The Gold Methodology v1.0
The Gold Barometer is a composite of seven measured conditions, each ranked as a percentile against its own history since 1971. Weights, evidence grades, and the exact rules for missing data and revisions are published on this page. Methodology version v1.0.
The Barometer answers one question: how do today's gold buying conditions compare with every day since 1971? It looks at seven conditions that history says matter to a long-term buyer, ranks each one against its own past, and averages the ranks with published weights. A high score means conditions today look like conditions that historically preceded better long-term returns. A low score means the opposite. It is a measurement, not investment advice.
1. The seven pillars, weights, and evidence grades
Weights are frozen at v1.0. The retail-premiums pillar ramps in from 5% to 10% as its own history accrues (12 months, approximately August 2027). All other weights renormalize accordingly.
| # | Pillar | Weight | Evidence grade |
|---|---|---|---|
| 1 | Real rates | 25% | strong, with a documented 2022 regime break |
| 2 | Entry price | 20% | strong on both sub-signals, at different horizons |
| 3 | Structural demand | 15% | strong for central banks (peer-reviewed sanctions channel, three years above 1,000 tonnes); weak for ETF flows (coincident with price) |
| 4 | US dollar | 10% | moderate |
| 5 | Positioning | 10% | weak to moderate |
| 6 | Volatility | 10% | weak as a return predictor |
| 7 | Retail premiums | 0% (ramp 5%) | cost argument incontestable; predictive value anecdotal |
2. The percentile transformation
Every pillar sub-score is the percentile rank of the latest smoothed value inside the full history of that indicator up to today. A percentile of 80 means today is higher than 80% of every previous observation of that pillar. Percentiles make pillars comparable, are robust to outliers, and satisfy the Cboe GVZ non-reversibility constraint recorded in the data-rights page.
Each pillar is then oriented so that a higher sub-score maps to conditions that historically preceded better long-term forward returns for a buyer. Orientations are set by documented economic mechanism, never curve-fit to gold returns. Real rates: low is better for a buyer. Entry price: cheap in real terms and below trend is better. Structural demand: rising is better. Dollar: weak is better. Positioning: washed-out longs are better (contrarian at extremes). Volatility: low is better as an entry-quality signal. Premiums: low is better (lower all-in cost).
3. Composite
The composite score is the weighted average of the available pillar sub-scores. A daily 5-day median smoother is applied to daily inputs; nothing else. Details of every pillar's measurement, orientation, and evidence are on that pillar's own page.
4. Zones
| Zone | Band |
|---|---|
| Historically very favorable | 80 to 100 |
| Favorable | 60 to 79 |
| Mixed | 40 to 59 |
| Unfavorable | 20 to 39 |
| Historically very unfavorable | 0 to 19 |
Language rule: "conditions", never buy or sell or should. Email alerts fire on zone change; see the home page signup form for the alert channel.
5. Missing-data policy
Every pillar has a published freshness limit. If the underlying source falls silent inside that limit, the last available value is carried forward and flagged aging. If the source stays silent beyond the limit, the pillar drops from that day's composite and remaining weights renormalize. The site publicly shows the count "N of 7 pillars used today" on the home page plaque and on the provenance page.
- Daily series: staleness limit 7 business days.
- Weekly series (COT): staleness limit 10 days.
- Monthly series: staleness limit 45 days.
6. Versioning and no-silent-revision policy
v1.0 is frozen at launch. Any change to weights, orientations, or the formula bumps the version number and appears in the changelog below. A 90-day parallel run publishes both versions before the new one becomes canonical. See the corrections page for the record of any published revision to values already released.
Changelog
| Version | Date | Change |
|---|---|---|
| v1.0 | 2026-08-05 | Initial published version. Seven pillars, weights 25/20/15/10/10/10/10 with retail premiums ramping from 5% to 10% as its own history accrues. |
7. Limitations and where this framework has been wrong
Every quantitative claim below is dated and sourced. The point of this section is to state the limits of the instrument in numbers, not adjectives.
7.1 Real-rate pillar is regime-dependent
The published caveat on pillar 1 is that the co-movement between real rates and gold is regime-dependent, and the site publishes the numbers. Chicago Fed Letter 464 (2021) documents that the inverse relationship is essentially absent before 2001. RBC Wealth Management reports the R-squared at 69% for 1997 to 2004, 84% for 2005 to 2021, 3% for 2022 to 2023, and 7% for 2024. JP Morgan Asset Management reports a similar figure moving from about 85% (1990 to 2021) to about 16% since 2022, with the beta flipped. Own backtest, 1971 to present at monthly cadence: correlation of the real-rates sub-score to subsequent 12-month nominal gold return is +0.237 for 1971 to 2000, -0.368 for 2001 to 2021, -0.804 for 2022 to 2024. The 1-year predictive claim has never been positively correlated in a full-era measurement.
7.2 The top zone is empirically untested
The top zone (80 to 100 "Historically very favorable") has N=0 across the full 1971-present backtest. The composite never crossed 80 in 667 monthly observations. That zone is therefore untested empirically. The instrument does not claim forward statistics for it until at least one such month materializes and its outcome is observed.
7.3 Band ordering inverts at 1 year
Median nominal 12-month returns by zone across 1971 to present: Favorable 5.2%, Mixed 4.6%, Unfavorable 8.6%. Short-run gold moves and the score point in opposite directions on median: momentum drags the score lower right when returns have started to accelerate, and vice versa. This is a genuine limit of the framework for anyone reading it as a tactical 1-year buy signal, and it is surfaced in the buyer copy. The long-horizon real thesis does hold: median 5-year real returns of 61.8% in the Favorable zone versus -13.9% in the Unfavorable zone.
7.4 COT contrarian-at-extremes is practitioner lore
The academic contrarian leg in COT data sits on hedgers (Wang, 2003), not on managed money. The pillar uses managed money because it is the cleanest data with the deepest history, and discloses the caveat. Weight is 10% for this reason.
7.5 No predictive evidence for volatility or premium LEVELS
There is no published evidence that implied-volatility levels (GVZ) predict forward gold returns. There is no return-prediction claim for retail premium levels either. Both pillars are framed as risk (volatility) and cost (premiums) gauges, not directional forecasts.
7.6 Retail premiums are regional
The retail-premiums pillar is US-only. In March 2020, Western coin demand rose 36% year-over-year (WGC Q1 2020) while India traded at a $70 per-ounce discount (SBMA). A single global premium signal does not exist; this site publishes the US read.
7.7 ETF flows are coincident
ETF flows follow price rather than lead it (WGC ETF flow data). The structural-demand pillar carries the ETF flow signal at low weight for this reason, and the central-bank sub-signal (peer-reviewed sanctions channel) does the heavier lifting.
8. Considered and not included
The v1.0 pillar set is deliberately small: seven measured conditions, each with a documented mechanism and a published evidence grade. The following candidates were evaluated and rejected for v1.0. Some are logged as candidates for v1.1.
| Candidate | Status | Reason |
|---|---|---|
| Inflation breakevens (10Y BEI) | Rejected | Already inside real rates by construction (nominal Treasury minus TIPS). Adding it would be double-counting. |
| M2 growth / fiscal deficits (debasement channel) | Rejected | The debasement channel is already captured by the CPI-deflated valuation anchor in the entry-price pillar. Adding a monetary aggregate would compound the same channel twice. |
| 10Y term premium (ACM or similar) | v1.1 candidate | Correlated with the rates pillar and would risk double-counting. The post-2022 regime is a case for considering a term-premium overlay; logged for v1.1. |
| Geopolitical risk index (Caldara-Iacoviello) | v1.1 candidate | Documented driver of gold at event dates but event-driven and untimeable at daily cadence. Logged for v1.1 as an event overlay, not a pillar. |
| Miner all-in sustaining costs (AISC) | Rejected | Quarterly, laggy, industry-reported. Weak evidence of predictive content and long publication lag. |
| Seasonality (month, day-of-month) | Rejected as a pillar | Real effect but too small to weight without dominating the composite. Covered editorially in the seasonality satellite page. |
| Physical market internals (lease rates, EFP spreads) | Rejected | Data is licensed or fragile and the retail-premiums pillar already proxies physical tightness at the consumer end. |
9. Attribution
Data sources are producer-first (Board of Governors of the Federal Reserve System, US Bureau of Labor Statistics, US Commodity Futures Trading Commission, International Monetary Fund, World Bank, Cboe). See the provenance page for the per-source status board, license class, and freshness limits. Cboe Gold Volatility Index (GVZ) is used internally under proprietary rights and is credited here in methodology text only; raw GVZ levels are not republished by this site.
See the history page for the point-in-time backtest 1971 to present, per-zone forward returns, drawdowns, DCA versus lump sum, and the failures section in full.