[mspinola]

Summary

• HG adds 085691 (pre-grade-Norgate integration: cotdata data layer (Norgate prices, CFTC ETL, Databento dormant, code-stitching) #1 copper, same 25,000 lb). Ten-week overlap with OI migrating old to new; the larger-OI rule from get_cot: overlap weeks keep the larger scaled open interest #122 hands off on 1989-11-30.
• ZR adds 039781 (MidAmerica / Chicago Rice & Cotton rough rice to 1994-09-27, same 2,000 cwt). OI and trader counts continuous at the seam (3,177 to 3,424).
• 6B adds 096741 at scale 0.4. The CFTC relabels the successor "BRITISH POUND STERLING- NEW" in 1988; OI drops 2.4x at the seam against the 25,000 to 62,500 GBP size ratio of 2.5. Stitched OI: 15,633 then 16,980 across the switch.
• All three now start 1986.
• Not stitched, documented inline: DX (098661, seam too noisy to pin a size ratio) and NKD (240741 under the 20-trader publication floor since 2026-03-03, not delisted; the yen contract's scale moves with USDJPY).
• reading-the-store.md: the 2016 to 2018 TFF regime change in the equity index futures was checked across all 22 TFF markets and has no common week, so it is a calibration boundary rather than a reclassification seam.

Test plan

• Registry test for the three stitches and the two deliberate non-stitches; the grain test no longer lists ZR as single-code
• Full suite passes; ruff check src tests scripts clean
• Producer run for the six codes into a scratch store; each seam inspected (no duplicates, one code label, continuous OI)

Deployment

Same as #123: the Windows box pulls, the next --cot run writes three new per-code files. Afterwards clear and rebuild the cotmetrics cache for HG, ZR and 6B (it does not bust on history changes).

🤖 Generated with Claude Code