Study: Filtered-Levered Gold vs Gold Buy-and-Hold
Code: https://github.com/receiptsandregimes/receipts-lab/tree/main/studies/filtered-levered-gold-vs-bnh (status: protocol-only, seeded 2026-10-12)
Question
Does a trend-filtered gold position, levered 2–3× with a stated financing model, beat gold buy-and-hold on CAGR while never producing a rolling-5-year loss worse than B&H’s — i.e. does the gold-as-call-option shape survive net of whipsaw and carry? (Frank’s heuristic, on trial.)
Claims tested
- gold-call-option-filtered-levered (primary)
- gold-trend-filter-underperforms-bnh (same baseline leg; unfiltered vs filtered)
Data & assumptions
- Gold monthly (and daily for whipsaw sensitivity), 1971-08 → present, per the lbma-gold-price composite leg: frozen PM-fix history 1968-04→2026-02 + seam-normalized GLD total-return live extension (GC=F month-ends second opinion); daily-MA variant runs fix-based to 2026-02, GLD/GC=F basis after — fidelity caveat recorded there. Bills: TB3MS. Margin-rate proxy for the levered leg; report bill-financed AND margin-financed variants — financing cost IS the premium in this trade.
- Filters: 12-mo total return > 0; 200-day MA (daily variant); 20-mo MA (monthly). Filter ON → target exposure L ∈ {1, 2, 3}; filter OFF → cash. Position from prior month-end data; costs 10 bps per unit exposure change.
- Engine: pandas under
studies/filtered-levered-gold-vs-bnh/code/. - Benchmarks: gold B&H, T-bills, and 60/40 (the honest “what else would you hold” comparator for a levered sleeve).
Method
- Unfiltered vs filtered 1× (reproduces/updates the Returnstream chart — answers gold-trend-filter-post-2023).
- Leverage sweep: CAGR, max DD, Sharpe, worst rolling-5-year total return per L × financing model. The claim survives only if worst-5yr never worse than B&H while CAGR ≥ 2× B&H-ish at L=2.
- Regime decomposition: 1980–2000 (the bleed), 2001–11, 2013–19, 2024–26 separately — where does leverage help vs where does it compound a whipsaw?
- Sensitivity: filter family × lookback × rebalance frequency; report the heatmap so we see if the surviving region is wide or one lucky cell.
Questions this study must answer or update
- gold-trend-filter-post-2023
- does-the-cash-leg-earn-its-keep (this is absolute momentum in its purest single-asset form)
Results
(pending — no code run yet)
Verdict
(pending)
Next tests prompted
- If survives: port to dual-momentum on gold/SPX/bills (proper Antonacci on our pair) — becomes its own study if the co-rally question demands hold-both variants.
- If financing cost kills it at 3× but not 2×: the claim’s honest form is rate-regime-dependent — check 1977–81 (double-digit bills) separately; that window likely falsifies naive leverage.