Gold/SPX Ratio-Momentum Rotation Replication (QuantifiedStrategies)

Study: Gold/SPX Ratio-Momentum Rotation Replication

Code: https://github.com/receiptsandregimes/receipts-lab/tree/main/studies/gold-spx-ratio-rotation-replication (status: protocol-only, seeded 2026-10-12)

Question

Does the fully-disclosed gold-spx-rs-ratio-rotation signal reproduce CAGR ≈ 9.9% / max DD ≈ 22% (1985–Jul 2020) — and does the headline edge survive dividend-adjusted SPX data, realistic costs, and different SMA lengths?

Claims tested

• qs-rotation-beats-both-buyhold
• qs-rotation-gfc-divergence
• qs-rotation-asset-contributions-equal

Data & assumptions

• Universe: S&P 500 price index (matches source) AND total-return variant (Yahoo ^SPX + dividend-reinvestment proxy, or SPY adjusted-close as proxy); gold monthly spot per lbma-gold-price composite leg (frozen LBMA fix history + seam-normalized GLD total-return live extension; the FRED GOLDPMGBD228NL named at proposal time is dead — log 2026-10-12).
• Period: 1985-01 → present (extend past source’s Jul-2020 cutoff to test robustness through gold’s 2024–25 run); monthly bars, positions from prior month-end data, executed next month.
• Costs: base run costless (matches source), sensitivity at 10 bps per switch.
• Engine: self-contained Python (pandas) under studies/gold-spx-ratio-rotation-replication/code/.

Method

1. Pin down ratio direction (body: SPX/gold above SMA → long SPX; verify against the published ratio chart gold-rotation.gif peak ≈5.5 in 2000 → confirms SPX/gold).
2. Replicate 20-month SMA rule; report CAGR, max DD, Sharpe vs SPX price B&H, SPX total-return B&H, gold B&H.
3. Grid SMA length ∈ {10, 15, 20, 30, 40} months — expose parameter sensitivity the author never tested.
4. Per-leg return decomposition (summed and log-accumulated) across start dates 1985/1990/1995/2000 — tests the “equal contributions” claim.
5. GFC facts: strategy equity high vs SPX low Mar-2009; rolling 5-yr relative performance.
6. Cross-check against gold-spx-switching-replication variants (does the Returnstream cash leg add anything over this simpler signal?).

Questions this study must answer or update

• dividend-exclusion-edge-hole (step 2 answers it directly)
• does-the-cash-leg-earn-its-keep (step 6)
• both-assets-down-regimes (add a both-down months table to results)

Results

(pending — no code run yet)

Verdict

(pending)

Next tests prompted

• Weekly/daily execution variant (monthly-bar smoothing).
• Add T-bill cash leg when both assets’ trend negative — does DD drop below 22%?
• If replication holds with dividends included, this becomes a rare vendor-claim that survives contact; promote confidence accordingly.