High Value Data Progress Review

Run high_value_data_progress_review_20260720T061235Z | generated UTC 2026-07-20T06:12:35.5733466Z | scheduled next run 07/20/2026 17:00:00

Pause Snapshot

Scheduled taskDisabled
Daily pull pauseTrue
Last task result267014
Active workers0
Active tasks0
Active provider leases0
Model rows ready0/24
Pause reasonControlled high-value data backfill, model, and report work; keep daily scheduled pulls paused until operator resumes.

Approach Recommendation

The broad data foundation is now good enough to stop measuring progress by raw pull volume alone. The next highest-value phase is narrower: paper-trading evidence, stop-loss labels, semantic relationship quality, and promotion gates.

  1. Pause broad historical pulls unless a specific model gate needs more rows.
  2. Put the next effort into paper-trading readiness: monitor, account snapshots, fills, slippage, and stop-loss case labels.
  3. Use the five positive component candidates and twenty stack-validation rows as the next model-training queue.
  4. Treat relationship graph growth as a semantic-quality problem now, not just an edge-count problem.
  5. Keep live/real trading disabled until paper fills, risk limits, kill switch, and operator review are all green.

High Value Workstreams

Workstream State Evidence Remaining Risk Best Next Move Report
Intraday market data foundationStrong local foundation1,374 symbols, 52,517,125 bar/trade/quote rows, cutoff 2026-07-17, zero orders in latest readiness report.Historical, bounded WebSocket, and supervised read-only service foundations are present. Market-hours soak evidence and downstream fill/risk gates remain.Use targeted intraday refreshes around top opportunities and stop-loss cases instead of broad pulls unless a coverage gate asks for it.Alpaca_Market_Data_Readiness.html
FMP bulk and fundamentals bridgeBulk bridge current; daily delta guarded18/18 datasets available or succeeded; latest rows published 3,617,716; latest observed 2026-07-19T23:54:52.870635+00:00. Daily delta bulk_first_with_recent_fallback_review: bridge 2026-07-17, latest-date symbol fallback rows 0, latest-date empty rows 0, bridge lag 0, report provider calls 0.Latest bulk date is bridged without broad symbol-level fallback; older fallback or empty coverage remains a targeted review queue until bulk-first sequencing is fully promoted.Latest bulk coverage is clean; recent older fallback or empty rows should stay in a small review queue rather than becoming daily baseline work.FMP_Daily_Delta_Readiness.html
Broader news and event historyProduction event-impact risk artifact; prospective evidence collectingLatest run has 1,876,996 article rows, 1,683,534 event evidence rows, 2,408 snapshots, and 16,105 component signals over 3,650 days. Governed impact model 20260720T053747Z: 168,700 matured event days across 2,322 symbols/3,414 dates; 63,808 matured historical shadow predictions; holdout RMSE lift vs constant 0.14003974209876024, incremental news lift vs price-only 0.0009431655694244201, rank correlation 0.5864598752118867, quintile spread 0.08334469006648253; strict prospective outcomes 0/0 dates; status production_model_collecting_strict_prospective_paper_evidence.The model artifact is production-pinned for advisory paper risk, but 500 matured strictly post-assignment predictions across 20 dates remain mandatory before any decision influence.Run daily outcome maturity and scoring once per completed news/price publication; do not retrain repeatedly or grant sizing/order authority while prospective gates are open.News_Event_Impact_Model.html
Live market data readinessSupervised read-only service controls ready; market-hours soak pendingStatus live_stop_research_only: bounded stream feeds 2, test/prod verification 1/1, stream data events/orders 2/0; supervisor feeds/controlled 2/2, test/prod/market-hours/kill runs 2/1/0/1, reconnect runtime runs 0, connections/events/orders 2/2/0; stop gates blocked/passed 2/4, signal dates 17/126, paper fill samples 2/100, ready quote symbols 0/50, orders submitted 0, submission enabled False; expected-cost paper shadow collecting_strictly_post_cutoff_paper_shadow, dates 0, remaining 20.The supervisor implementation now has heartbeat, reconnect, bounded-buffer, stale-feed, backpressure, and kill-switch controls. A real market-hours soak, account-risk limits, and enough paper fill/slippage samples remain before real trading.Run a supervised SIP soak during market hours, keep order submission disabled, and promote only monitoring/reconciliation/reporting until fill and account-risk gates clear.Alpaca_Live_Trading_Readiness.html
Execution and fill data readinessThin, gated; paper collection queue visible3 fill feature rows, 2/100 candidate training samples, 3 quote matches, 3 trade matches, 3 open gates. Collection queue 4: 0 paper-fill candidate, 1 proxy-review intent, 1 blocked intents, 2 no-fill-needed rows; paper execution paper_plan_ready, plan-only True, orders 0, provider calls 0; collection stage proxy_review_required_before_paper_collection, allowed now False, possible when authorized False; block reasons sample_depth_gap_98; wide_spread_review_rows_1; slippage_training_guardrail_blocked; paper_order_submission_switch_disabled; no_clean_collection_candidate; proxy_review_intents_1. Immutable expected-cost model 20260719T172030Z is pinned to paper_shadow; historical promotion ready True, prospective shadow dates 0, remaining 20, status collecting_strictly_post_cutoff_paper_shadow.Expected-cost pretraining is promoted to paper shadow, but 20 future shadow dates and actual clean broker-fill calibration remain required before production sizing review.No clean paper-fill candidates are available yet; refresh/review market-hours intraday proxy context for 1 eligible intent(s), then collect 98 additional linked, fresh-quote, tight-spread paper fills before slippage-sensitive training.Alpaca_Fill_Slippage_Features.html
SEC filing text depthSection-aware semantic change foundation working19,740 filing evidence rows, 2,405 reviewed symbols, 3 unknowns, 23,293 SEC component-signal rows. Section layer: 18,876 documents, 2,378 symbols, 57,312/73,909 sections extracted (0.7754), 46,936 prior-filing comparisons, high/change review 2,161/16,798, filings 2000-06-29-2026-07-01, status section_change_features_ready_for_research.Review issuer-format extraction gaps and calibrate semantic changes against matured post-filing outcomes before model weighting.Join section-change candidates to matured returns and ontology concepts; keep pattern topics as research flags, not allegations.SEC_Filing_Section_Change_Foundation.html
SEC post-filing downside-risk modelProduction paper-risk model; prospective evidence collectingVersion 20260719T191251Z: 12,901 matured events over 1,315 dates and 2,174 symbols; historical holdout 3,721, RMSE/MAE lift 0.029217989839482428/0.12403258859423949, rank correlation 0.3168720849870035, risk-quintile spread 0.08077541190613204, 80% envelope coverage 0.8414404729911314; semantic residual active false after validation lift 0.0025292181601336283; prospective rows/dates/matured 0/0/0; orders 0.Collect 100 strictly prospective matured predictions and integrate gap/fill behavior before this risk envelope can influence paper stop placement.Run only daily label maturity and scoring; keep semantic residuals disabled until they prove incremental validation value, and keep order authority at zero.SEC_Filing_Downside_Risk_Model.html
Comtrade and global trade foundationResearch feature foundation readyComtrade has 2,719 observations, 12 reporters, 2,719 commodity-code observations, 5,320 projected graph edges, years 2015-2025; TiVA has 109 observations, 4 measures, and 109 graph edges. Direct YoY cohort: 1,563 eligible rows, 950 shock labels, 613 reference rows, 12 reporters/10 years; 238 confirmed absences retained as context and 1,363 unresolved gaps excluded; status trade_shock_training_foundation_ready.The direct-observation training cohort is usable, but unresolved historical gaps stay excluded and the lake remains intentionally priority-industry shaped rather than globally exhaustive.Run sector and horizon ablations on the eligible direct-year cohort; use further Comtrade pulls only for a named energy, metals, semiconductor, shipping, agriculture, or supply-chain model gap.Global_Trade_Data_Readiness.html
Entity and relationship graphLarge graph; research_feature_ready_quality_reviews_governed2,138,296 relationships across 192 grouped rows, 36 source families, and 81 relationship types; concept/meaning coverage 1.0/1.0; model usability quality_reviews_governed_candidate_research_allowed, training allowed False, ready horizons 2, next horizon/date gap 1/0, quality review rows 74 covering 519,045 relationships; manifest research_feature_ready_quality_reviews_governed, policy governed_candidate_relationship_features_allowed_for_research_training_only, candidate/review relationship rows 1,619,211/519,045, research training True, challenger promotion False, governed/ungoverned families 141/0, latest ablation rejected_at_selection_gate with 0/216 passing trials, provider calls 0, orders 0.All populated relationship families now have explicit full, reduced, or zero-weight dispositions. The latest 216-trial H1 quality ablation produced zero passing trials, so challenger promotion remains rejected even though governed research training is allowed.Keep the current H1 baseline and governed research weights. Do not repeat the rejected same-block ablation; reopen promotion only after a new temporal block, materially new relationship features, or source improvements satisfy the documented family conditions.Entity_Relationship_Readiness.html
Options data planning and feature plumbingHistorical OPRA training foundation ready; secondary features maturingSnapshots: 1,038 underlyings, 639,362 rows, latest 2026-07-19T20:22:19.722102+00:00, readiness stale_option_chain_seed: 1,122. Historical OPRA: 5 underlyings, 768,670 bars/trades, 204 contracts, max 135 market dates, latest 2026-07-19T20:43:47.736444+00:00, readiness collecting_historical_depth: 4; historical_option_training_foundation_ready: 1. Daily features: 200 rows across 5 underlyings/135 dates through 2026-07-17. Components production-ready 0/6, max evaluated/remaining dates 3/18.Historical depth now supports secondary feature research, but option components still need independent evaluated dates and ablation evidence before promotion.Mature existing option component outcomes and validate them as secondary stock features; avoid broad new pulls unless a coverage gate asks for them.Alpaca_Options_Data_Readiness.html
Model promotion and stop-loss researchResearch actions visible; production blocked0/24 model rows ready for review; component triage: inspect_inverse_or_risk_orientation: 11; train_next_positive_candidate: 13; validate_in_stack: 37. Critical path: 3 actionable research rows, 11 blocked gates, 1 review gates; direct challengers 1, inverse/risk tests 1, stack ablations 17, optimizer-date gap 109, paper-fill gap 98, top-15 execution-proxy gap 15. Training preflight: 13 groups, 14 missing-policy queue rows, 22 tested-not-weighted rows, 2 row-depth wait groups, largest row gap 1,552. Latest queue ablation: 1 rows, 1 evaluated, 1 positive-review rows, best MSE lift 0.046523983079177884, best directional delta 0.011938202247190999. Stop selector: 22 priority rows, 7 coverage-backfill rows, 1 positive clean / 3 positive-with-friction / 9 friction-rejection rows, 145 training examples, launch gates blocked/review/passed 7/0/2, selector signal-date gap 109, execution-proxy ready top-15/top-50 0/0, wide-spread review 12; paper/live/artifact allowed 0/0/0. Intraday stop proxy: 41/50 symbols covered, 4 full-depth, 27 partial-depth, latest proxy date 2026-07-17, evidence rows shown 25; proxy evidence is not paper-fill/slippage evidence.Production promotion remains blocked by zero ready rows, stop-loss selector date/temporal gates, paper-fill sample depth, and execution-proxy review; intraday stop proxy evidence helps research but does not replace actual fill/slippage evidence.Use bounded queue ablations to separate candidates worth review from neutral/no-lift candidates, let h20/h63 rows mature, and keep selector training, artifact writes, live trading, and order submission disabled until hard gates clear.Model_Promotion_Readiness.html
Component outcome maturity diagnosticsDiagnostic surface workingLatest diagnostic component_outcome_sync_diagnostics_20260718T004510Z: inspected 47,358 pending rows, wrote 0 matured evaluations, left 47,149 session-immature rows, and found 209 stale-price rows across 15 symbols; missing-asset/unreadable/invalid rows 0/0/0; stale actions: review_provider_activity_flag_conflict: 15; triage stages: manual_identity_review_after_local_bar_evidence: 2; manual_identity_review_after_local_bridge: 13; no-repeat provider-call symbols 15; local repair candidates 0; training-label excluded symbols 15.Some queued model inputs are true exact-session maturity waits, while stale-price rows need price-history repair or provider/source review before training; the repair queue and Instrument Source Review now separate no-repeat provider-call symbols, local repair candidates, and rows excluded from training labels.Do not spend more price-provider calls on symbols marked no-repeat until identity/activity review resolves them; repair only local-repair candidates, rerun diagnostics, then train candidates whose evaluated rows and date gates actually improve.Component_Outcome_Sync_Diagnostics.html

Key Public Surfaces

ReportExistsAge HoursBytes
Pipeline_Health.htmltrue1.140,863
High_Value_Data_Progress_Review.htmltrue1.221,128
FMP_Daily_Delta_Readiness.htmltrue5.617,540
FMP_Bulk_Bridge_Readiness.htmltrue5.617,176
Global_Trade_Data_Readiness.htmltrue1.342,213
Alpaca_Market_Data_Readiness.htmltrue8.129,428
Alpaca_Execution_Training_Foundation.htmltrue13.25,879
Alpaca_Execution_Cost_Model.htmltrue12.93,553
Alpaca_Live_Trading_Readiness.htmltrue12.343,460
Alpaca_Fill_Slippage_Features.htmltrue32.231,780
Alpaca_Options_Data_Readiness.htmltrue9.8167,480
News_Event_Training_Coverage.htmltrue0.432,845
News_Event_Impact_Model.htmltrue0.65,799
SEC_Filing_Intelligence.htmltrue33.029,142
SEC_Filing_Section_Change_Foundation.htmltrue1.931,319
SEC_Filing_Downside_Risk_Model.htmltrue4.35,566
Market_Relationship_Weight_Training.htmltrue7.368,363
Entity_Relationship_Readiness.htmltrue6.751,685
Model_Promotion_Readiness.htmltrue0.01,088,213
Component_Queue_Ablation.htmltrue14.851,477
Component_Outcome_Sync_Diagnostics.htmltrue53.532,477
Instrument_Source_Review.htmltrue53.522,980
Paper_Strategy_Optimizer.htmltrue66.4375,820