Bitcoin Correlations

Five lines, five macro stories. Each is Bitcoin’s rolling 90-day Pearson correlation with a single reference market: GLD for the digital-gold thesis, TLT for long-duration / safe-haven flows, SPY and QQQ for broad-market and tech-beta risk, and IWM for small-cap risk. The linear right axis spans −1.00 to +1.00 with dashed reference lines at ±0.5 and a solid zero baseline. There is no price overlay — the chart is dense enough with five correlation lanes.

Today’s strongest correlate is GLD at +0.81; the gold reading is +0.81, the Nasdaq reading is −0.05. The regime classifier — defined in the next section — reads Gold-linked. The series stretches from 15 Sept 2010 to the most recent close on 12 Jun 2026.

For each daily close t and each reference market X, the rolling 90-day Pearson correlation is:

ρ90d(BTC, X)t = Cov(rBTC, rX) / (σBTC · σX)

where r denotes daily log returns (ln(close / close−1)), and the covariance and standard deviations are computed over the trailing 90 aligned trading days. Equity returns are New-York-close-to-New-York-close; Bitcoin’s 24/7 returns are folded onto the equity calendar so the Friday-to-Monday bar carries the weekend price action. The 90-day window is the conventional choice — shorter windows whip on every macro surprise, longer windows lag too much to catch regime turns.

The five reference markets are exchange-traded funds rather than the underlying indices: GLD (SPDR Gold Shares), IWM (iShares Russell 2000), QQQ (Invesco QQQ Trust), SPY (SPDR S&P 500), and TLT (iShares 20+ Year Treasury Bond). ETFs are used so the dividend-and-distribution-adjusted total-return matches what an investable benchmark would actually deliver. The series itself is the CoinGlass 90-day correlation feed; the alignment and reconstruction match mainstream research-desk practice, and the full derivation lives on the methodology page.

The regime label on the reading row is deterministic from the per-asset coefficients. If the strongest correlate by absolute value is below 0.30, the regime reads Idiosyncratic — Bitcoin is moving on its own. Above that threshold the label follows the top correlate: a high TLT coefficient prints Risk-off linked; GLD prints Gold-linked; SPY, QQQ, or IWM print Risk-on linked. Because the named branches all require a positive top coefficient above +0.30, the only way to reach Mixed is a strongly negative top correlate — a deeply negative GLD or TLT reading below −0.30.

Reading every cycle anchor against the live series surfaces how unstable Bitcoin’s correlation profile is. The 2017 cycle top read deeply gold-decoupled — a strongly negative BTC–GLD coefficient — while the broad-equity (SPY) reading sat well above +0.8; by 2022, the network was coupled to the Nasdaq at coefficients above +0.6 for months. The 2024 pre-halving window saw gold and equities both elevated. Prices are our own daily closes; coefficients are pulled from the rolling-90-day series at the named date or the most recent prior close.

The single observation most competitor pages omit: Bitcoin’s gold correlation and Nasdaq correlation cross repeatedly. Each crossover is a regime flip from “digital gold” to “tech beta” or back. The table below walks the full daily-close history and emits a row for every regime that lasted at least thirty days. Gold-led regimes are the rarer of the two: across 5,750 trading days, Bitcoin spent 984 of them — 17.1% — with its 90-day GLD coefficient above its 90-day QQQ coefficient.

Three observations carry through every crossover regime on the table. First, the “digital gold” thesis is genuinely visible — just intermittently. Gold-led runs show up around banking stress, dollar weakness, and the months around the January 2024 spot-ETF launch. Larry Fink described Bitcoin in October 2024 as “an alternative to other commodities like gold”; the data places that statement inside a real but bounded regime, not a permanent reclassification.

Second, Nasdaq-led runs dominate the time-in-regime budget — especially since mid-2020. The Covid-era liquidity flush, the 2022 Fed hiking cycle, and the 2024–25 AI-led rally all produced multi-month windows where the QQQ coefficient sat above +0.5 and the GLD coefficient was a sideshow. Lyn Alden’s 2024 essay on Bitcoin as a global liquidity barometer argues the deeper driver is global money-supply growth: tech equities and Bitcoin both depend on cheap discount rates, and the equity-correlation is downstream of that shared sensitivity rather than the cause.

Third, the framing flips faster than narrative allows. Arthur Hayes frames Bitcoin as a global-fiat-liquidity barometer; the canonical institutional research note Bitcoin First positions it as an uncorrelated portfolio diversifier; Peter Schiff has argued Bitcoin is “negatively correlated” with gold in his debates with Saifedean Ammous. Each framing is a window in this data, not a constant. Today’s GLD reading of +0.81 and QQQ reading of −0.05 place the network in a gold-led regime; the next regime change will arrive on its own schedule.

An institutional research note on the factors driving gold and Bitcoin documents the long-run baseline plainly: across the post-2015 record, the 90-day BTC–GLD correlation has averaged about +0.10, with a peak of +0.57 and a trough of −0.37. The digital-gold thesis is a regime that fires, not a structural state.

Pearson is symmetric and direction-blind. A high BTC–QQQ reading tells you the two co-moved; whether tech dragged Bitcoin or Bitcoin dragged tech is a question this coefficient cannot answer. For causality, the appropriate tool is a Granger-causality test or a vector autoregression, not rolling Pearson; the coefficient sets the bound, not the direction.

Rolling windows smooth through tail-coupling. The March 2020 Covid liquidity flush is the canonical case: on 12 March 2020 Bitcoin fell roughly 40% in a single day alongside an equity crash, yet the 90-day coefficient barely moved at the time because the bulk of the trailing window was the placid pre-shock January and February tape. The two assets coupled violently in the tail and the rolling Pearson reported it three months late. Crisis correlation needs a shorter window or an explicit drawdown overlay; this series will always understate it.

The 2023 banking-stress gold signal was brief and easy to over-read. When Silicon Valley Bank failed on 10 March 2023 and USDC briefly depegged that same weekend, Bitcoin rallied hard while gold caught a haven bid — the textbook “digital gold” print. But the GLD lead lasted only weeks before the QQQ coefficient reasserted itself into the summer-2023 risk-on tape. Anyone who read that spring as a permanent reclassification was wrong by the next quarter; the crossover table dates exactly how short the window was.

The reference universe is selective. The five ETFs cover gold, US equities (broad, tech, small-cap), and long bonds. They do not cover the dollar index (DXY), oil, EM equities, the yen carry, or credit spreads — each of which has had a real claim on Bitcoin’s correlation surface in specific regimes (the August 2024 yen-carry unwind is a clean example the chart cannot see). Add an asset and you may find a higher coefficient than anything on this page; the absence is not a denial.

Calendar-alignment loses weekend information. Bitcoin trades 24/7; equity ETFs do not. Folding the 24/7 returns onto Monday-to-Friday closes parks weekend price action inside the Friday-to-Monday bar. Major weekend moves (the October 2023 Hamas-attack repricing, the March 2023 USDC depeg) hit the chart as a single elevated Monday return rather than as the multi-day events they were. The coefficient absorbs this; the narrative attribution does not.

If you size a portfolio, treat the digital-gold and tech-beta framings as regime-dependent rather than structural. Today’s coefficients say something about today’s diversification properties; nothing in the crossover record argues they will hold through the next macro regime change. Position-sizing on a permanent +0.5 Nasdaq correlation, or on a permanent zero, will both be wrong inside two years — the table shows the regime turning over on a multi-month, not multi-year, clock.

If you trade macro, the crossover table is the cleanest catalogue of Bitcoin’s regime-shift cadence on this site. Pair it with global M2 for the liquidity lens, with dominance for the cryptoasset-internal share, and with spot ETF flows for the institutional-flow lens. Crossovers from gold-led to Nasdaq-led inside a thirty-day window have historically marked the start of a wider risk-on impulse; the reverse direction has marked dollar-stress windows.