Gold's 30-Day Correlation With Bitcoin Reached 0.64. With Long Treasuries It Is 0.10.

By Kresmion Research, July 23, 2026

Yemen's Iran-aligned Houthis said on Wednesday that they had attacked two Saudi oil tankers in the Red Sea. Saudi authorities confirmed damage to one of them, the Encelia. The UK Maritime Trade Operations office reported a vessel struck about 70 nautical miles southwest of Al Shuqaiq by what it described as an unknown projectile, causing a fire on board, with no casualties and no environmental damage. The second strike the Houthis claimed has not been confirmed.

Separately, and before that strike, President Trump had warned that any time Iran shoots at a ship in the Strait of Hormuz, the United States would bomb and destroy one bridge or power plant. That warning concerns Iran and the Strait of Hormuz. The tanker attack was claimed by the Iran-aligned Houthis in the Red Sea. Different actors, different waters.

Crude rose. As of 12:52 UTC on Thursday, West Texas Intermediate was trading near 90.62 dollars, up about 4.4 percent on the day and about 14.8 percent over five sessions. The VIX was up about 12 percent at 18.69.

Gold was down about 1.8 percent at 4,072 dollars. Silver was down about 2.9 percent. Those are intraday readings taken at the same moment, before the US cash session opened, so they are a snapshot of Thursday morning rather than a closing level.

The reflex is to call that a failed safe haven. One day of data does not support the reflex. Across 2,520 sessions since July 2016, WTI rose 4 percent or more on 101 days, and gold closed lower on 43 of them. That is 42.6 percent of the time. Gold's average return on those days was positive but small, 0.14 percent against an unconditional daily average of 0.05 percent, and the full-sample daily correlation between gold and WTI is 0.075. A decline of 1.8 percent is about 1.7 standard deviations against gold's 1.047 percent daily deviation, and if anything that understates the move, since a partial session should carry less variance than a full one. On its own it is an ordinary day.

What is not ordinary is the company gold has been keeping for the past two months.

Key takeaways

What the correlation tape shows

Kresmion computes rolling 30-day correlations across 843 asset pairs daily. On the July 22 snapshot, the last complete run, gold was the second most broken symbol on the board behind energy: 11 of its 41 pairs registered as correlation breaks.

Every one of those 11 breaks moved in the same direction. There were zero sign flips. Measured against the engine's reference baseline for each pair, gold's correlation to Bitcoin went from 0.223 to 0.637, to XRP from 0.270 to 0.759, to Dogecoin from 0.178 to 0.685, to Ethereum from 0.198 to 0.612, and to Solana from 0.215 to 0.667.

Those baselines are the trailing 252-day mean of the same 30-day correlation, not the reading on a particular past date. Measured instead against a fixed date, May 20, gold's correlation to Bitcoin has moved from 0.179 to 0.637.

Aggregated across 19 crypto majors, gold's average correlation is 0.5409. Against the 513 daily snapshots Kresmion holds back to July 2024, that is the 99.8th percentile, with a median of 0.1076.

It is worth being precise about what that record means. The all-time high on this measure is 0.5940, and it was set on July 20, two days earlier. The six highest readings in the entire 513-day history all fall within the past eight days. The current reading is not an isolated spike against a calm past. It is near the top of an episode that has been building since June, and the record it is being measured against belongs to the same episode.

The path from May: 0.146 on May 20, 0.443 on July 6, 0.475 on July 15, 0.541 on July 22.

The control that matters

A correlation cluster usually means one of two things. Either a specific relationship changed, or everything started moving together and the pair in question came along for the ride. The second explanation is far more common and it is the one that kills most correlation stories.

It does not hold here, and the cleanest way to show that is to put the S&P 500 through the identical measure, on the identical 19 symbols, on the identical dates.

The S&P's average 30-day correlation to those same 19 crypto majors is 0.3573. Against its own 513-day history that is the 42.3rd percentile, below its median of 0.3856. The equity relationship with crypto is unremarkable right now.

Gold's is 0.5409. The gap between the two, gold minus the S&P, is 0.1836, and that is the widest reading in the entire 513-day record. The median gap is -0.2476, meaning that in the ordinary state of the world the S&P is the more crypto-correlated of the pair, by about 0.25 correlation points. Gold has been the higher of the two on 5.5 percent of all sessions since July 2024, and on 18 of the last 30.

Gold's relationship with fear moved the same way. Its correlation to the VIX is -0.6249, the 3.6th percentile of 816 daily snapshots back to April 2023, against a median of -0.1393. An asset held as insurance typically rises when volatility rises. This one has been doing the opposite for most of the summer. That particular figure carries a caveat covered further down: it was surfaced by a scan across 843 pairs, which makes an extreme percentile much less surprising than it looks.

The mechanism

One genuinely independent constraint bears on this, and one weaker observation that should not be leaned on. Taking them in that order matters, because they are not equal evidence.

The independent one is positioning. The CFTC's Commitments of Traders report for July 14 puts gold non-commercial net length at 186,682 contracts on 383,689 of open interest, 48.65 percent, with a 52-week z-score of 0.45 and a 3-year COT index of 34.6. That is the middle of the range. Silver sits lower still at 25.6. This is a government publication collected separately from any price feed, so it is a real second source rather than another view of the same tape. If a speculative mania in gold futures were driving the convergence with crypto, positioning would be stretched. It is not, which points toward spot and exchange-traded fund demand, the access route gold and crypto share.

The weaker observation is gold's decoupling from duration. Its correlation to long Treasuries fell from 0.703 on May 20 to 0.0989 on July 22. That looks like gold leaving the rates complex, and it is tempting to build on. It does not hold up, for reasons set out in the next section.

The real-rate explanation is also not dead, and the article would be dishonest to bury it. As of July 22 the 10-year real yield stood at 2.35 percent against 1.88 percent a year earlier, up 47 basis points, while the 10-year breakeven inflation rate was 2.26 percent, below the 2.34 percent of a year ago. A zero-coupon asset priced off real rates would have struggled against that, and gold did struggle. It is down 13.0 percent over three months, 15.4 percent over six, and 23.2 percent from its record close of 5,318.40 on January 29. Gold is up on the week, which runs the other way, but a week is the wrong horizon to set against a year of real-rate movement.

So the rates story accounts for gold's level rather well. What it does not account for is the company gold keeps. An asset selling off on rising real rates has no particular reason to tighten its correlation with Bitcoin to a two-year high in the same window, while the S&P's correlation to the same coins sits at its own median.

That gap between the two is where the inference sits: the price at the margin looks like it is being set by something closer to the crypto bid, a position indifferent to short-term volatility, expressed through spot and fund vehicles rather than futures.

That last step is an inference, and it should be read as one. Kresmion can observe what moves together and what futures positioning shows. It cannot observe who is buying. The correlation and positioning data are consistent with a changed marginal buyer, and they are also consistent with an unchanged set of buyers responding to a common impulse that happens to move gold and crypto the same way. The tests at the end are how to tell those apart.

What cuts against this read

The Treasury leg is the weakest part of the story and it fails on inspection.

Gold's correlation to long Treasuries fell from 0.703 to 0.0989 over the two months. But the S&P 500's correlation to long Treasuries fell almost exactly as far over the same window, from 0.773 to 0.199. Long Treasuries decoupled from gold and from equities at close to the same rate, so the collapse in gold's bond correlation is substantially a bond-market event and cannot be claimed as evidence about gold specifically. What survives is narrower: when the equity complex and the bond complex separated, gold tracked the equity side.

Gold has not been performing well. It is down 23.2 percent from its January record, 13.0 percent over three months and 2.1 percent over the past month, against a gain of 20.1 percent over twelve. So this is not a story about an asset going up for a new reason. Gold has been falling, roughly in line with what rising real yields would do to it, and the notable part is the correlation structure rather than the direction.

The precious metals complex is internally intact. Gold against silver is 0.897, up from a 0.799 baseline. Gold has not detached from its own asset class, which argues against reading it purely as a crypto proxy.

The dollar relationship is unchanged. Gold against the dollar index is -0.4072 against a baseline of -0.4124. A full repricing of gold as a risk asset would show up as a looser monetary link. It has not moved.

There is also a search-space caveat. The gold and VIX correlation extreme was surfaced by a scan across 843 pairs, and a 96th or 99th percentile reading found inside a scan that large is much less surprising than the same reading found in isolation. The gold and crypto aggregate, the S&P control and the CFTC cross-check were tested afterward as directed questions rather than pulled from the scan, which is why they carry more weight here than the scan itself.

What would change the read

Three observable things would confirm or break this.

An equity drawdown that does not hit crypto is the cleanest test. If gold falls with stocks while crypto holds, the shared-buyer explanation survives. If gold rises while stocks fall, the hedge is intact and the summer's co-movement was a coincidence of flows.

The next Commitments of Traders release, due Friday July 24 covering July 21 positioning, tests the independent leg. A sharp jump in gold speculative length would weaken the claim that futures money is not driving this.

A return of gold's correlation to long Treasuries toward its May reading of 0.70, alongside a fading crypto correlation, would indicate the rates buyer came back and this was a rotation rather than a change in who holds the asset.

Frequently asked questions

Does a high correlation between gold and Bitcoin mean they are the same asset?

No. Correlation measures whether two prices move together over a window, not whether the assets are equivalent. Gold's 30-day correlation to Bitcoin at 0.637 means the two have tended to move in the same direction recently. Their volatility, liquidity, custody and holder base remain very different.

Why is a 30-day correlation of 0.54 across crypto majors significant?

Because of where it sits in its own history. Across 513 daily snapshots since July 2024 the median reading is 0.1076. The current 0.5409 is the 99.8th percentile of that distribution. The record of 0.5940 was set on July 20, and the six highest readings all fall within the past eight days.

Did gold fall because of the Red Sea attack?

The single-day move does not support that conclusion either way. Gold has fallen on 42.6 percent of the 101 days since 2016 when WTI rose 4 percent or more, and Thursday's intraday decline of about 1.8 percent is roughly 1.7 standard deviations, inside the normal daily range.

Is this just every asset correlating with every other asset?

The control argues against it. The S&P 500's average correlation to the same 19 crypto majors is 0.3573, the 42.3rd percentile of its own 513-day history and below its median. Gold's is 0.5409. The gap between them is the widest of all 513 snapshots, against a median gap that runs the other way.

Has gold been rising during this period?

No. Gold is down 13.0 percent over three months and 23.2 percent from its January 29 record close, though it remains up 20.1 percent over twelve months. The finding here concerns the company gold keeps. Its direction over that period is explained reasonably well by rising real yields.

What is the CFTC Commitments of Traders report and why does it matter here?

It is a weekly US government publication showing how futures positioning is split between commercial hedgers, large speculators and smaller traders. It is collected separately from price data, so it serves as an independent check. Gold speculative positioning at a 3-year index reading of 34.6 is mid-range, which argues against a futures-driven explanation.

Sources

Correlation figures are Kresmion 30-day rolling calculations as of the July 22, 2026 close, the last complete 843-pair run. Price quotes are intraday as of 12:52 UTC on July 23, 2026. Gold return horizons and the record close are from daily futures closes through July 21, 2026. Positioning is CFTC data for the week ended July 14, 2026. Rates are as of July 22, 2026.

External reporting cited above:

• Houthi claim of attacks on two Saudi tankers, and Saudi confirmation of damage to the Encelia: Al Jazeera, July 22, 2026
• UK Maritime Trade Operations warning on the vessel struck southwest of Al Shuqaiq: UKMTO advisory 095/26
• President Trump's warning regarding Iran and the Strait of Hormuz: CNBC, July 22, 2026
• Oil market reaction: CNBC, July 23, 2026
• CFTC Commitments of Traders release schedule: CFTC