Enhance daily report with per-group hotspot detail and dual signals

Add daily Telegram report + backtest experiment scripts

C31 PASS: SHAP re-pruning drops peer_ret features; add nightly full precompute script

Wire C30 peer returns into production + add C29 MetaLabel log

Add C30 sector peer return features (PASSED 5-stock and 12-stock)

Merge C19: adaptive triple-barrier labeling into old_wang_dev

Add C19 adaptive triple-barrier labeling (PASSED 5-stock and 12-stock)

Add news sentiment pipeline (C28) and macro factor experiments C24-C27

Merge HF artifact commits; keep login-removal codebase

Daily precompute refresh: --max-age-hours replaces --resume

Add --push flag to full-universe precompute; push 2249 results to HF queue

Add full TW universe precompute + midnight LaunchAgent

C20: sector-conditional model — pool semis/financials/ETFs intra-sector

Fix full precompute wrapper defaults

Enable full-universe daily precompute run

Improve dashboard latency caching

Integrate Old Wang trading strategy overlay

Backup current stock predictor strategies

Apply optimized factor weight overlay

Add daily Hugging Face sync automation

Temporarily skip login and add validation config report

Merge remote-tracking branch 'hf/main' into codex/self-refine-validation-gate

feat: integrate local architecture with HF Space

feat: daily 00:00 watchlist precompute via LaunchAgent

feat(query_history): sync HF Space queries to Mac via queue

chore: sync md/harness/memory/rule architecture to current state

Fix HF queue prediction placeholders

fix: HF Space always returns Mac ML result on first call

fix: move @asynccontextmanager to lifespan; add watchlist + status/push commands

feat: HF Dataset queue — Mac worker pushes fresh predictions to HF Space

feat(C12): 1-day horizon; FAILED

feat(C11): full ensemble walk-forward; FAILED — RF already representative of ensemble

feat(C10): asymmetric barrier grid search; FAILED

feat(C9): multi-stock training; FAILED

feat(C8): Optuna RF tuning; FAILED — fixed params already near-optimal

feat(C7): add calendar/seasonality features + 36m lookback; PASSED

feat(C2): add PTT sentiment scraper and backtest; FAILED — PTT rows=0, features not added to model

feat(C5): SHAP pruning — remove 11 zero-contribution features, +1pp metrics

feat(C6): add HMM regime detection — FAILED (improvement <1.5pp aggregate)

refactor: add improvement_harness.py (shared walk-forward engine), clean settings.local.json

feat(C4): add meta-labeling secondary classifier — PASSED (up_prec 51.7→55.8%)

feat(C3): add triple barrier labels — PASSED (dir_acc 31.4→41.2%, up_prec 43.0→51.7%)

feat(C1): add fracdiff features — FAILED validation, not added to FEATURE_COLUMNS

feat: add research survey scripts + docs/survey outputs for NotebookLM

feat: feature audit + 3-way backtest, update FEATURE_COLUMNS to winner subset

feat: Phase 4 — securities lending, block trades, large holder features

feat: add walk-forward backtest scripts and fix margin_flow neutral row

feat: add chip_score, MA120/240, margin flow, and TA distillation