Skip to content

P0 Evidence: licensed historical source rights, original vintages and TBSP income-index methodology for research #116

Open

Description

@gniewkob

Cel

Udokumentować rzeczywiste, niezależne dowody dla kwalifikacji historycznych źródeł badawczych M1–M3. Zielone CI Draft I1–I7 to wyłącznie weryfikacja oprogramowania na danych syntetycznych, nie prawo do automatycznego pobierania danych, niezależny historic PIT ani dowód uzyskanych stóp zwrotu.

Powiązane: I7 Draft #115, I5 Draft #113, I6 Draft #114, istniejące M1 source #97. I7 zweryfikował wzorzec NBP dirYYYY.txt bez pozyskiwania prawdziwych notowań.

P0 — trzy niezależne osie dowodów (właściciel danych / niezależny reviewer)

  • Source identity / methodology: oficjalny identyfikator indeksu, dokładna waluta, price/NAV/total-return, data wejścia metodologii w życie i ewentualna zmiana administratora. Nie łączyć indeksu dochodowego z ceną transakcyjną.
  • Rights / entitlement: umowa, licencja, regulamin albo pisemne potwierdzenie upoważnionego dostawcy obejmujące dokładnie: dataset, kanał pobrania, przechowywanie, przetwarzanie analityczne, automatyzację, funkcje AI i dystrybucję. Sam publiczny URL czy pomyślny download nie wystarcza.
  • Historical PIT / original vintage: dla każdej obserwacji first_available_at, identyfikator i skrót oryginalnej wersji, sposób potwierdzenia historycznej dostępności przed decision_at, ścieżka rewizji oraz obsługa opóźnionych publikacji. Gdy brak dowodu, zachować HOLD_CASH, nie przypisywać domyślnego historycznego czasu publikacji.

P0 — dostawcy / źródła pierwotne

  • TBSP.Index / PL_BONDS_TR_PLN: BondSpot 2019, §3.1 opisuje benchmark dochodowy (zmiany cen, narosłe odsetki, reinwestowane kupony), lecz jego wartości nie są ceną wykonania. BondSpot wskazuje GPW Benchmark jako administratora od 2019-12-01. Potrzebne: aktualna metodologia GPW Benchmark, uprawniona historia i dokument dopuszczalnych zastosowań/licencja oraz daty wersji i publikacji.
  • MSCI ACWI NETR / GLOBAL_EQUITY_TR_PLN: MSCI Terms of Use, §2.1/2.4, aktualizacja 2026-07-15, wyraźnie ograniczają nieautoryzowane boty, AI agents i MCP. Nie zbierać automatycznie bez właściwego upoważnienia. Poprzedni pięciokolumnowy eksport nie ma kolumny waluty; wymagane źródłowe potwierdzenie waluty i basis, a nie wnioskowanie z NETR lub dopasowania dat USD/PLN. Rozważyć alternatywne źródło z jasną licencją i równoważną semantyką, nie kopię obchodzącą ograniczenia.
  • Vanguard ETF NAV (alternatywny benchmark, nie MSCI ACWI): Vanguard Terms, User license ogranicza wykorzystanie danych, a strona ETF osobno pokazuje NAV i ceny rynkowe. Potrzebne: nazwa/ISIN, waluta, dokładny zakres praw do danych, rozdzielenie NAV od cen giełdowych i faktyczna historia dostępności.
  • NBP USD/PLN, gold: API daje EffectiveDate, instrukcja archiwum daje dirYYYY.txt, terminy publikacji opisują obecny kalendarz. I6 i I7 weryfikują lokalną spójność, nie autentyczność ani historyczną godzinę. Potrzebny dowód PIT oraz ocena zasad użycia danych w zakresie projektu. Kurs FX jest referencyjny; złoto NBP ma semantykę PRICE_ONLY.

P1 — proces odbioru

  • Review źródeł przez właściciela danych (identyfikator + prawa + zakres wykorzystania) i review PIT przez niezależną osobę/technikę.
  • Bezpieczna ewidencja dokumentów dowodowych (minimum metadanych + hash, bez wrzucania licencjonowanych szeregów, sekretów i prywatnych umów do GitHub).
  • Pary okres observation_date / first_available_at, wersjonowanie i test historycznej chronologii na rzeczywistych uprawnionych danych; nie dawać zgody tylko na podstawie syntetycznych fixture.
  • Dopiero po P0 planować nowy audyt backtestu M1–M3 w trybie edukacyjnym / symulacyjnym z kosztami, ujemnymi scenariuszami i sensitivity. Bez handlu, zleceń, portfela realnego ani transferów.

Acceptance definition: dla każdej rodziny GLOBAL_EQUITY_TR_PLN, PL_BONDS_TR_PLN, GOLD_TR_PLN osobno jawnie APPROVED_FOR_SPECIFIED_RESEARCH_SCOPE albo BLOCKED, z dokładnym odwołaniem do dowodu i recenzenta. Ogólny source_sha256 lub link publiczny nie jest certyfikacją. Istniejące I4/I5/G1 nadal fail-closed. main i środowiska produkcyjne nietknięte.

Activity

  1. gniewkob commented on Nov 10, 2026

    @gniewkob
    OwnerAuthor

    Decision record — real, permitted market-data sources (2026-11-10)

    Decision: do not add further synthetic qualification gates until we have an actionable, permission-aware source pilot. No real-data series is newly qualified by this review. Source readings are primary-provider disclosures, not signed rights or PIT certificates issued to this project.

    Verified facts, assumptions and route

    Source Provider's official statement Appropriate research role Critical gap / decision
    EODHD (candidate) Personal use FAQ, AI/cloud FAQ: a qualifying individual may privately analyse, backtest and automate data, including use of their own AI tools; no distribution to others. EOD API includes raw close and dividend/split-adjusted close, with varying market coverage. CONDITIONAL_PRIVATE_RESEARCH_PROXY for selected actually listed ETFs and time series only if account use/eligibility and exact feed rights permit it. Raw files, credentials and data-bearing derived reports must stay private, not in public GitHub/CI. Methodology FAQ explicitly states NO historical point-in-time vintages, overwrite on corrections, and backfilled/recomputed adjusted_close. Historic adjusted close downloaded today is not an independent as-of data series. Never imply MSCI ACWI index equivalence, true historical PIT, or executable total-return results. Plans FAQ: free plan is 1 year EOD history; >=36 months not guaranteed on free plan. No subscription/payment is authorized here.
    NBP FX/gold (official) NBP API offers public FX and gold data. NBP website regulation, §1(6) states NBP does not object to saving/copying site materials for personal or professional non-commercial use; §1(8) requires source/link attribution. §5(1) cautions against excessive automated load. REFERENCE_ONLY, potentially usable for bounded, attributed, non-commercial research after confirming these terms apply to the exact api.nbp.pl access channel. I6/I7 handle local chronology but not independent historical release hour. FX mid / NBP gold quote are not full investable total returns or execution prices. A current API response is not an immutable historical vintage. Whether website terms extend fully to API and our intended publication/cloud storage needs specific verification.
    GPW Benchmark / TBSP.Index BondSpot historical TBSP methodology, §3.1 documents an income benchmark with accrued interest/reinvested coupons. KNF list of benchmark administrators confirms GPW Benchmark's benchmark-administrator status and official contact. POLISH_BOND_TR_BENCHMARK_CANDIDATE, distinct from a tradeable bond ETF/security price. Need current index methodology, lawful rights to precisely named historical values, distribution/AI clauses, historical original publication and revision records; contact administrator instead of using scraped or re-hosted copies.
    MSCI ACWI NETR MSCI Terms §2.4: automated extraction and AI/MCP access without rights is restricted. HOLD_UNTIL_EXPLICIT_ENTITLEMENT. The available five-column research CSV does not declare currency and has no independent historic PIT or use rights. Never infer USD from matched NBP dates or try an unlicensed mirror.
    Vanguard historical NAV Vanguard Terms and ETF product show site-use restrictions and NAV vs market-price distinction. SEPARATE_NAV_REFERENCE_CANDIDATE under existing Issue #97. Existing export still not independently validated. NAV is not ETF exchange execution price; no silent substitution of GLOBAL_EQUITY_TR_PLN.

    Important economic-methodology caveat: adjusted EOD prices

    EODHD's own methodology FAQ states adjusted_close is recalculated across the whole history when new dividend/split information arrives and corrections can overwrite bars without version history. It would create historical look-ahead/revision leakage if we simply use the latest adjusted series and label the result PIT_BACKTEST. A current historic series is at most RETROSPECTIVE_RECONSTRUCTION_CURRENT_VINTAGE, unless independent period-specific evidence exists. Prospective daily captured snapshots can establish our own first_observed_at, but cannot reconstruct what the vendor published before those captures.

    Do not treat provider's personal-use FAQ as user-specific entitlement: product availability, plan eligibility, audience, cloud/AI processing and access must be resolved first. No provider credentials/raw licensed data may enter GitHub, issue comments, public actions logs or shared artifacts.

    Concrete next work, in the order that unlocks evidence (without buying anything)

    1. Rights & usage scope: confirm personal, one-user private research scope and any account-specific restrictions for a candidate EODHD feed. Separate any commercial/shared or published use; do not use a personal entitlement for the latter.
    2. Coverage discovery, metadata only: check whether exact ETFs/securities representing global equities, Polish bonds and investable gold are supported, each instrument's ISIN/exchange/trading currency/inception/share class and actual 36–72 completed monthly observations. No benchmark substitution without separate label.
    3. Reproducible returns: distinguish close, adjusted_close, paid dividends, splits, accumulating/distributing fund structure, FX conversion using eligible NBP dates, costs and whether adjustments are latest-vintage. No PIT_CERTIFIED claim.
    4. Prospective snapshot pipeline design: local/private, source bytes+timestamp+hash+retrieval log (never Git), idempotent immutable new captures, detection of later revisions; only run with proper rights and no access tokens in PR/CI. First observed is not independently first published.
    5. TBSP approach: obtain current GPW Benchmark methodological/rights confirmation for index level history; retain benchmark separate from investable instrument.
    6. Research decision: if exact candidate feed and terms do not satisfy the scope, stop provider evaluation rather than fabricate/mirror/scrape; maintain models INSUFFICIENT_DATA/HOLD_CASH.

    Approval and integration

    No real trades, orders, wallet/broker/asset transfer, paid subscription, source download, provider activation, scheduling, Neon/cloud write or main merge was authorized or done in this research. I4/I5/I6/I7 and G1 gates remain fail-closed; no synthetic test validates live market-data rights. Status: SOURCE_SELECTION_RESEARCH_IN_PROGRESS; QUALIFIED_SERIES=0.

  2. gniewkob commented on Nov 10, 2026

    @gniewkob
    OwnerAuthor

    Stage 2 source qualification: open macro/reference lane vs investable total-return (2026-11-10)

    Draft PR #118 now includes docs/EODHD_STAGE0_RIGHTS_INSTRUMENTS_PIT_DECISION_20261110_PL.md §7 as a source-backed, documentation-only alternative-source decision. Exact SHA d3559325cd76a591e7b4cadb4a35668ea67e13d4. No supplier API calls, credentials, data downloads, purchases, executable integrations, new schedules or external DB writes.

    Positive evidence that does not qualify TR

    • OECD share prices are PRICE_INDEX, excluding reinvested dividends. Official OECD Financial Market dataset identifies the SDMX dataflow OECD.SDD.STES,DSD_STES@DF_FINMARK. OECD Terms §3 permit broad reuse where no third-party conditions override, but require examining per-series ownership and restrictions. Country share-price index ≠ global ETF return.
    • ECB policy, statistics reuse allows free reuse of publicly available ESCB statistics with citation, subject to originals/third-party and revisions. FX and interest rates are research-context inputs, not *_TR_PLN instruments.
    • Eurostat copyright allows attributed reuse with stated exceptions, but the Eurostat API introduction says only latest datasets are served, not vintage versions. CPI/inflation is macro context only.
    • FRED API does not supersede third-party rights. No verified official Stooq extraction/redistribution authorisation; no implicit permission from a CSV endpoint.

    Concrete gate structure

    1. Source catalog: exact dataflow_id, series_key, measure_type, publisher/third-party owner, rights scope, attribution, frequency, revision policy and first observed timestamp before any new automated ingestion.
    2. Separate MACRO_CONTEXT_REFERENCE namespace; cannot alias GLOBAL_EQUITY_TR_PLN, PL_BONDS_TR_PLN, GOLD_TR_PLN or unblock C1/A1.
    3. Original TBSP total-return benchmark rights, ETF cashflow adjustments, executable prices and independent PIT evidence remain P0 blockers; no actual strategy edge confirmed.

    Decision: OPEN_MACRO_RESEARCH_FEASIBLE_CONDITIONALLY; REAL_TOTAL_RETURN_SERIES=0; LICENSED_ETF_HISTORY=0; NO_EODHD_INGEST; HOLD_CASH.

    No new issue or competing Draft PR was opened for this review; the Stage 2 evidence is contained in existing #118 and #116.

  3. gniewkob commented on Nov 10, 2026

    @gniewkob
    OwnerAuthor

    Stage 3 — P0 exact OECD series rights/metadata gate (2026-11-10)

    Source-selection record: Draft PR #118, updated research doc §8, pinned HEAD 4e76aa8119fbe7af3c08ebf42e54189098f2570a. No code, provider API request, prices, financial operations, or schedules were added.

    Precisely identified candidate for independent MACRO_CONTEXT_REFERENCE only:

    • OECD Share prices indicator definition: PRICE INDEX, not dividend-reinvested total-return, monthly generally an arithmetic average of daily prices, not month-end execution close.
    • Official OECD Financial market dataflow: OECD.SDD.STES,DSD_STES@DF_FINMARK,4.0.
    • Series dimensions as independently displayed on FRED's OECD series metadata page: REF_AREA=POL, FREQ=M, MEASURE=SHARE, UNIT_MEASURE=IX, ACTIVITY=_Z, ADJUSTMENT=_Z, TRANSFORMATION=_Z, TIME_HORIZ=_Z, METHODOLOGY=N. Candidate SDMX key POL.M.SHARE.IX._Z._Z._Z._Z.N is constructed from those nine filters but not directly OECD-API-tested; do not claim working endpoint/coverage.

    Blocking rights constraints, with exact source:

    1. OECD Terms §3 Data / Permitted Use explicitly requires the user to examine series Source metadata and secure permission for third-party-owned content. The underlying original Polish equity index owner / exact rights for use in ISL are not independently established. OECD portal visibility does not clear them.
    2. FRED Terms §II Prohibited Use / FRED API restrictions (k)–(l) restrict AI-related development/use of its services/content/API, and FRED marks the relevant series Copyrighted: Citation Required. FRED is a metadata corroboration link, NOT an authorised alternative automated data source, and must not be integrated into the AI/agent workflow.
    3. No verified historically frozen first_available_at for each observation, official revision archive or authenticated real data sample. A current monthly price index alone is not GLOBAL_EQUITY_TR_PLN.

    Definition of next evidence-based milestone — no new parser first:

    • Verify exact POL.M.SHARE.IX._Z._Z._Z._Z.N selection in the official OECD Data Explorer/SDMX structure; capture its true per-series SOURCE, rights notice and upstream compiler identity.
    • If rights unclear, obtain permission clarification from the official OECD Financial market statistics contact stat.contact@oecd.org, referring to full dataflow/series key, use for one-user private AI-assisted research, retention and derived results. No email sent; ask for permission and any necessary original-owner contact.
    • Only after permitted use is documented, evaluate a bounded official OECD observation sample for coverage/revisions and define explicit MACRO_CONTEXT_REFERENCE storage with citation and no promotion into A1/C1 total-return source inputs.
    • Until then DIRECT_SDMX_RESPONSE=UNVERIFIED; ORIGINAL_OWNER_RIGHTS=UNVERIFIED; HISTORICAL_PIT=UNVERIFIED; PERMITTED_SOURCE_INGEST=NO_GO; QUALIFIED_TR_SERIES=0; M1–M3=HOLD_CASH.

    Stage 3 decision: Exact macro series identification has advanced materially, but legal/source acceptance remains blocked. No market or broker API integrations, orders, account actions, real trades/transfers or changes to main.

  4. gniewkob commented on Nov 10, 2026

    @gniewkob
    OwnerAuthor

    2026-11-10 — P0 discovery: OECD Poland series may not be price-only despite OECD generic label

    New finding, source-backed: OECD definition of Share prices explains national price indices vs total-return indices. But GUS official WIG definition states WIG is a total-return index that includes dividends and subscription rights, calculated from the 1991 opening session, while GUS WIG20 definition defines WIG20 as a price index, initiated in 1994. The selected OECD Poland monthly SHARE series secondary FRED metadata starts May 1991.

    This is an identity/semantics ambiguity, not proof OECD uses WIG. The original OECD SOURCE metadata and underlying Polish compiler were not independently verified. Possible scenarios WIG rebasing, separate price-only original, or splicing/methodology change are hypotheses, not accepted facts. An attempted direct OECD SDMX read could not be processed by the available web tool, so do not report the constructed POL.M.SHARE.IX._Z._Z._Z._Z.N as HTTP-verified or the feed as invalid.

    Defensive source gate: the POL series must not enter even MACRO_CONTEXT_REFERENCE production until we know exactly which primary index/algorithm was used; it must never be silently labelled GLOBAL_EQUITY_TR_PLN nor used to prove profits or signal live investment. Current QUALIFIED_TR_SERIES=0, all M1–M3 remain INSUFFICIENT_DATA/HOLD_CASH.

    Completed action: Updated Draft PR #118 Stage 4 at pinned head 49766d2d564f5cefe80adf7e584586012ab2d9da with explicit GUS/OECD source comparison and a ready-to-send, not sent, four-question clarification email to the OECD dataset contact stat.contact@oecd.org. Questions cover (1) original index/compiler/price-vs-TR and exact rights holder, (2) methodology changes/splicing since 1991, (3) exact SDMX key + source/revision/PIT metadata, and (4) permissible one-user private AI-assisted research/retention and any third-party licences.

    Issue #116 P0 acceptance additions:

    • Obtain original SOURCE field/primary compiler and official description for exact Poland SHARE SDMX series.
    • Verify historical PRICE vs TR category and any series changes from 1991; do not infer WIG solely from start date.
    • Confirm third-party licensing, automation/AI/retention and citation conditions with data owner.
    • Confirm direct official SDMX response and vintage/metadata after rights check.
    • Then and only then consider an isolated, bounded official macro-context pilot (not trading).

    No email was sent, no API source values were downloaded, no Git main/CI/provider security changes, no account registration/paid rights, no financial operations.

  5. gniewkob commented on Nov 10, 2026

    @gniewkob
    OwnerAuthor

    OECD source clarification — email status and next evidence gate (2026-11-10)

    Gmail verification: searched messages to:stat.contact@oecd.org, from:stat.contact@oecd.org, and OECD-related drafts. No existing sent inquiry, reply or draft was found in the connected Gmail search at this check.

    Outbound communication: an attempt to send the already documented source/methodology inquiry via Gmail required additional user authorization and was declined by the connected service. EMAIL_NOT_SENT; no resend or bypass attempted. The complete four-question text remains available in Draft PR #118, Stage 4. The OECD data owner/methodology is therefore still UNVERIFIED.

    Additional public evidence: historical WIG20 index methodology says WIG20 is a price index with dividend income excluded, consistent with the prior GUS glossary WIG20. GUS WIG definition treats WIG as an income index. This does not identify which index OECD's Poland SHARE series uses or clear rights/PIT. The secondary FRED documentation monthly POL SHARE confirms the field structure only and must not be used as an ingestion route.

    Decision: OECD_INQUIRY_UNSENT_AUTHORIZATION_DECLINED; OWNER=UNVERIFIED; SEMANTICS=UNVERIFIED; INGEST=BLOCKED; QUALIFIED_TR_SERIES=0; HOLD_CASH.

    Next unblocked action: arrange direct, user-authorized clarification with OECD or GPW Benchmark using official channels; request underlying index compiler, full historic methodology/rewisions and reuse permissions for this exact series. No real price feeds, AI or rehosted CSV scraping, source-licence inference, account operations, payments or trades.

  6. gniewkob commented on Nov 10, 2026

    @gniewkob
    OwnerAuthor

    P0 source discovery — two explicitly licensed historical TOTAL-RETURN research datasets (2026-11-10)

    Meaningful change in evidence: rather than repeatedly constructing synthetic proxy source gates or conflating prices/NAV with total returns, identified two publisher-hosted primary datasets with explicit publication licences and actual historical equity/bond return semantics. No actual files have yet been fetched or audited. Neither meets the 2021-11–2026-10 monthly Poland/world-equities/gold target; they are historical methodology-validation baselines, not an automatic M1–M3 source unblock.

    A. JST Macrohistory R6 — official CC BY-NC-SA 4.0, annual asset TOTAL RETURNS

    • Publisher/primary authority: MacroFinance & MacroHistory Lab — JST Database R6, full Licence Terms. Official § 'Rates of Return' specifies equity total return (capital gain + dividends), government bond total return, government bill and other asset returns. 18 developed economies, annual from 1870, latest R6 updated to ~2020, not Poland. All nominal returns are local currency.
    • Licence: official Terms of Use explicitly adopt Creative Commons Attribution–NonCommercial–ShareAlike 4.0 International; Licence §2(a)(1) grants reproduction/adaptation/sharing for NonCommercial purposes; §3(a) requires attribution, indication of changes and links; §3(b) ShareAlike for distributed adapted material. Commercial data services / reselling excluded. Cite Jordà, Schularick, Taylor (2017) generally and Jordà, Knoll, Kuvshinov, Schularick, Taylor (2019) for asset returns. Publisher advises documenting access date because statistics may be revised.
    • Classification: HISTORIC_ANNUAL_TOTAL_RETURN_RESEARCH_REFERENCE. This licence is verified for an eligible noncommercial research scope, but the XLSX/DTA source file, columns, row coverage, source SHA, quality and historical release vintages are not verified. Not a monthly current proxy; no Poland, no 2021–2026 monthly periods, no direct tradable execution prices.

    B. Swinkels / Erasmus University Rotterdam — official CC BY 4.0, monthly government-bond returns

    • Publisher's original dataset: Erasmus — Historical Data: International monthly government bond returns, DOI 10.25397/eur.8152748.v6. Published 2023; series updated through December 2022. Monthly estimated investor bond holding-period returns based on YTM and documented bond-pricing formula, not raw yield and not exchange-traded fund close. Countries: US, Germany, Japan, Australia, France, Canada, UK, Norway, Sweden; no Poland.
    • Licence: original provider labels the dataset Creative Commons Attribution 4.0 International (dataset licence, not a licence inferred from an unofficial copy). Author's methodological paper / dataset source explains approximations and comparability caveats. Dataset distribution/citation must preserve CC BY attribution; verify original download file + any file-embedded notices before using.
    • Classification: HISTORIC_MONTHLY_GOVT_BOND_RETURN_METHOD_VALIDATION. Good candidate for a real source-backed, offline 36-month educational regression of monthly return calculations over an older period, not PL_BONDS_TR_PLN and not 2026 source/PIT certification. Full original workbook has not been downloaded, parsed or SHA-pinned by this review.

    C. Kenneth R. French — dividend-inclusive USD developed-equity returns, rights still unclear

    • Dartmouth original Data Library, Developed portfolio construction note explicitly say market-factor series are monthly from 1990, USD, include dividends and capital gains. Mkt-RF is excess return and needs same-source RF addition to describe market return; it is not a world ETF or MSCI ACWI NETR. Publisher states full historical return files are reconstructed on updates, subject to revisions; present export is NOT a historical-vintage snapshot. US format changes FIZ→CIZ in 2025 are flagged and must not be generalized without checking the exact Developed series.
    • Important legal gap: no clear independent reuse/automation/AI/redistribution licence was found on the source pages. Public downloadable ZIPs are not proof of rights. RIGHTS_UNVERIFIED / NO_AUTOMATED_INGEST; ask source owner before agent ingestion, avoid mirrored ZIPs.

    Exact decisions

    Candidate Semantics independently supported? Publisher licence? Monthly contemporary target? Use
    JST R6 YES, annual equities + gov bonds total returns YES CC BY-NC-SA 4.0 for eligible noncommercial scope NO historical annual independent methodology check only
    Erasmus Swinkels YES, monthly modeled bond investor returns YES CC BY 4.0 NO — only up to Dec 2022, not PL prioritized monthly historic analytical benchmark to validate source handling, after original-file acceptance
    Ken French Developed YES, dividend-inclusive monthly USD factor, not ETF UNVERIFIED nominal time coverage YES, rights/vintage NO NO INGEST, exact permission/data method review required

    Required next physical evidence (no new synthetic adapter): obtain one permitted ORIGINAL source file from its academic publisher using the stated licence; record URL, licence, original filename, local SHA-256, actual format/columns, unit, country and observed window; ensure no unauthorized third-party or AI-use terms in the file. Keep data and unreviewed derived values out of public GitHub/CI and all brokerage/wallet systems. Use the pre-existing offline scientific review process before any new parser. Actual file not obtained or tested yet. See focused implementation issue to follow.

    Status: TWO_LICENSED_HISTORICAL_RETURN_SOURCE_FAMILIES_IDENTIFIED; REAL_ORIGINAL_FILES_VERIFIED=0; QUALIFIED_M1_M2_M3_SERIES=0; HOLD_CASH. All Draft PRs unmerged, financial operations 0.

  7. gniewkob commented on Nov 11, 2026

    @gniewkob
    OwnerAuthor

    2026-11-11 evidence checkpoint — P0 still OPEN

    All M1–M3 research engine/backtest PRs closed via original or clean integrations #125–#129, final main 7ffc43e23d0f4376bcf2c4ea6860c1f88d8772fc, offline regression 43085587408 1,237 tests PASS, one skipped. Original DataverseNL Swinkels CC BY ODS/README and 360 historical modeled bond-return observations validated in merged #128, but non-Polish and historical through 2022, not eligible PLN 2021–2026 M1–M3 source. P0 still requires exact data-owner entitlements and permitted AI/automation uses, current TBSP methodology, properly typed global equity and bond TR, original point-in-time vintages and first-available dates. ALFRED PIT remains BLOCKED_NO_FRED_API_KEY. No assumed rights, inferred release timestamps, model-source promotion or financial operations. qualified_model_series=[], INSUFFICIENT_DATA/HOLD_CASH.

  8. added a commit that references this issue on Nov 11, 2026
  9. gniewkob commented on Nov 11, 2026

    @gniewkob
    OwnerAuthor

    2026-11-11 — precise source-rights decision, original-file evidence reconciled

    Engineering baseline updated: All post-merge P2 fixes #134–#137 are merged; final main 3f809b5016f63c527ad14d19276e156e6d9e4d2e, offline regression #43091417131 1247 tests run, one skipped, result SUCCESS. This closes no owner entitlement/PIT gate.

    Actual private Vanguard NAV data should not be counted as "missing altogether." Original XLSX SHA eadddaf00d6c9ac14fefb3a396ffb521cb82bbaa1d1b71289c650060712e64, actual file layout 1886 entries, 60/60 monthly USD NAV observations and exact-day NBP conversion were operator-tested in #97 receipt; current freeze remains intentionally PENDING until the first complete post-freeze November month. This is existing original-file observation evidence, not third-party identity, permission for automated reuse, benchmark TR, or historic publication vintages.

    Verified first-party site/legal constraints:

    • Vanguard UK Professional Terms, "User license": limited personal, informational, non-commercial use or express written authorization; restrictions on reuse/distribution and repeated automated access. This is the UK Professional website's terms, not automatically proof of the exact Polish retail download licence or of a contract held by this owner. Obtain the applicable actual source-channel terms and assess private AI-assisted analytics, storage, derived statistics and automation separately.
    • MSCI Terms, §§2.1 and 2.4, revised 15 July 2026: absent an Order Form, use only internally for non-production purposes; restrictions include unauthorized AI agents, bots, MCP extractors and redistribution. A public index export alone is not a general authorization to build our data pipeline.
    • BondSpot official TBSP.Index description: confirms GPW Benchmark S.A. responsible for index provision since 1 Dec 2019 and directs users to its current index method/rules. Confirms administrator identity, not a data redistribution licence, precise vintage calendar, or executable asset prices.
    • Erasmus University source states explicit CC BY 4.0 for Swinkels modeled government-bond data; it covers 9 foreign countries through 2022, not current PLN bonds. A current Dataverse original ODS is SHA-pinned; the mapping from Erasmus v6 to Dataverse remains unverified in P0 REAL DATA: audit original CC-BY monthly government-bond return workbook (Erasmus Swinkels) offline #123.

    Actual external evidence still missing: dataset-specific permission/entitlement and AI/automated use decision per candidate, original published historical vintages with first_available_at and revision history, current GPW Benchmark TBSP method + release schedule. The independent owner must review provider terms and underlying data-source identity; absence of such approval is not a software error.

    Disposition by required model input:
    GLOBAL_EQUITY_TR_PLN=BLOCKED_RIGHTS_PIT_SEMANTICS;
    PL_BONDS_TR_PLN=BLOCKED_RIGHTS_CURRENT_METHOD_PIT;
    GOLD_TR_PLN=BLOCKED_REAL_TOTAL_RETURN_SOURCE.
    NBP FX and gold PRICE_ONLY remain descriptive public reference inputs, not total-return indexes. All M1–M3 INSUFFICIENT_DATA/HOLD_CASH. No live trading, financial operations, provider automation, downloads bypassing access controls, or external DB writes.

  10. gniewkob commented on Nov 11, 2026

    @gniewkob
    OwnerAuthor

    2026-11-11 — VERIFIED applicable POLISH retail website Terms (source-channel rights correction)

    Exact first-party primary source: Vanguard Poland Individual Investors — Warunki i zasady, section “Ograniczone licencje i ograniczenia dotyczące użytkowania materiałów”, and scope preamble. The publisher is Vanguard Group (Ireland) Limited, and these terms explicitly govern this Polish retail website itself (the original NAV XLSX acquisition channel), not the separate UK Professional website previously cited.

    What the Polish section actually says: it grants a limited, revocable, non-exclusive, non-transferable licence to browse/store/bookmark/download/print website pages for personal, informational, noncommercial purposes or as explicitly authorized in writing. Unless otherwise authorized it restricts reproduction, derivative use, transfer/distribution of site-generated data/material, and repeated automated access to the website via third-party tools. It also identifies intellectualproperty@vanguard.com as a possible written-permission channel. Do not treat the website's terms as an automatically acquired API/data-reuse contract.

    Impact on this project's private evidence: The operator's actual retail XLSX original SHA eadddaf00d6c9ac14fefb3a396ffb521cb82bbaa1d1b71289c650060712e64 and verified 60/60 local month/FX conversion are genuine technical source-ingest evidence but do not themselves authorize public dissemination, shared AI-agent data loading, persistent automated collection, licensed third-party redistribution, or qualification of an investable total-return model. Local/private educational statistical calculations and further publisher-scope questions must be assessed under these exact terms and any additional terms that applied during XLSX export. No repetitive automated downloader should be activated without explicit applicable permission.

    Remaining review: save a sanitized rights decision referencing the exact Polish section and version/access date (no raw proprietary NAV rows), identify the intended private AI-assisted research processing boundary, and seek written clarification only if that scope is not covered. Independent NAV historic first_available_at vintages and first post-November freeze observation remain separate open checks. SOURCE_RIGHTS_FOR_AI_AUTOMATION=NOT_APPROVED, M1-M3=INSUFFICIENT_DATA/HOLD_CASH, financial operations 0.

  11. gniewkob commented on Nov 12, 2026

    @gniewkob
    OwnerAuthor

    2026-11-12 — current TBSP.Index / GPW Benchmark methodology located (source identity sub-gate advanced)

    New primary administrator publication (replacing reliance solely on legacy 2019 BondSpot document): GPW Benchmark, Rules of the Treasury Bond Index Family, consolidated 29 July 2026, applicable changes from 17 August 2026. A read-only text extraction of the original official publisher PDF confirms:

    • History of resolutions (front matter): 76/2022 replaced the prior Treasury BondSpot Poland Index rules, effective 28 Nov 2022; 22/2026 effective 9 Mar 2026; 127/2026 of 29 July 2026 added GPWB-B3Y, effective 17 Aug 2026. The English publication explicitly notes that the Polish-language version is binding in the event of doubt. For formal source approval, independently obtain/review the matching Polish text.
    • §2.1(1): TBSP.Index means Treasury BondSpot Poland Index covering PLN fixed-rate and zero-coupon Treasury Bonds with at least six months' maturity.
    • §§3.2.1–3.3.4: organized wholesale Polish Treasury bond market (BondSpot), bond/portfolio capitalization methodology, and total-return semantics explicitly including price changes, accrued interest and reinvested interest coupons. Do not treat reference-index points as a purchasable bond or executable ETF fill.
    • §§3.4.1–3.4.3: administrator derives reference prices from TBSP market fixing transactions/binding offers. This supports intended underlying index data semantics, not proof of licenses for redistribution or automated downloads.
    • §§3.7.1, 3.8.1–3.8.7: initial/final/closing publication concepts. Standard publication times 10:15 / 16:45 / 17:20 trading day; §3.8.2–3.8.4 explicitly allow change/delay/cancellation, and §3.8.7 identifies the closing index as market-performance reference. Critically: a published schedule is not an authenticated historical first_available_at for each day and version. Check source archives and exception logs rather than backdating all 2021–2026 observations to 17:20.
    • The English PDF has a potentially inconsistent 3.8.8 cross-reference to §3.8.1(b) while §3.8.7 references §3.8.1(c) for the closing index. Treat as review required against the controlling Polish text; do not infer index eligibility from an English translation inconsistency.

    Verification limits: Administrator PDF text was retrieved via the official site and independently inspected in extract; this environment did not support visual PDF screenshot rendering, so there is no visual-page attestation in this checkpoint. The document is marked public, but this is NOT an explicit licence for private AI/agent ingestion, storage, derived commercial use, onward transfer or recurring API collection. The author's/administrator's actual terms, signed dataset-specific entitlement, data source raw vintages and verified historic publication exceptions are still missing. No values, licensed market-data files, scraper, raw dataset, financial operations, provider scheduler, or database writes were created.

    Evidence scope decision:
    TBSP_CURRENT_ADMINISTRATOR_AND_METHOD_DOCUMENT_IDENTIFIED = documented;
    TBSP_PLN_BOND_INDEX_TOTAL_RETURN_SEMANTICS = supported by §§3.3.4;
    TBSP_SOURCE_REUSE_AI_AUTOMATION_LICENSE = UNVERIFIED;
    TBSP_HISTORIC_PIT_VINTAGES = UNVERIFIED;
    TBSP_POLISH_CONTROLLING_TEXT_INDEPENDENT_REVIEW = PENDING;
    PL_BONDS_TR_PLN = NOT QUALIFIED; M1–M3 INSUFFICIENT_DATA/HOLD_CASH.

  12. gniewkob commented on Nov 12, 2026

    @gniewkob
    OwnerAuthor

    2026-11-12 — binding Polish 2026 TBSP methodology and licensing gate verified

    Authoritative Polish original (supersedes prior reliance on the English translation): GPW Benchmark S.A., Regulamin Rodziny Indeksów Obligacji Skarbowych, tekst jednolity 29.07.2026; uchwała 127/2026 effective 17.08.2026. Verified against primary-publisher document text; the publisher's archive separately records prior rule vintages (9.03.2026–16.08.2026, 2.05.2023–8.03.2026 and earlier). The document is public, but publication does not equal market-data reuse entitlement.

    Control Exact controlling paragraph Supported fact Acceptance
    Identity §2.1(1), §3.2.1 TBSP.Index refers to PLN-denominated wholesale Polish Treasury fixed-rate/zero-coupon bonds, minimum 6-month maturity METHODOLOGY VERIFIED
    Return semantics §3.3.4 “indeks dochodowy”: bond price changes, accrued interest and reinvested coupons. It is not a tradeable security or broker execution price TOTAL-RETURN-TYPED BENCHMARK VERIFIED
    Reference data §§3.4.1–3.4.3 BondSpot TBSP.Price and TBSP.fixPrice underpin calculation; this does not license copying the publisher's dataset METHOD VERIFIED, SOURCE DATA NOT APPROVED
    Values and publication §§3.8.1–3.8.7 10:15 initial, 16:45 final, 17:20 closing index value; §§3.8.2–3.8.4 allow timing changes, postponement or cancellation; §3.8.7 explicitly treats §3.8.1(c) closing value as relevant to underlying market NORMAL SCHEDULE VERIFIED; HISTORIC PIT NOT VERIFIED
    Index usage rights §§1.5.1–1.5.3 Administrator grants usage rights; use within the Benchmarks Regulation scope is based on a separate “Umowa o Stosowanie”, with contractual fee provisions DO NOT ASSUME FREE REUSE
    Controlling-language cross-reference §3.8.7 vs §3.8.8 The Polish ORIGINAL also contains the mismatch: §3.8.7 uses §3.8.1(c) for closing value, but §3.8.8 refers to §3.8.1(b) while naming the closing value. This is not merely an English translation inconsistency. Request publisher clarification; conservative model policy uses §3.8.7 closing-price semantics, never claims the inconsistency grants data rights PUBLISHER CLARIFICATION OPEN

    Further primary licensing documentation: GPW/WSE Market Data License Agreement Appendices 2026 includes Product J BondSpot Indices and separate GPW Benchmark index license provisions in the product annexes. This establishes that contractual market-data and benchmark/index use rights are distinct. It does not prove that any current ISL user holds either entitlement or that every private, non-commercial statistics use requires the same contract. GPW Market Data offering identifies analytical/non-display and automated use as distinct categories. Official administrator kontakt publishes kontakt@gpwbenchmark.pl.

    Decision / next evidence: PL_BONDS_TR_PLN_TBSP_METHOD_CURRENT_VERIFIED=true; POLISH_BINDING_RULE_REVIEW=PASS; DATA_LICENSE_FOR_PRIVATE_AUTOMATED_AI_RESEARCH=UNCONFIRMED; PIT_RELEASE_HISTORY=UNCONFIRMED; CLOSING_VALUE_CROSSREF_ERRATUM=REVIEW_REQUIRED; qualified_model_series=[]; M1–M3=INSUFFICIENT_DATA/HOLD_CASH; financial operations 0. No license, contract, purchase, scrape, history file, broker action, asset transfer or production source activation is authorized.

    Contact can be pursued via a separately approved email; no external email was sent during this evidence audit.

  13. gniewkob commented on Nov 12, 2026

    @gniewkob
    OwnerAuthor

    Public-first policy update — 2026-11-12

    Decision: With the present 100 PLN/month research budget, do not purchase GPW/TBSP/index access or assume a commercial data licence merely to progress the reference analysis. Work on official public data first (NBP gold/FX #142, Eurostat HICP #144), with properly typed PRICE_ONLY, FX_MID_REFERENCE, CONSUMER_PRICE_INDEX outputs. This does not waive the GPW Benchmark rules §§1.5 or third-party rights, and no TBSP scraper/download will be enabled without an approved entitlement.

    Keep #116 open as an optional future evidence/entitlement track for fully qualifying PL_BONDS_TR_PLN, not a blocker for the independently labelled public-reference research. If there is a concrete later model need, make a cost/value analysis and ask the rights owner directly; no blanket subscription. M1–M3 currently INSUFFICIENT_DATA/HOLD_CASH, no financial actions.

  14. gniewkob commented on Nov 12, 2026

    @gniewkob
    OwnerAuthor

    Public-first supporting source observation module #150 / PR #151 now MERGED main 4aa4f1d619f73b618bed7ef041906a15061528a5, exact main 1329 offline tests SUCCESS Actions. It captures one-shot version fingerprints for public NBP and Eurostat and same-window later comparison; NOT historical original PIT, source license scope admission, verified total-return data, M1–M3 eligibility, or real financial operations. #116 remains open with all original source-rights/PIT conditions untouched.

  15. gniewkob commented on Nov 12, 2026

    @gniewkob
    OwnerAuthor

    Public-first P0 #116 source research partial evidence update, NOT source admission (12 Nov 2026).

    Found and source-tested a genuine free official Polish long-term government bond YIELD series: Eurostat irt_lt_mcby_m, freq=M/int_rt=MCBY/geo=PL. Official methodology, sections 3.4/4/17/18, confirms monthly national-currency ~10-year gov bond secondary-market yield percent per annum from ECB/national central banks. Eurostat reuse §2–3 generally allows credited reuse with exceptions (not general license of any downstream market indices).

    Real original API run #43325200200 59/59 published Polish months 2021-11…2026-09 on 2026-11-12, raw SHA acc302b641ca981b5c8e22259a613d092e5c12d0cb01e39e7e093c07c59c3abf. Mean 5.6115% p.a., range 3.12%–7.82%, max absolute adjacent-month yield level change 154 bps; none are investment performance or holding-period total return. Working code and source/negative tests in PR #153, issue #152. Current 1348 offline full tests pass, 1 skipped on PR SHA f2aa795693872a6d37acce7d1a36d3bb8df4dfab, remaining G1+stack macOS queued before merge.

    Classification for #116: EUROSTAT_PL_LONG_GOV_YIELD_PCT_PA = ORIGINAL_PUBLIC_SOURCE_VERIFIED_YIELD_REFERENCE_ONLY with currently retrieved-vintage time and SHA. NOT PL_BONDS_TR_PLN / TBSP income index; no coupon cashflows/reinvestment, no realized total return, no executable price and no evidence that today's historic response existed on original decision dates. Do not flip any source rights/PIT acceptance checkboxes; #116 stays OPEN, M1–M3 HOLD_CASH, zero orders/wallet/DB operations.

  16. gniewkob commented on Nov 12, 2026

    @gniewkob
    OwnerAuthor

    Research source evidence checkpoint: #154 and PR #155 now merged main SHA 1a9aabb969dd1444552c9bdf4bf21058db654209; original official NBP (gold, USD/EUR) + Eurostat HICP/PL 10Y yield five series jointly validated in one run, 1360 tests exact main PASS Actions. Found actual BondSpot TBSP.Index archive per-day index values and GPW historical XLS archive, with BondSpot explicit market-data policy including licence/user declaration/fee schedule. Both exchange bulk ingestion + historical source vintages still BLOCKED pending authorized rights, not circumvented. NBP/Eurostat derived macro references NOT Polish total-return benchmark, cannot close #116; M1–M3 HOLD_CASH, no financial operations.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions