Cross-asset · ETH/BTC vs copper/gold

The Copper/Gold
of Crypto

Copper/gold is the macro world's internal risk dial: growth metal over haven metal. ETH/BTC is crypto's version of the same question. If both answer to one real-rate drum, they should rise and fall together — and when they stop agreeing, one of them is lying. This page keeps score.

Updated 08 September 2026 · 20:37 UTC · Prices as of 08 September 2026 · Sources: Yahoo Finance (BTC-USD, ETH-USD, HG=F, GC=F), FRED (DFII10)

The divergence gap

Each ratio is z-scored against its own trailing year, then the two z-scores are subtracted. The alarm trips at ±2σ — a level gap can never trigger it, only a genuine disagreement about direction.

-1.52σWITHIN BAND

Current gap: ETH/BTC minus copper/gold, in standard deviations. The 90-day rolling correlation of their daily changes is -0.06 (full 3-year sample: +0.08). 8 divergence episodes beyond 2σ are on record for this window.

ETH/BTC ratio0.0317Crypto's internal risk dial.
Copper/gold ratio0.0015Macro's internal risk dial (HG=F ÷ GC=F).
90d corr of changes-0.06Are they moving to one drum right now?
10y TIPS (DFII10)2.43%Real-rate drum, observed 04 September 2026 (latest published FRED print; carried forward between prints).

Two dials, one chart

Both ratios rebased to 100 at the start of the window so their shapes are comparable. They will not sit on top of each other — the question is whether they turn at the same times.

ETH/BTC (rebased)Copper/gold (rebased)

When the drumbeat fades

The 90-day rolling correlation of daily ratio changes, then the z-score gap itself with the ±2σ alarm band. Episodes above the band are the rows in the ledger below.

90d correlation
Z-score gap±2σ alarm band

Divergence ledger

Every run of at least three trading days with the gap beyond 2σ: when it opened, when it closed, how wide it got, and which ratio was the rich one.

OpenedClosedDaysWidest gapRich side
2025-03-112025-03-1772.19σCopper/gold rich
2025-03-192025-03-2682.38σCopper/gold rich
2025-07-312025-10-28903.78σETH/BTC rich
2025-10-302025-11-0352.18σETH/BTC rich
2025-11-052025-11-20162.34σETH/BTC rich
2025-11-242025-11-2852.26σETH/BTC rich
2026-01-262026-01-2942.64σETH/BTC rich
2026-06-012026-08-17783.23σCopper/gold rich

The real-rate drum, tested

If the one-drum hypothesis holds, changes in the 10-year TIPS yield should lean on both ratios in the same direction. Correlation of daily changes, full sample and latest 90 days, printed side by side so neither window can flatter the story alone.

DFII10 vs ETH/BTC · full / 90d-0.06 / -0.20Correlation of daily changes: 10-year TIPS yield against crypto's risk dial.
DFII10 vs copper/gold · full / 90d+0.03 / -0.02Same test against the macro dial — the benchmark the hypothesis has to clear.

Methodology & limitations

Method. Daily closes from Yahoo Finance (3-year window, export timestamped 08 September 2026 · 20:37 UTC). Copper/gold is HG=F ÷ GC=F on futures trading days; ETH/BTC is ETH-USD ÷ BTC-USD carried onto those days. Correlations run on daily log changes of the ratios. The divergence gap z-scores each log ratio against its own trailing 252 days and subtracts. Real rates are FRED series DFII10 (10-year TIPS, daily), observed 04 September 2026. An episode is a contiguous run beyond ±2σ lasting at least 3 days.

Limits. ETH/BTC is dominated at times by crypto-internal events (upgrades, ETF flows, staking) that no real rate can explain; copper/gold carries Chinese credit and supply shocks. Front-month copper embeds roll yield. A 90-day window is short enough to be seduced by a single shared shock. Correlation of changes is not co-movement of levels, and neither is causation. If the FRED or price feed fails, the affected section is labelled unavailable — missing data is never shown as zero.

Research by for Charlie Quant Lab · Updated