§ 08 · INVESTMENT RESEARCH · RECORD
Research record
The audit trail behind the Investment Research page: 3 research snapshots with their content digests, the frozen July 2026 study with the 18 findings of its audit, the withdrawn statistics, the correction history, the verification reports, the registered forward study and the provenance of the issuer fields.
§ 08.R1 · SNAPSHOTS
Research snapshots and their content digests
| Snapshot | Class | Computed | BHI vintage | Status | Content digest (SHA-256) |
|---|---|---|---|---|---|
| IS-2026Q3-v1 | HISTORICAL ORIGINAL (qualified) | 5 Jul 2026 | Q3-2026 | frozen | fb23004c9636dd16ff6eb8c223060563ae09a81c4a1eee78110b799cdef1d3f5 |
| IS-2026Q4-v1 | CURRENT DESCRIPTIVE | 7 Nov 2026 | Q4-2026 | validated · superseded for publication by IS-2026Q4-v2 (same results) | a403abd5db51f54fd1ce72067091ec23eeded74b46c875c5b5fc56e91d441a73 |
| IS-2026Q4-v2 | CURRENT DESCRIPTIVE | 8 Nov 2026 | Q4-2026 | validated · awaiting scientific approval and rights clearance | 8378212c801153497d80754517d15be530785701d458391c01ce093e721649c8 |
§ 08.R2 · JULY 2026 STUDY
IS-2026Q3-v1: the original record, its defects and its corrections
Computed 5 July 2026 on the Q3-2026 vintage over 141 platform rows. The record is frozen; its digest is fb23004c9636dd16ff6eb8c223060563ae09a81c4a1eee78110b799cdef1d3f5. Its numerical figures are withheld pending the data-rights decision; the record, its defects and the list of withdrawn metrics stay public.
Historical calculations
| Quantity | Value |
|---|---|
| Rows of the July universe | 141 platform rows; 124 distinct tickers; 8 tickers repeated across 25 rows |
| ρ(B, return) recomputed at platform level | withheld pending the data-rights decision |
| Operating margin, one row per issuer | withheld pending the data-rights decision |
| Corrected analytical cohort | 114 issuers; figures withheld |
Findings of the forensic audit
The audit of 7 November 2026 recorded 18 findings: 14 are defects of the July figures and stay attached to the frozen record; the other 4 concern the process around it or record a check that passed. Severity is the audit’s own ranking, P0 highest.
| Finding | Severity | Kind | Component | What the audit found |
|---|---|---|---|---|
| F-01 | P0 | defect of the July study | investment universe | 8 rows (7 tickers; BNB twice) carry returns of instruments that are not the platform's issuer: FAA NextGen ticker 'Govt' = iShares U.S. Treasury Bond ETF (GOVT; B [withheld]; Q4 basket); Chainlink 'LINK' = Interlink Electronics; TRON 'TRX' = TRX Gold Corp (+[withheld]; Q1 basket); ADA and LDO = [withheld] from unresolved quotes; BNB (x2) and UNI = unresolved ECN quotes. 3 more rows are crypto asset ETFs (ETH; BTC; XRP) |
| F-02 | P1 | defect of the July study | issuer mapping | TikTok mapped to Alibaba shares |
| F-03 | P1 | defect of the July study | issuer mapping / universe membership | stale or malformed tickers excluded platforms: Block 'SQ' (renamed XYZ Jan 2025; DB has XYZ but script overrides with SQ); Canvas 'INST' (Instructure taken private 2024); WeChat '700.HK' (Yahoo symbol 0700.HK); SOL/HYPE/USDT/DOGE/AAVE plain symbols; 'UBER sub'; 'ORCL sub'; 'SNPS/CDNS' etc. |
| F-04 | P1 | defect of the July study | statistical unit | 141 rows = 124 tickers; 8 duplicate groups (25 rows); AMZN x3 in Q4; fundamentals correlation dedups by LOWEST-B platform (AMZN gets AWS's [withheld] not Amazon's [withheld]) |
| F-05 | P0 | defect of the July study | portfolio metrics (IR [withheld]; TE [withheld]; Sharpe [withheld]; LS Sharpe [withheld]; maxDD [withheld]/[withheld]/[withheld]; ann returns/vols) | series of 119 tickers (Jul 2025-Feb 2026) averaged with the Bitwise XRP ETF (listed 2025-11-20; 153 closes) by index; window 152 days labelled 'annualised'/'trailing 12 months'; values reproduced under the method (Sharpe [withheld] etc.) and differ under date alignment (full year Sharpe [withheld]; LS +[withheld]; IR [withheld]) |
| F-06 | P0 | defect of the July study | capacity [withheld] | non-USD caps summed without FX (RELIANCE.NS INR ~[withheld] raw today; Kakao KRW [withheld]; Foxconn TWD; China Mobile HKD; CATL CNY); AMZN counted 3x; GOVT ETF included; today's replication raw [withheld] vs USD+dedup [withheld] |
| F-07 | P1 | defect of the July study | Signal Evolution Q2 column | Q2 op-margin rho typed as [withheld] (= Q3 value); recomputed from the Q2 vintage (DB) + 4-Jun fundamentals = [withheld]; 'held ([withheld])' unsupported; return figures match the bulletin but a second June run gave [withheld]/[withheld]; Q2 output file overwritten (fixed filename) |
| F-08 | P1 | defect of the July study | fundamentals labels | non-USD revenues stored in local units under '$B' (TSM [withheld]; RELIANCE [withheld]; Samsung [withheld]; Sony [withheld]; GoTo [withheld] ...); revBasis is the BHI release label not a fiscal period; bubble size uses rev as 'mcap' |
| F-09 | P2 | defect of the July study | Spearman convention | rho [withheld] vs [withheld] tie-corrected; [withheld] vs [withheld]; 17 tied returns from duplicate issuers |
| F-10 | P2 | defect of the July study | governance of investment outputs | investment snapshots outside the append-only research record; Q1 and Q2 outputs lost |
| F-11 | P1 | defect of the July study | page wording | 'Investment Signal'; 'Does structural lock-in predict equity performance?'; 'strongest cross-sectional signal'; 'market-beta effect not a breakdown of the lock-in thesis'; 'Top quartile · current basket'; 'Capacity'; 'held ([withheld])' |
| F-12 | P2 | process or verification | refresh mechanism | manual one-off runs at Q1/Q2/Q3; Q4 release omitted the step; policy written only after the fact |
| F-13 | P2 | process or verification | data rights | redistribution status untested |
| F-14 | P2 | process or verification | stored returns | 136/141 within 1 pp of today's close-based 52-week change (median [withheld]); last sessions 2 Jul (US) / 3 Jul (non-US); price return (no dividends) |
| F-15 | P2 | defect of the July study | missing-data handling | ADA and LDO stored as [withheld] returns and counted in Q1 |
| F-16 | P3 | defect of the July study | currency of returns | mixed-currency cross-section |
| F-17 | P2 | process or verification | BHI vintage identity | 141/141 within [withheld]; Q3->Q4 0 changes |
| F-18 | P2 | defect of the July study | version/time fields | one label carries BHI release and observation month; n switches between rows and issuers without saying so |
Withdrawn statistics
| Metric | Reason |
|---|---|
| sharpeQ4 | WITHDRAWN — daily series aligned by array index and truncated to the shortest series (Bitwise XRP ETF, listed 2025-11-20; 152 sessions); the statistic describes no calendar period |
| sharpeLS | WITHDRAWN — same method defect; sign depends on the window |
| informationRatio | WITHDRAWN — same method defect |
| trackingError | WITHDRAWN — same method defect |
| maxDrawdownQ4/Q1/Universe | WITHDRAWN — same method defect |
| annReturn/annVol (Q4, Q1, universe) | WITHDRAWN — same method defect; labelled annualised over 152 index-aligned days |
| capacityB | WITHDRAWN — sum of provider marketCap over the Q4 ticker LIST (AMZN counted three times) without currency conversion (INR, KRW, TWD, HKD, CNY summed as USD), including a Treasury-bond ETF, times [withheld]; not a capacity measure in any case — CAPACITY NOT ESTIMATED |
| q2_column.rho_b_opMargin = [withheld] | WITHDRAWN — typed constant equal to the Q3 value; recomputation from the Q2-2026 vintage and the 4 June 2026 fundamentals gives [withheld] (RECONSTRUCTED, see derived/) |
Correction history
CR-2026-11-08-01 · recorded 8 November 2026 · RECORDED — the components above are not shown as results; the replacements are computed · snapshot digest fb23004c9636dd16ff6eb8c223060563ae09a81c4a1eee78110b799cdef1d3f5
| Component | Classification | Finding |
|---|---|---|
| Figure 4.1 "Cumulative return, indexed to 1.00" (equity curve) | FABRICATED VISUALISATION — WITHDRAWN | not computed from prices: each line was a straight interpolation from [withheld] to an endpoint derived from a withdrawn annualised return (annReturn × 6/12), with a deterministic sine term added as noise; the month labels were typed |
| Figure 3.1 "Parameter attribution to the Q4 − Q1 spread" (bar chart) | UNSOURCED CONSTANTS — WITHDRAWN | the eleven bar values were typed constants in the page source, not computed from the scores |
| Table "Capture-axis attribution" (Q4 mean, Q1 mean columns) | FABRICATED VALUES — WITHDRAWN | the two mean columns were generated from the typed deltas by arithmetic (delta + [withheld]; 10 − |delta| + [withheld]), not measured |
| Institutional-access API sample ("count":100) | STALE TYPED FIGURE — CORRECTED (the new page computes the count) | a typed platform count shown beside a 184-platform release |
Evidence cited by CR-2026-11-08-01
- Figure 4.1 "Cumulative return, indexed to 1.00" (equity curve). data/research/investment/snapshots/IS-2026Q3-v1/code/InvestClient.v1.tsx line 290 (sha256 e3173adb484e726c…): const months = ['Oct 25', 'Nov', 'Dec', 'Jan 26', 'Feb', 'Mar', 'Apr'];; data/research/investment/snapshots/IS-2026Q3-v1/code/InvestClient.v1.tsx line 292 (sha256 e3173adb484e726c…): const q4Final = 1 + advData.annReturnQ4 / 100 * (6 / 12);; data/research/investment/snapshots/IS-2026Q3-v1/code/InvestClient.v1.tsx line 301 (sha256 e3173adb484e726c…): const noise = Math.sin(i * [withheld]) * [withheld];
- Figure 3.1 "Parameter attribution to the Q4 − Q1 spread" (bar chart). data/research/investment/snapshots/IS-2026Q3-v1/code/InvestClient.v1.tsx line 82 (sha256 e3173adb484e726c…): const ATTRIBUTION = [
- Table "Capture-axis attribution" (Q4 mean, Q1 mean columns). data/research/investment/snapshots/IS-2026Q3-v1/code/InvestClient.v1.tsx line 648 (sha256 e3173adb484e726c…): <td className="r">{a.delta > 0 ? (a.delta + [withheld]).toFixed(1) : (10 - Math.abs(a.delta) + [withheld]).toFixed(1)}</td>; data/research/investment/snapshots/IS-2026Q3-v1/code/InvestClient.v1.tsx line 649 (sha256 e3173adb484e726c…): <td className="r">{a.delta > 0 ? (a.delta + [withheld] - a.delta).toFixed(1) : (10 - Math.abs(a.delta) + [withheld] + Math.abs(a.delta)).toFixed(1)}</td>
- Institutional-access API sample ("count":100). data/research/investment/snapshots/IS-2026Q3-v1/code/InvestClient.v1.tsx line 723 (sha256 e3173adb484e726c…): <span style={{ color: 'var(--ink-4)' }}>{'{'}</span> "instrument":"Black Hole Index", "count":100, "platforms":[...] <span style={{ color: 'var(--ink-
§ 08.R3 · VERIFICATION
Verification reports of the current study
IS-2026Q4-v2 · validated · BHI vintage Q4-2026 · financial cutoff 6 November 2026 · scientific approval not given · data rights UNRESOLVED.
| Content digest | 8378212c801153497d80754517d15be530785701d458391c01ce093e721649c8 |
|---|---|
| Level 1 · IS-2026Q4-v2 | 8378212c801153497d80754517d15be530785701d458391c01ce093e721649c8 |
| Extensions oracle | PASS — 2976 of 2976 comparisons within tolerance (ext-oracle-1.0 (python); report sha256 331816f1b7874b6c…) |
| extensions oracle report sha256 | 331816f1b7874b6c12a9411f13a7d1192435664149b3fbb9cea05ae55ef6a2dd |
| Negative controls | PASS — 110 of 110 controls detected, 0 blind (report sha256 32f21130363aadbc…) |
| negative controls report sha256 | 32f21130363aadbc0e09dacb4af66f80cfdb1c095f1d0b8acafd7680c701fd93 |
| Independent oracle | PASS — 558 of 558 comparisons within tolerance (oracle-2.1 (python); report sha256 6d086e69a21ebc90…) |
| oracle report sha256 | 6d086e69a21ebc90560c1ed772905fdf46f34b6eb3576bd4be02b98adfd73899 |
| recorded at | 2026-11-08T10:03:37Z |
| recorded by | scripts/research/promote_snapshot.py record-verification |
§ 08.R4 · FORWARD STUDY
The registered forward study
- Registration
- PR-2026-11-09-v5
- Status
- REGISTERED — design (2026-11-08T06:10:17Z) and approval (2026-11-08T06:51:51Z) before the origin session (cutoff 2026-11-08T23:00:00Z)
- Registration digest
- 537640f399e75b85754ac179755a23d5ecbe9c92703a66d5c4793be53e045370
- Protocol
- PROTOCOL_v1.4.md · e8bfe672560fb940d98bd2898ecc29cdf4990242cbbd7bf65024c7a41eea4022
- Design frozen
- 2026-11-08T06:10:17Z
- Cutoffs
- design before 2026-11-08T23:00:00Z; outcome information from 2026-11-09T23:00:00Z
- Origin prices
- PENDING — to be captured after the origin session with `capture-origin` from a lawfully obtained price record; the design is frozen without them
- Maturity rule
- first recorded trading session of the issuer ON OR AFTER the target date, no later than target + 7 calendar days; otherwise MISSING for that horizon; never a session before the target
§ 08.R5 · ISSUER REFERENCE
Where the issuer fields come from
issuer_reference_v1: 117 securities; 96 identified by SEC CIK and 21 identified by LEI. Company names: 50 typographic renderings of the official name, 66 as recorded in the official record, 1 BHI editorial English label (the official record holds only the Chinese legal name).
Display rule R1 (2026-11-08): strip EDGAR state tags (" /DE/", "/UK", trailing "/"); a name whose letters are 80 % or more upper case is set in title case word by word, with the phrase and token tables recorded in the manifest; any other name is kept as recorded.
| Registry digest | dfabee2837815bd81fcc64cd3b30622353f5814bf86e02c4f9f427874d3fe696 |
|---|---|
| Builder digest | 5c44575afb9ce89c279d0c75fec23b8366ddd6a1ebb766eee5da1485d6cad134 |
| gleif lei records | https://api.gleif.org/api/v1/lei-records/<LEI> · retrieved 2026-11-08T13:48:51Z · 21 records · GLEIF: "The data available through the Access Service are provided under the CC0 licence" (https://www.gleif.org/en/meta/lei-data-terms-of-use, retrieved 2026-11-08) |
| iso10383 mic | https://www.iso20022.org/sites/default/files/ISO10383_MIC/ISO10383_MIC.csv · retrieved 2026-11-08T13:37:57Z · sha256 09a0d07b7b8348147453a2d8b18bc6638e3c5b16f749462582c9b1edf0fd0e38 · ISO 10383 Registration Authority list, published for automated identification of markets; only the code, operating MIC, market name and acronym of the venues used are recorded |
| iso4217 | currency codes of ISO 4217; the trading currency of each venue is recorded by BHI in MIC_CURRENCY (venue convention); LSE quotes in pence (GBX), a market convention, not an ISO 4217 code |
| sec company tickers exchange | https://www.sec.gov/files/company_tickers_exchange.json · retrieved 2026-11-08T13:37:57Z · sha256 e1469df6a5e24ec90676d0bbd0059f857167207722b4e9952852b8f0806a9472 · sec.gov: "Information presented on sec.gov is considered public information and may be copied or further distributed by users of the web site without the SEC’s permission." (https://www.sec.gov/about/privacy-information, retrieved 2026-11-08) |
Field classification
| Field | Class | Source |
|---|---|---|
| security_ticker | BHI-authored record key | exchange-assigned local code with a market suffix (common market convention); the join key of mapping_v1 |
| issuer_name_official | Official factual reference | SEC company_tickers_exchange.json name (US registrants) or GLEIF legal / English legal name (others) |
| issuer_id (SEC_CIK / LEI) | Official factual reference | SEC CIK; GLEIF LEI (CC0) |
| jurisdiction | Official factual reference | GLEIF entity.jurisdiction (LEI rows only) |
| display_name | BHI-authored (typographic rendering of the official name) | rule R1 + phrase/token tables; one BHI editorial English label (CATL) named in NAME_OVERRIDES |
| venue_mic / venue_operating_mic / venue_name_iso | Official factual reference (ISO 10383 code) | US rows: SEC exchange field; other rows: BHI-recorded from the record key (venue_attribution states which) |
| venue_label | BHI-authored (the ISO 10383 acronym, else the ISO market name set by rule R1) | |
| local_code | Official factual reference (exchange-assigned code) | US rows: SEC ticker; other rows: the record key without its market suffix (unverified against the venue list) |
| trading_currency | Official factual reference (ISO 4217 code) | venue convention table MIC_CURRENCY recorded by BHI |
| price_unit | BHI-authored (market convention) | GBX = pence for the London venue |
| category, platform_id, platform_name, b, zone, quartile, capture, escape, b_by_vintage, mapping_class, inclusion_policy, identity_verified, confidence, valid_from, valid_to, corporate_actions, legacy_* | BHI-authored | BHI index and mapping registry mapping_v1 |
| listed_issuer, exchange, trading_currency of mapping_v1 and of the Level-1 file; availability flags; fundamentals coverage counts; provider short names, quote types, listing dates | Restricted third-party (provider quote metadata, 2026-11-07) | not exported; replaced by the fields above |
Correction: until this release the company names, exchange labels and trading currencies on the Investment Research page, in its API and in the universe CSV came from provider quote metadata recorded when registry mapping_v1 was built (7 November 2026), while the API and the CSV described them as public facts recorded by BHI and the CSV stated that no provider-compiled fields were exported. They are no longer published; the official reference above replaces them.
§ 08.R6 · DATA RIGHTS
Publication levels and their standing
| Level | Standing |
|---|---|
| 1 · BHI-native | PUBLISHED · IS-2026Q4-v2 · PASS — 2976 of 2976 comparisons within tolerance (ext-oracle-1.0 (python); report sha256 331816f1b7874b6c…) |
| 2 · Financial aggregates | VERIFIED — NOT PUBLISHED · IS-2026Q4-v2 · scientific approval not given · data rights UNRESOLVED |
| 3 · Prospective | REGISTERED · PR-2026-11-09-v5 · 0 horizons evaluated |
Prices and fundamentals of the current study were retrieved from Yahoo Finance through an open-source client. Its terms do not permit redistribution: per-security values stay in restricted files, and the financial aggregates wait for a CLEARED rights record. A licensed or official source changes the snapshot identity: it would be computed as a new snapshot with its own oracle and controls, never written into IS-2026Q4-v2.