Commit Graph
12 Commits
Author SHA1 Message Date
Kunthawat Greethong 7f175e5a05 feat(scoring): flexible source fallback (no board crash) + per-source calc audit (Q2, Q3)
Q2 flexible scoring: _fetch_with_cache now degrades instead of raising
DashboardError — a source that fails with no cached value returns {} so the
theme scorer drops that source's factors; a previously-good value is kept as
stale by the daily cache. Verified: all-sources-down still builds 13 themes.

Q3 per-source audit: new themes.factor_source_breakdown(fetched, theme) shows
per factor source/raw/normalized/weight/contribution; dashboard exposes
fetch_data + factor_sources; per-symbol modal renders symbolDetail.factor_sources
(e.g. retail: te_thailand ยอดขายปลีก -14.5 -> -1.0 x 0.7 = -0.7).

Suite 369 green; independent review passed: true.
Q1 (HAR for deferred sources) spike recorded: method works, REIC needs deeper
interaction; NBTC 403 likely unbpassable without a session.
2026-08-29 12:15:54 +07:00
Kunthawat Greethong 12b34929d7 feat(factor): add energy_irpc (IRPC net margin) as 2nd Thai refiner signal
- new energy_irpc collector parsing IRPC performance-highlights table
  (net profit/EBITDA/ROE margins, latest period 3M26: +10.27%)
- factor energy_irpc_net_margin (sign +1) wired into refining_energy/
  exploration/utilities, extending the energy theme beyond TOP
- scheduler job + dashboard fetch + sources table row (now 9 sources)
- tests: parse (incl paren-negatives), value-key resolution, direction;
  suite 368 OK. Independent review passed: true
- Phase B feasibility: REIC/EPPO/NBTC/PTTEP are JS-rendered or anti-bot
  (recorded deferred in plan); IRPC was the clean server-rendered win
2026-08-29 11:13:17 +07:00
Kunthawat Greethong fcc0da9c8d feat(factor): te_thailand rate/credit/retail/property/confidence + thai_trade external sector; fix sign inversion on bearish factors
- add te_thailand collector (TradingEconomics) -> 8 factors: interest rate,
  business loan growth, consumer credit, household debt/GDP, retail sales YoY,
  consumer confidence, residential property prices, business confidence;
  feed banks/retail/consumer_staples/nonbank_finance/property/telecom/healthcare
- add thai_trade collector (TradingEconomics external sector) -> exports/
  imports/current-account factors (concurrent in-tree work, verified green)
- fix sign inversion: theme weights were negative on sign:-1 factors (NPL,
  inflation, unemployment) so higher NPL/inflation RAISED scores; direction now
  lives only in factor sign, theme weights positive (regression-locked)
- tests: te_thailand parse+direction, value-key resolution contract, dashboard
  8-sources, scheduler vintage counts; suite 362 OK
2026-08-29 09:18:55 +07:00
Kunthawat Greethong 8db3d48ae2 [verified] P0-B registry-driven scoring + P3 PIT backtest + P4 factor-weight learning
P0-B (registry is the single source of truth for scoring):
- FACTORS now carries center/span normalization spec; unused hand-written
  per-theme surprise blocks in dashboard.py replaced by one registry-driven
  compute_theme_surprises() (themes.py).
- THEMES['banks'] adds bank_npl weight so NPL is genuinely blended.
- factor_value/normalize hardened against NaN/inf (finite guards).
- Board re-ranks (TRUE/GULF up, TOP->3) per registry weights; 3 new tests
  incl. 'changing a registry weight changes output'.

P3 (point-in-time backtest):
- run_backtest is now a real multi-rebalance engine (reallocates every window,
  reconciles holdings, marks to market) instead of allocate-once+break.
- Added leakage_guard (False unless a PIT score_fn is supplied), planned vs
  actual rebalances, and momentum_at() true 12-1 (skips last month, PIT).

P4 (factor-weight learning):
- weight_learning.py: cross-sectional Spearman IC, forward-return builder,
  IC aggregation + t-stat, and apply_weight_update (new = clip(old*(1+shrink*IC))).
- GET /api/v1/learning/momentum endpoint. Live result: momentum IC=0.012
  t=0.132 over 22 periods -> momentum has no reliable predictive power here.
  Macro/demographic factors blocked (no historical factor vintages yet).

Two independent review gates passed (deleg_fe6f45cd, deleg_718218f8): empty
security/logic arrays; their non-blocking suggestions applied (finite guards,
dedupe leakage_guard resolution). 226 tests pass; Vite build passes.
2026-08-27 07:12:18 +07:00
Kunthawat Greethong 325e164dd3 [verified] Fix P1-P2-P5 audit findings: simulation reuses board, source_summary clarity, dead-code removal + conftest
- P1: /api/v1/simulation now uses the canonical board score (default_scores)
  instead of a divergent 3-theme recompute -> 'จำลอง' can't disagree with board
  (live check: sim top pick PTT == top board combined 1.600). Removes binary
  auto/en signs, restores quality+momentum+dividend screen consistency.
- P2: dashboard emits source_summary{factor_keys, rows}; frontend shows
  'N ปัจจัย · M แหล่ง' so the 7-vs-5 count confusion is impossible.
- P5: removed dead themes.list_themes()/Theme/build_theme_scores/_map_index and
  the tests that locked them; added tests/conftest.py so pytest needs no PYTHONPATH.
- docs: audit-and-plan-2026-08-26.md (full P0-P5 plan) + engineering-log entry.
- 203 backend tests pass; Vite build passes. Independent reviewer: no security or
  logic blockers (minor error-leak suggestion applied: 503 message no longer leaks
  exception detail).
2026-08-27 03:21:22 +07:00
Kunthawat Greethong 6e78b6acb5 [verified] Apply R1-R5 (factor formula) + real bank-sector NPL collector
(a) R1-R5 (factor-refinement, grounded in methodology-research.md):
- R1 (PEAD): EPS-growth weight raised 1.0->1.5 in build_siamchart_score / symbol_breakdown (Bernard-Thomas 1990, Livnat-Mendenhall 2006)
- R2 (momentum): 12-1 momentum factor from Yahoo price snapshot (Jegadeesh-Titman 93; lite weight 0.5)
- R3 (regime): binary bear gate -> continuous stress = negative-themes fraction, smooth LONG/SHORT shift
- R5 (dividend screen): non-dividend / cut-yield names no longer go LONG (screen-off)
- R4 (earnings-revision) deferred: no free EPS-forecast source yet (documented)

(b) bank-sector NPL collector (BOT reportID 794, financial&insurance sector):
- refactored auto_npl to expose shared _parse_sector; new bank_npl.py reuses it
- registered bank_npl FACTOR -> auto-appears in sources table (6 rows) + blends into banks theme surprise (real NPL)
- +unit tests (test_bank_npl), test_dashboard updated (6 sources)

205 tests pass; verified live API (banks surprise incl. NPL 1.07, 6 sources).
2026-08-26 19:56:39 +07:00
Kunthawat Greethong fc592d8aa9 [verified] Real backtest engine + backtest UI section (start/end dates, P&L, persisted)
- backtest.py: buy-and-hold backtest over [start,end] — allocates 50/20/30 at first available rebalance date, marks to market to end, accrues dividend, reports {final_value, price_pnl, dividend_income, net_return, trades, holdings}
- Fixed double-spend bug (was allocating full capital every rebalance -> negative cash)
- dashboard.default_scores(): per-symbol combined/dividend/yield baseline for backtest
- POST /api/v1/backtest + GET /api/v1/backtest/runs (results persisted in app state -> survive refresh)
- Frontend: backtest section w/ start/end/capital/freq inputs + P&L KPIs + run history table
- Honest note: uses current combined scores as static baseline (non-PIT); PIT score_fn pluggable
- Verified: 1M -> 1.088M (+8.80%) over 2024-06..2026-06; history persists across refresh
2026-08-26 16:00:17 +07:00
Kunthawat Greethong 2da73b8a8c [verified] Sources table auto-derived from FACTORS registry + next-update column
- _build_sources now iterates FACTORS registry (grouped by fetch module) instead of hardcoded 5-row list -> adding a FACTOR auto-appends its source row (single source of truth, answers 'เพิ่มแหล่งควรอัตโนมัติ')
- Each source row gains ความถี่ + อัปเดตครั้งต่อไป (from frequency cadence)
- Frontend sources table: 6 columns (ข้อมูล/แหล่ง/ช่วงข้อมูล/ความถี่/อัปเดตครั้งต่อไป/อัปเดตล่าสุด), all from /api/v1/dashboard (not mock)
- Verified live: 5 sources auto-derived w/ next-update; browser shows 6 cols
2026-08-26 15:55:30 +07:00
Kunthawat Greethong d850955c44 [verified] Declarative factor engine + per-symbol stock selection (full-app consistency)
- factors.py: FACTORS registry (10 declarative entries: source/fetch/frequency/sign/weight) + normalize/z-score helpers. Add a source = one dict entry, no scoring-function edit.
- themes.THEMES: 13 themes reference FACTORS with per-theme weights (flexible), replacing hardcoded _theme_surprises/_theme_narrative.
- themes.quality_within_theme(): per-symbol quality vs theme cohort (ROE/EPS) -> real stock picking. dashboard board now surprise×quality (BBL 0.5 vs KTB 1.5 in banks).
- board rows carry per-symbol themes[]; /api/v1/themes delegates to RealDashboard.build() -> 13-theme consistency with /api/v1/dashboard (removed 115 lines dead dup logic).
- frontend: deleted THEME_BY_SYMBOL/themeLabelById hardcode; theme column + modal labels+quality all from API. Modal shows surprise×quality=theme_score.
- Tests: 202 OK (quality selection, breakdown quality, themes/dashboard consistency).
- Verified: BBL modal 1.00σ×0.5=0.50σ; KTB 1.5 vs BBL 0.5, PTT 2 themes; 49/49 rows theme from API.
2026-08-26 15:17:41 +07:00
Kunthawat Greethong e7819a35dd [verified] All 13 SET50 themes now have real surprise (macro-proxy) — no more 'ยังไม่มีข้อมูล'
- dashboard._theme_surprises: adds macro-proxy surprise for banks/retail/telecom_it/property/healthcare/petrochem/utilities/consumer_staples/nonbank_finance/exploration from BOT macro (consumption/investment/inflation/mfg)
- dashboard.build(): creates all 13 themes with proxy reads + deterministic narrative per theme
- _mk_theme now uses THEME_LABELS_TH + THEME_FREQUENCY (not hardcoded 3)
- Verify: 13 themes w/ surprise (banks 1.0, retail 0.19, utilities -0.31, exploration 1.62); BBL modal (ธนาคาร 1.00σ, combined 0.467)
- Full suite 199 OK; fixed test_build themes=13
2026-08-26 14:19:59 +07:00
Kunthawat Greethong 33a4662cd4 [verified] Add per-symbol analysis breakdown endpoint + per-theme narrative
- themes.symbol_breakdown(): transparent scoring derivation (theme_score, siamchart_score components, combined = 0.6*theme + 0.4*siamchart npolut)
- GET /api/v1/symbols/<symbol>: themes + theme surprise contributions + fundamentals + price + weights (ข้อ 7)
- dashboard.py _theme_narrative(): long-form Thai explanation of each theme's analysis outcome + implication for its stocks (ข้อ 5)
- 2 tests; full suite OK; live verified (AOT: combined 0.107 = 0.6*0.571 + 0.4*(-0.588))
2026-08-25 21:18:20 +07:00
Kunthawat Greethong 9739849f68 [verified] Add real multi-theme dashboard (3 themes + macro + board + sources) — req #6/#8/#9/#10
- dashboard.py: RealDashboard assembles real Thai data (tourism + auto+NPL + energy TOP + macro BOT) with uniform z-score surprise per theme, per-theme thesis, sources provenance table, 49-symbol combined board
- macro_thai.py: BOT Thai Economy macro backdrop (consumption +4.9%, inflation 1.95%, unemployment 0.93%, tourists 16.2mn)
- GET /api/v1/dashboard endpoint (real data, no fixture fallback per user)
- 7 new tests; full suite 195 OK; live verified (3 theme surprise: 0.571/0.81/1.623)
2026-08-25 20:30:14 +07:00