Wire synthetic_levered histories into disclosed pre-IPO levered path fallback

Objective

Teach the levered path-return layer to use already-imported __SYNTHETIC_LEVERED_<BASE>_<L>X__ series only under an explicit modeled-history policy. Default behavior stays live OHLCV plus existing Testfol SIM merge. Live LETF returns win overlap. Report outputs label the ideal daily-reset/DFF method.

Approach

• research/levered_path_returns.py accepts modeled_history off (default) or synthetic_levered. Merge order is synthetic, then Testfol SIM, then live so live wins every overlap month.
• Ambiguous live tickers such as DRN are not silently proxied; a unique registry path or explicit base/leverage is required.
• Origin method and funding come from sim_series_meta.origin_column (daily_reset_linear/v1 + DFF, or testfol_daily/v1 JSON).
• Catalog, Meta, and weighted-Meta producers pass modeled_history=synthetic_levered and write modeled_history.policy plus window-rule disclosure. Report/read paths do not mutate data/market.sqlite.

Results

• Focused path-return tests cover default-off, live-over-synthetic overlap, no silent DRN proxy, and compose disclosure of daily_reset_linear/v1 + DFF.
• docs/site-data/catalog-leverage.json records method_id=substitute-path-returns/v1 and modeled_history.policy=synthetic_levered.
• Catalog/Meta rebuild changed 73 of 336 L>1 levels; example: faber-sector-rs L=2 window is 303 months (1993-12–2026-07). Weighted-Meta collapsed-cohort windows moved 54/49 → 127/136 months.

Validation

• python3 -m pytest -q tests/test_levered_path_returns.py tests/test_catalog_leverage.py tests/test_catalog_leverage_meta.py tests/test_catalog_leverage_weighted_meta.py — controller receipt check-1 passed.
• python3 -m pytest -q tests/test_levered_path_returns.py — controller receipt check-2 passed (temporal overlap/causality).
• python3 -m research.reports.results and python3 research/scripts/check_static_reports.py docs rebuild the public changelog and static bundle.

Web plan

Existing catalog.html leverage select and Meta/CW <em>-lev-</em> pages remain the discoverable entry. No new modeled-history toggle: a user-operated opt-in would imply tradable pre-IPO LETF history. Producers opt in explicitly; pages already linked from catalog/Meta navigation. Public JSON is producer-owned. Modeled prefixes are labeled estimates, not native LETF history.

Limitations

Synthetic prefixes can differ from live LETF returns because of tracking error, index construction, fees, swaps, taxes, and later provider revisions. Missing or ambiguous synthetic series fail closed rather than substituting another proxy.