Graveyard / Macro
Gold versus real yields
Can the gap between gold and real yields be traded?
The idea
Gold usually moves against real interest rates and the dollar. When GLD is far above or below what those two predict, bet on the gap closing, or on the move continuing.
How we tested it
| Data | GLD against the 10-year real yield and a broad dollar index. A 156-week regression residual z-score (the entry threshold is not stated in the note). Contrarian (primary) and follow variants, frozen in advance. |
|---|---|
| Period | 2010 to November 2026. Holdout January 2019 onward. |
| Costs | Costs and financing charged, plus a 1.5 times stress case. |
| Holdout | January 2019 onward, one run, nothing changed afterwards. |
| Control | Random twins: the signal shifted 1 to 8 years, 300 times. Bonferroni alpha 0.025. |
Result
| Holdout Sharpe, contrarian / follow | −0.67 / +0.64 |
|---|---|
| Holdout annual net | −10.4% / +10.0% |
| Holdout entries | 10 |
| Twin p, contrarian / follow | 0.89 / 0.146 |
| Average Sharpe of the random twins | 0.39 |
Why it failed
Only 10 entries in the holdout, so the verdict is insufficient evidence rather than a clean failure. The follow variant's +0.64 is explained by gold's drift in 2019-2025: twins with no information averaged 0.39. The contrarian variant lost.
What would make it worth retesting
At least 25 holdout entries (for example by adding silver and miners, or five more years) and a twin p below 0.025.
Test id: goldgap-gld-event-v1. Past results of a backtest say nothing certain about the future.