Gold has lost a property it genuinely had, and Bitcoin never demonstrated: the same test across 239 shock days

Executive summary

• The definition is identical, unchanged in every respect: a “shock day” is any session in which Brent crude closes up 3% or more from the prior session's close. It is the definition we used to measure gold, and retaining it unchanged makes the two results directly comparable.
• The averages are close. Across 239 shock days, Bitcoin's average same-day return was +0.316%, and it rose on 56.1% of those days; gold, over the same time series, returned +0.399% and rose on 64.4% of them.
• The difference is in dispersion, not the average — and that is the central point. The standard deviation of Bitcoin's return on those days is 4.69, versus 1.06 for gold, or roughly four times higher. The t-statistic is 5.85 for gold and 1.04 for Bitcoin.
• In non-statistical terms: gold's tendency to rise on oil-shock days was a genuine tendency that could be relied upon; Bitcoin's was positive on average but swamped by noise — an average that cannot support a conclusion because the volatility around it overwhelms it.
• In 2026, the two assets became the same trade. Across 40 shock days this year: Bitcoin −0.552% and gold −0.558%. The figures are effectively identical. This is convergence in weakness, not safety.
• The rates channel reaches gold, but has yet to be shown to reach Bitcoin. Gold's return correlation with the change in the two-year yield over the past 30 sessions is −0.403 and statistically significant; Bitcoin's is −0.254 and not significant. For all of 2026, Bitcoin's correlation falls to −0.070; in 2025 it was positive — that is, the opposite sign.
• The live test took place today: monthly core inflation came in at 0.3% versus 0.2% expected — the only line that missed expectations — sending Bitcoin to 76,046 dollars within minutes — or −1.28% — before it not only recovered the decline but ended the same hour above its starting point at +1.29%. A sharp response fully reversed within one hour, not a stable repricing.