TS-GMRI™ — Errata & Corrections
A complete, dated, signed record of every methodology error, data correction, and implementation bug-fix that has affected the TS-GMRI™ family of indices. We document every mistake openly. This is how a benchmark earns trust.
Correction policy (what counts, what we publish)
The TS-GMRI™ family follows the IOSCO Principles for Financial Benchmarks (§11 Methodology Changes, §12 Transition, §13 Errors). Every event below falls into one of three categories:
- Methodology change — a deliberate update to the rules (e.g. adding a constituent, refining weighting). Always issued as a numbered addendum (v1.1, v1.2, ...) with SHA-256 + Bitcoin timestamp before taking effect.
- Implementation bug-fix — code computed values that did not match the published methodology. Published series is retired and backfilled with full disclosure. Underlying rules unchanged.
- Data correction — an upstream source (CoinGecko, LBMA, CFB) revised a value after we ingested it. Snapshot is rebuilt; previous value archived.
What we never do. We never silently overwrite a published index value, never delete prior snapshots, never adjust base date once locked, never alter constituent membership rules retroactively to flatter a chart. Every snapshot from base date 22 June 2026 onward is preserved in data/gmri/history/<date>.json in the public repository.
Materiality threshold. Any error that changes a published index value by more than 0.10 points is disclosed here within 48 hours of identification. Smaller corrections (rounding, typography, link rot) are corrected in place without a separate entry.
Chronological log
MOD-2026-v1.4 · 29 June 2026 · Fix-aligned snapshot standardization (methodology enhancement)
Category: Methodology change (not an error) · Severity: Disclosed under IOSCO Principle 12 materiality threshold (max drift 3.70pp on TS-GMRI-CAP20, 23 June) · Scope: All 6 live TS-GMRI™ family members · Resolution: Backfilled 22–28 June 2026 + refresh timing updated
What changed. The daily TS-GMRI™ family snapshot is migrated from an arbitrary midnight UTC close to a fix-aligned 14:00 UTC convention coherent with the LBMA Gold/Silver/Platinum/Palladium PM fixes. Two concrete sources change: (1) PAXG, the largest single constituent at ~80% of base weight, is now sourced from the CF Benchmarks PAXGUSD_RR Registered Benchmark — an FCA-authorised, UK BMR-regulated daily settlement at 15:00 London — instead of an unregulated CoinGecko midnight quote. (2) Six other liquid token fixes (XAUT, KAU, XAUM, KAG, XAGM, PGOLD) are taken from a CoinGecko 14:00 UTC hourly snapshot. The 11 tail tokens with combined weight <3% remain on the existing daily close (documented under §13 of the addendum — thin order books make hourly snapshots noisier than the daily close for them).
Why a methodology enhancement, not an erratum. The v1.0 published values for 22–28 June 2026 were internally consistent under their own snapshot convention. v1.4 replaces an ad-hoc convention with a regulator-grade, fix-aligned one. The v1.0 series remains preserved immutably in data/gmri/history-pre-v14.json and on-chain via the v1.0 OpenTimestamps proof. No published value was wrong under v1.0; the new convention is simply better aligned with how comparison benchmarks (LBMA fixes, ETF closes, CFB PAXG) settle.
Backfill impact (TS-GMRI flagship).
| Date | v1.0 | v1.4 | Drift (pp) |
|---|---|---|---|
| 2026-06-22 | 100.00 | 100.00 | 0.00 (base) |
| 2026-06-23 | 98.50 | 98.24 | −0.26 |
| 2026-06-24 | 96.78 | 95.96 | −0.82 |
| 2026-06-25 | 94.52 | 95.53 | +1.01 |
| 2026-06-26 | 95.03 | 96.85 | +1.82 |
| 2026-06-27 | 96.89 | 96.93 | +0.04 |
| 2026-06-28 | 96.75 | 96.73 | −0.02 |
Refresh window updated. The daily refresh now closes shortly after the LBMA fix, so all future daily snapshots inherit the v1.4 timing convention natively.
Disclosure. See methodology-v1.4-addendum.md (SHA-256 b2eae103725d3a0b6172f58b4cfcab1eb45624d99c663318221e3726676ec57d, Bitcoin-stamped via OpenTimestamps on 29 June 2026 — awaiting block confirmation). Reproducible: feed the .md file through sha256sum to verify the checksum, then run ots verify methodology-v1.4-addendum.ots against any Bitcoin node.
MOD-2026-v1.2 · 22 June 2026 · Series A baseline lock (methodology addendum)
Category: Methodology change (not an error) · Severity: Informational · Scope: Internal site composites — out of TS-GMRI™ benchmark scope · Resolution: Locked Series A baseline at 100 from base date
What changed. On 22 June 2026 we locked the Series A baseline for TSM internal site composites so that they rebase to 100 on ISO week 2026-W25 (15–21 June 2026). These composites are not tradable benchmarks and do not fall under the TS-GMRI™ methodology. They are documented here for full audit transparency: every numbered version of any TSM-published index is Bitcoin-stamped, regardless of whether it is in the GMRI family or not.
Why timestamp it. IOSCO Principle 11 (Content of the Methodology) requires the administrator to publish methodology changes before they take effect. We satisfy this for the GMRI family via the SHA-256 + OpenTimestamps pipeline. We apply the same discipline to the internal site composites so that, if a Series B re-base is ever proposed, the original Series A constants cannot be silently rewritten.
Disclosure. SHA-256 1477d602e213a9b6b95251480b7171d165ee1720201f9e617878cf983119d108, Bitcoin-stamped via OpenTimestamps, confirmed in block #955798 on 28 June 2026 at 11:26 UTC.
Effect on TS-GMRI™ family. None. v1.2 does not change any constituent, weight, calculation rule or snapshot in the TS-GMRI™ benchmark family. Listed in this errata for completeness of the methodology version chain.
ERR-2026-001 · 28 June 2026 · TS-GMRI-EW formula correction
Category: Implementation bug-fix · Severity: Material (26.83-point overstatement) · Affected member: TS-GMRI-EW only · Resolution: Backfilled full 22–28 June series
What happened. The TS-GMRI-EW (equal-weighted) variant was computed by averaging per-constituent market-cap ratios (mcaptoday / mcapbase) instead of per-constituent price ratios (Ptoday / Pbase). On 26–28 June 2026 a single microcap constituent, TGLD (base mcap ≈ US$74,886), underwent a 5.4× supply expansion. Its market-cap ratio jumped to 5.51 while its actual token price moved only +1.46%. Averaged across 18 constituents this contributed +25.05 points to TS-GMRI-EW, pushing the published value to 122.73 on 28 June.
Impact.
| Date | EW (published, retired) | EW (corrected) | Delta |
|---|---|---|---|
| 2026-06-22 | 100.00 | 100.00 | 0.00 (base) |
| 2026-06-23 | 99.72 | 97.72 | −2.00 |
| 2026-06-24 | 99.58 | 97.13 | −2.45 |
| 2026-06-25 | 97.03 | 94.82 | −2.21 |
| 2026-06-26 | 122.26 | 94.50 | −27.76 |
| 2026-06-27 | 124.63 | 96.52 | −28.11 |
| 2026-06-28 | 122.73 | 95.90 | −26.83 |
Root cause. The function computeEW(records) in scripts/gmri-family.js aggregated market-cap ratios. Market-cap is the product of price and circulating supply; on a microcap constituent a supply emission therefore propagated 1:1 into the index return. This violated the published methodology, which defines EW as price return.
Fix. The function now computes EW = (1/N) × Σ (Ptoday / Pbase) × 100, matching S&P 500 Equal Weight, MSCI Equal Weighted and FTSE EW conventions. When a constituent has no price on a snapshot date, a per-constituent mcap-ratio fallback applies. Base prices are stored in data/gmri/base.json alongside base mcap.
Disclosure. See methodology-v1.3-addendum.md (SHA-256 de230770f6afa7fea4fcbafc2a56c5db2f574f32d0fa503a572df5ebb54c8d14, Bitcoin-stamped via OpenTimestamps, confirmed in block #955798 on 28 June 2026 at 11:26 UTC). Source commit (internal repository): b90b40d. Backfilled snapshots are preserved in data/gmri/history/ with both price_usd and mcap_usd per constituent. Retired EW values are not adjusted within the original snapshots; they were never representative of the underlying methodology and are listed in this errata for full transparency.
Outliers driving the gap.
- TGLD — price +1.46% / mcap +450.91% (5.4× supply expansion)
- GGBR — price −2.86% / mcap +29.19%
Members not affected. Cap-weighted family members (TS-GMRI, TS-GMRI-CAP20, TS-GMRI-AU, TS-GMRI-PRE, TS-GMRI-AG) correctly aggregate via the sum of market caps. They were never miscomputed. No values retired.
Methodology change log
Numbered methodology versions and addenda issued since base date. Each is locked with SHA-256 and (where confirmed) a Bitcoin timestamp via OpenTimestamps. Earlier versions remain referenced in the methodology document.
| Version | Issued | Scope | SHA-256 (first 16) | OTS |
|---|---|---|---|---|
| v1.0 | 16 June 2026 | Base methodology, family of 16 members, base date 22 June 2026 = 100 | dc8a47bf7f532641 |
confirmed #953968 |
| v1.1.2 | 17 June 2026 | Wrapper dimension + MEQ/ETF/ROY slots (family grows to 19) | f5f1dafe3354ca21 |
confirmed #954027 |
| v1.2 | 22 June 2026 | Series A baseline lock (internal site composites — out of GMRI scope) | 1477d602e213a9b6 |
confirmed #955798 |
| v1.3 | 28 June 2026 | TS-GMRI-EW formula correction (this errata, ERR-2026-001) | de230770f6afa7fe |
confirmed #955798 |
How to report a suspected error
If you believe a TS-GMRI™ value is wrong, email info@truesourcemetals.com with the date, member, observed value and your reasoning. We respond within 48 hours and, if the report is material, publish a numbered ERR entry above with full disclosure — whether the cause is on our side or upstream.