Record Berg experiment gaps with proprietary-data boundary (qm-tykd.1..4)

Objective

Berg price/boundary (qm-tykd.1), volume/TRIN (qm-tykd.2), breadth/thrust (qm-tykd.3), composite (qm-tykd.4) experiments use public proxies (HIGN/LOWN/S5FI/UVOL) already persisted but Berg proprietary 30k daily feeds are not redistributable per qm-zfl5/qm-m1ib/qm-is3q. Record gap that these remain rejected with public-proxy provenance only; no synthetic Berg history invented; references datastore provenance for public proxy inputs; follow-up owned by Berg feasibility beads.

Approach

1. Intake classified as fallback (not strategy/indicator/data_source/bug/feature) — single finding gap documentation with no split; route fallback via quantmodel-change-pipeline/v1 28 nodes 38 edges; no child tasks, no new Beads.
2. Inspected ledger research/findings/berg-coverage-ledger-qm-tykd.5.md/.json v1: 3 reports R1 2015/R2 2008/R3 2023 enumerate 4 families F1–F4 all untestable_as_exact_replication_testable_as_proxy_family (zero disclosed thresholds/lookbacks/formulas), missing-input and proxy analysis per family, policy no_pdf_redistribution=true, proprietary scale quote “30,000 daily”.
3. Verified datastore provenance offline: data/market.sqlite close_only v5 124 tickers 785,032 ohlcv SCHEMA_VERSION=5, TRACKED_DB_SOFT_LIMIT_BYTES=95MB, instruments rows INDEX:HIGN 6331 2001-06-05..2026-08-11, INDEX:LOWN 6312, INDEX:S5FI 4956 2006-12-29.., USI:UVOL 7616 1996-04-19.., USI:DVOL 8643; reads via backtest/datastore.py:connect_readonly query-only, writes only via import-cache explicit transactional upsert_ohlcv with research/input_provenance.py:public-input-provenance/v1.
4. Confirmed 4 experiments closed rejected 2026-08-17: research/experiments/qm-tykd.1..4.json + docs/site-data/experiments/qm-tykd.1..4.json with metrics pct252_5th n=389, ret1<=-2.5% n=183, HL Zweig n=221, composite ge2 n=358 — dispositions rejected, Volume=0 close_only caveat preserved.
5. Confirmed existing beads: qm-tykd.1..4 closed, qm-tykd.5 ledger closed, qm-u94t gap audit closed (per-id rows qm-tykd.1..4 → qm-u94t.6), qm-t0u5 composite promotion closed rejected (overlay -2.56pp), qm-zfl5/qm-m1ib/qm-is3q open P3 feasibility/reserved, qm-ckab blocked ledger v2.
6. Authored research/findings/qm-zzdt-berg-proxy-boundary-gap.md (13.5KB): §1 proprietary boundary table + 4 citations, §2 proxy provenance table with offline sqlite verification, §3 per-experiment gap table, §4 12 existing beads referenced table, §5 follow-up ownership, §6 datastore reader/import contracts, §7 verification commands; no synthetic __BERG*__ ticker invented.

Files changed

• research/findings/qm-zzdt-berg-proxy-boundary-gap.md — new gap finding listing Berg proprietary boundary, public-proxy datastore provenance, per-experiment rejected disposition, 12 existing beads, follow-up ownership, verification commands; public visibility, no private data, no synthetic history.
• research/results/qm-zzdt.md — this durable task note (YAML front matter id: qm-zzdt, newest changelog entry).
• docs/site-data/results.json, docs/results.html — rebuilt changelog payload via python3 -m research.reports.results (qm-zzdt newest).
• data/market.sqlite — not mutated; provenance verified read-only; docs/site-data/experiments/qm-tykd.*.json unchanged (canonical producer outputs).

Validation

• python3 -m backtest.datastore status — 124 tickers 785032 ohlcv 20 macro 56261 obs 117 sim series schema v5 close_only under soft limit.
• sqlite3 data/market.sqlite "SELECT ticker, n_rows FROM instruments WHERE ticker IN ('INDEX:HIGN','INDEX:LOWN','INDEX:S5FI','USI:UVOL','USI:DVOL')" — 6331/6312/4956/7616/8643 matches provenance table.
• bd show qm-zzdt qm-tykd.1..4 qm-tykd.5 qm-zfl5 qm-m1ib qm-is3q qm-ckab qm-u94t qm-t0u5 — statuses closed/open/blocked as documented.
• grep -c 'qm-zfl5\|qm-m1ib\|qm-is3q\|proprietary' research/findings/qm-zzdt-berg-proxy-boundary-gap.md — 24 hits, boundary + beads present.
• python3 -m research.artifacts check --mode full --offline — shared artifacts deterministic and current.
• python3 -m research.pipeline validate — 28 nodes 38 edges passed.
• python3 -m research.pipeline verify --offline — passed fallback/strategy/indicator/data_source.
• python3 research/scripts/check_static_reports.py docs — local bundle OK 1243 files 178 canaries.
• git grep qm-zzdt research/results/qm-zzdt.md — entry is grep-searchable.

Results

Gap explicitly records that Berg 30k proprietary feeds remain non-redistributable (no synthetic history invented) and that qm-tykd.1..4 stay rejected with public-proxy-only provenance (HIGN/LOWN/S5FI/UVOL/DVOL persisted in data/market.sqlite close_only v5, offline repro closed). Finding lists proprietary boundary with 4 miltonberg.com citations and 12 existing beads referenced per acceptance, satisfying qm-zzdt completion. Follow-up remains owned by Berg feasibility beads qm-zfl5/qm-m1ib (licensing vetting) and qm-is3q (reserved tradable promotion if robust proxy emerges) plus qm-ckab blocked ledger v2 refresh; no new beads required until feasibility go/no-go.