schema / output change. Re-run after every quarterly cohort audit (next: 2026-08-19) + after each Phase 2.x recalibration PR lands. Closes the Phase 3 deliverable in epic #150. https://claude.ai/code/session_01Nj5sMzisnqDmF46g5ckEJn Co-authored-by: Claude <noreply@anthropic.com>

Second deliverable from the 14-subagent full self-audit (2026-05-21). Doc drift surfaced by `docs-reviewer` + `methodology-scientist` against the actual production state on `a16c887` (2026-05-20 cron) + the schema version on `compute/config.py:30`. Five reconciles in lockstep: (a) PHASE_STATUS.md §Current state — schema `0.9.2-phase4h.2` → `0.9.4-phase4h.4`; defense layer headline 17 declared → 27 emitted flags (per PR #154 reconcile); skill inventory 38 → 42; subagent inventory added (14 in 4 tiers); active-vetoes list explicit (7 named); recently-merged block refreshed from PR #146-back to PRs #148-#169 (~20 PRs since `v1.2.0-phase4.5`); next-deliverables list updated to drop the PR-#148-closed Epic #125 Item 3 entry and add the `loss_avoidance_pattern` size-invariant follow-up. (b) SKILL.md schema-version table — 2 missing rows added: `0.9.3-phase4h.3` (PR #160, explicit `Metadata.tier2_enabled` field, closes #117/#155) and `0.9.4-phase4h.4` (PR #161, additive `valuation_methods_applicable` field on `StockDetail` and nested in `fair_price` dict, additive only — no consumer migration). (c) WORKFLOW.md Phase 4.5d `loss_avoidance_pattern` task — threshold description `$5M / $0.05` → `$50M / $0.50` (PR #163 Phase 2.4 10× rescale). Adds the Phase 4 size-invariant follow-up note (NI / TotalAssets) carried from CLAUDE.md §Gotchas since the rescale still fires 0% on production. (d) docs/METHODOLOGY.md — Active vetoes 4 → 7 rows: adds `beneish_manipulation_veto` (Beneish 1999 *FAJ*), `dechow_manipulation_veto` (Dechow 2011 *CAR* Model 1), and `data_quality_input_corruption` (with the dual-path rule — `risk_overlay.py` snapshot-level TBVPS guard + `ensemble.py` output-level guard). Known-calibration-drift block refreshed: `value_trap_risk` (Issue #11 closed via PR #166), `going_concern_disclosure` (FP rate now 1.0% within Mayew band — verified 5/502 on production cron), `loss_avoidance_pattern` (still 0% after Phase 2.4 rescale, follow-up to size-invariant), `restatement_history` (Phase 2.2 high-confidence complement via PR #165). (e) README.md §Honest Limitations — adds Phase 2.2 `restatement_high_confidence` and Issue #11 closure (PR #166) entries; mentions PR #163's 10× rescale on the `loss_avoidance_pattern` line. CLAUDE.md §Phase status + AGENTS.md §Phase + version state — matching "Phase 2 in flight" entries per §Conventions "ship with every PR" rule. Deferred to a follow-up Phase 2.x PR (too large for this scope): METHODOLOGY.md annotate-only section full refresh (7 → 14+ flags) + missing citation blocks (full Hennes-Leone-Miller 2008 *TAR*, Cohen-Malloy-Pomorski 2012, etc.) — needs `methodology-scientist` sign-off per the new-defense-flow rule. Pre-commit verification: - docs-reviewer NEEDS-CLARITY-PASS → fixed `data_quality_input_corruption` rule description in METHODOLOGY.md to cover both veto paths (snapshot-level `risk_overlay.py` + output-level `ensemble.py`). - methodology-scientist findings: data_quality veto classification CONFIRMED via dual-surface emission; value_trap / loss_avoidance / restatement_history all CLEAN. - phase-coordinator Mode A: COLLISION-RISK with PR #170 (Phase 1 ops hardening on `claude/enable-subagents-standby-7lMN4`) on CLAUDE.md + AGENTS.md §Phase status block — both PRs append to the tail. Strategy: PR #170 merges first, then this branch rebases onto updated main; conflict is mechanical 2-paragraph merge. No compute / schema / output / dep change. Lint passes (no Python / TypeScript edited). https://claude.ai/code/session_01HtcGy4F59TaawNQ7V1Eysc Co-authored-by: Claude <noreply@anthropic.com>

…ion (#171) Second deliverable from the 14-subagent full self-audit (2026-05-21). Doc drift surfaced by `docs-reviewer` + `methodology-scientist` against the actual production state on `a16c887` (2026-05-20 cron) + the schema version on `compute/config.py:30`. Five reconciles in lockstep: (a) PHASE_STATUS.md §Current state — schema `0.9.2-phase4h.2` → `0.9.4-phase4h.4`; defense layer headline 17 declared → 27 emitted flags (per PR #154 reconcile); skill inventory 38 → 42; subagent inventory added (14 in 4 tiers); active-vetoes list explicit (7 named); recently-merged block refreshed from PR #146-back to PRs #148-#169 (~20 PRs since `v1.2.0-phase4.5`); next-deliverables list updated to drop the PR-#148-closed Epic #125 Item 3 entry and add the `loss_avoidance_pattern` size-invariant follow-up. (b) SKILL.md schema-version table — 2 missing rows added: `0.9.3-phase4h.3` (PR #160, explicit `Metadata.tier2_enabled` field, closes #117/#155) and `0.9.4-phase4h.4` (PR #161, additive `valuation_methods_applicable` field on `StockDetail` and nested in `fair_price` dict, additive only — no consumer migration). (c) WORKFLOW.md Phase 4.5d `loss_avoidance_pattern` task — threshold description `$5M / $0.05` → `$50M / $0.50` (PR #163 Phase 2.4 10× rescale). Adds the Phase 4 size-invariant follow-up note (NI / TotalAssets) carried from CLAUDE.md §Gotchas since the rescale still fires 0% on production. (d) docs/METHODOLOGY.md — Active vetoes 4 → 7 rows: adds `beneish_manipulation_veto` (Beneish 1999 *FAJ*), `dechow_manipulation_veto` (Dechow 2011 *CAR* Model 1), and `data_quality_input_corruption` (with the dual-path rule — `risk_overlay.py` snapshot-level TBVPS guard + `ensemble.py` output-level guard). Known-calibration-drift block refreshed: `value_trap_risk` (Issue #11 closed via PR #166), `going_concern_disclosure` (FP rate now 1.0% within Mayew band — verified 5/502 on production cron), `loss_avoidance_pattern` (still 0% after Phase 2.4 rescale, follow-up to size-invariant), `restatement_history` (Phase 2.2 high-confidence complement via PR #165). (e) README.md §Honest Limitations — adds Phase 2.2 `restatement_high_confidence` and Issue #11 closure (PR #166) entries; mentions PR #163's 10× rescale on the `loss_avoidance_pattern` line. CLAUDE.md §Phase status + AGENTS.md §Phase + version state — matching "Phase 2 in flight" entries per §Conventions "ship with every PR" rule. Deferred to a follow-up Phase 2.x PR (too large for this scope): METHODOLOGY.md annotate-only section full refresh (7 → 14+ flags) + missing citation blocks (full Hennes-Leone-Miller 2008 *TAR*, Cohen-Malloy-Pomorski 2012, etc.) — needs `methodology-scientist` sign-off per the new-defense-flow rule. Pre-commit verification: - docs-reviewer NEEDS-CLARITY-PASS → fixed `data_quality_input_corruption` rule description in METHODOLOGY.md to cover both veto paths (snapshot-level `risk_overlay.py` + output-level `ensemble.py`). - methodology-scientist findings: data_quality veto classification CONFIRMED via dual-surface emission; value_trap / loss_avoidance / restatement_history all CLEAN. - phase-coordinator Mode A: COLLISION-RISK with PR #170 (Phase 1 ops hardening on `claude/enable-subagents-standby-7lMN4`) on CLAUDE.md + AGENTS.md §Phase status block — both PRs append to the tail. Strategy: PR #170 merges first, then this branch rebases onto updated main; conflict is mechanical 2-paragraph merge. No compute / schema / output / dep change. Lint passes (no Python / TypeScript edited). https://claude.ai/code/session_01HtcGy4F59TaawNQ7V1Eysc Co-authored-by: Claude <noreply@anthropic.com>

vercel Bot deployed to Preview May 20, 2026 19:06 View deployment

dackclup marked this pull request as ready for review May 20, 2026 19:11

dackclup merged commit a16c887 into main May 20, 2026
4 checks passed

dackclup deleted the claude/session-close-audit-wrap-QkvFz branch May 20, 2026 19:12

dackclup mentioned this pull request May 21, 2026

docs(phase-3): Epic #150 Phase 3 — flag-correlation baseline analysis #164

Merged

5 tasks

This was referenced May 21, 2026

feat(agents): add .claude/agents/ — 4 project-specific Claude Code subagents #168

Merged

docs: Phase 2 doc-drift reconcile — 5 files match current implementation #171

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(schema): Epic #150 Phase 2.1 — explicit valuation_methods_applicable field#161

feat(schema): Epic #150 Phase 2.1 — explicit valuation_methods_applicable field#161
dackclup merged 1 commit into
mainfrom
claude/session-close-audit-wrap-QkvFz

dackclup commented May 20, 2026

Uh oh!

vercel Bot commented May 20, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

dackclup commented May 20, 2026

Summary

Why

Scope: additive-only

Tests

Test plan

Notes

Uh oh!

vercel Bot commented May 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

vercel Bot commented May 20, 2026 •

edited

Loading