Bitcoin Did Not Break Its Correlation With Gold, It Broke Its Beta: 30-Day BTC/Gold Return Correlation Is +0.65, a 2-Year High (99th Percentile), 20-Day BTC/S&P +0.62, 30-Day BTC/DXY -0.43; the +33% vs -1%/+1% Gap Since Aug 18 Is BTC's 1.45x Beta to Gold at 54% Vol, Not Decoupling

price

The claim, and what the return data actually shows. Santiment's 23 Sept (05:08 UTC) post says Bitcoin "has completely broken away from stocks and precious metals": BTC market cap +36.0% since 18 Aug versus the S&P 500 +0.8% and gold -1.5%. CryptoPotato relayed it the same morning as Bitcoin having "broken its correlation with gold, stocks and the dollar". But the chart Santiment links is a price chart (btc-spx-gold-prices), not a correlation series, and a correlation computed from returns tells the opposite story. Using Yahoo Finance daily closes (BTC-USD, GC=F, ^GSPC, ^NDX, DX-Y.NYB; 2 years to 22/23 Sept 2026; Pearson on log returns, common trading days): the 30-day BTC/gold correlation is +0.65, it peaked at +0.71 on 17 Sept, the highest reading in the entire 2-year sample (median +0.11, range -0.40 to +0.71, current reading at the 99th percentile). The 90-day BTC/gold figure is +0.49 (97th percentile). The 20-day BTC/S&P correlation is +0.62 and 20-day BTC/Nasdaq-100 +0.64. The 30-day BTC/DXY correlation is -0.43, at the 10th percentile of two years, i.e. the inverse-dollar relationship is stronger than usual, not weaker. Nothing "broke" in September on the measure the headline names.

Why performance can diverge by 34 points while correlation sits at a high. Correlation measures whether two assets move in the same direction, not by how much. BTC's 30-day beta to gold is 1.45 and to the S&P 500 1.85; realised vol over the window is 54% annualised for BTC against 20% for gold and 10% for the S&P. Look at BTC's six biggest days since 18 Aug: 21 Aug +7.0% (gold +2.4%), 19 Aug +6.8% (gold +2.8%, DXY -0.8%), 21 Sept +6.5% (gold -0.9%, Nasdaq-100 +2.8%), 18 Sept +5.7% (gold +0.6%), 20 Aug +5.3% (gold +0.6%), 3 Sept +5.0% (gold +2.8%). Five of the six were gold up-days; the one exception, 21 Sept, was the Nasdaq's best session of the stretch and the $999M ETF inflow day. An asset that rises roughly three times as much as gold on shared up-days compounds a large performance gap while remaining tightly correlated. The price-level correlation over the last 25 sessions is -0.17 for gold and +0.05 for the S&P, and that is precisely what a price chart shows: trend divergence. Trend divergence at high return-correlation is a beta and drift story, not decoupling.

What did change, and when. The genuine shift is a summer de-rating of the equity link, not a September event. The 90-day BTC/Nasdaq-100 correlation is +0.30, the 9th percentile of two years, down from +0.40 on 1 July; 90-day BTC/S&P is +0.34 (12th percentile, from +0.46 on 1 July). The Kobeissi Letter flagged the same numbers on 5 Sept (90-day BTC/gold +0.50, near the 2020 pandemic high; BTC/Nasdaq-100 ~0.30, a one-year low) and tied the acceleration to the Treasury's 19 Aug announcement. That announcement is real and primary: Treasury press release sb0607 raises the maximum size of long-end liquidity-support buybacks (10-20y and 20-30y sectors) from $2B to at least $4B per operation, effective 9 Sept through the 4 Nov refunding. On 19 Aug BTC rose 6.85%, gold 2.78%, and DXY fell 0.83%; that is a liquidity-and-debasement day on which the two "hedges" moved together. The early-September "digital gold" coverage landed after the gold correlation had already risen, and the 30-day reading kept climbing to its 17 Sept peak and is still +0.65. CryptoPotato's "reversal in a matter of weeks" does not exist in return data.

The two stress tests confirm co-movement, not independence. The FOMC raised the target range 25bp to 3.75-4.00% on 16 Sept (12-0 vote, "inflation remains elevated"): BTC +0.7%, gold +1.25%, S&P -0.45%. CLARITY cloture failed 49-50 on 15 Sept: BTC -3.3%, gold -0.4%, S&P -0.45%, Nasdaq-100 -0.65%. On both days gold and BTC shared a sign; BTC simply moved more. Context on levels: gold's own peak was $5,318 (GC=F close, 29 Jan 2026); at $4,355 on 23 Sept it is 18% below that and flat since 1 Sept ($4,396). BTC at roughly $85,500 is 31% below its 6 Oct 2025 high of $124,753 and 46% above the 30 June 2026 low of $58,559. Both are drawdown-recovery assets responding to the same rates and liquidity impulses, which is exactly what a high correlation looks like.

Numbers in circulation to distrust. 247wallst (15 Sept) states BTC's "QQQ correlation flipped from +0.80 to -0.30". No 30-day window of daily returns in 2026 shows BTC/Nasdaq-100 below +0.08 (13 May) or above +0.70 (9 Mar); no 90-day window is below +0.24. Santiment's +36.0% market-cap gain versus my close-to-close +33.2% ($64,681 on 18 Aug to $86,172 on 22 Sept) is intraday timing (BTC printed above $87,000), not an error. The parent card's "improved diversification value" is backwards for gold: on the current data BTC's marginal diversification against gold is the lowest in two years. Against equities it is fair: a 90-day reading of +0.34 is the low end of the 2-year range (+0.21 to +0.59).

Base rate and what it means for positioning. Rolling correlations mean-revert: the 2-year median for 30-day BTC/gold is +0.11, so the +0.65 reading is more likely to fall than rise. The tradeable implication is the reverse of the headline. If BTC is at peak correlation with gold with a 1.45 beta, and gold is being repriced for "tighter for longer" after a 12-0 hike, then a gold drawdown is a BTC risk rather than something BTC is immune to. The bullish reading survives only in the narrower form: BTC's link to equities is at a 2-year low while its link to the debasement trade is at a 2-year high, which is a change in what BTC is correlated with, not a break from correlation. Method note: all correlations here are reproducible from the Yahoo chart endpoint cited below by swapping the ticker; the same endpoint supplied GC=F, ^GSPC, ^NDX and DX-Y.NYB.

Sources (6)

https://cryptopotato.com/bitcoin-just-broke-its-correlation-with-gold-stocks-and-the-dollar-what-changed/https://x.com/SantimentData/status/2102626192746549403https://x.com/KobeissiLetter/status/2096269626015629679https://home.treasury.gov/news/press-releases/sb0607https://www.federalreserve.gov/newsevents/pressreleases/monetary20260916a.htmhttps://query1.finance.yahoo.com/v8/finance/chart/BTC-USD?range=2y&interval=1d

AI Research

Key Takeaway

Santiment's 'Bitcoin broke away from gold and stocks' chart shows cumulative price, not correlation: 30-day BTC/gold return correlation is +0.65 (2-year high, 99th percentile) and 20-day BTC/S&P +0.62. The 34-point performance gap since 18 Aug is BTC's 1.45x beta and 54% vol, which makes a gold drawdown a BTC risk, not something BTC has escaped.

The claim, and what the return data actually shows. Santiment's 23 Sept (05:08 UTC) post says Bitcoin "has completely broken away from stocks and precious metals": BTC market cap +36.0% since 18 Aug versus the S&P 500 +0.8% and gold -1.5%. CryptoPotato relayed it the same morning as Bitcoin having "broken its correlation with gold, stocks and the dollar". But the chart Santiment links is a price chart (btc-spx-gold-prices), not a correlation series, and a correlation computed from returns tells the opposite story. Using Yahoo Finance daily closes (BTC-USD, GC=F, ^GSPC, ^NDX, DX-Y.NYB; 2 years to 22/23 Sept 2026; Pearson on log returns, common trading days): the 30-day BTC/gold correlation is +0.65, it peaked at +0.71 on 17 Sept, the highest reading in the entire 2-year sample (median +0.11, range -0.40 to +0.71, current reading at the 99th percentile). The 90-day BTC/gold figure is +0.49 (97th percentile). The 20-day BTC/S&P correlation is +0.62 and 20-day BTC/Nasdaq-100 +0.64. The 30-day BTC/DXY correlation is -0.43, at the 10th percentile of two years, i.e. the inverse-dollar relationship is stronger than usual, not weaker. Nothing "broke" in September on the measure the headline names.

Why performance can diverge by 34 points while correlation sits at a high. Correlation measures whether two assets move in the same direction, not by how much. BTC's 30-day beta to gold is 1.45 and to the S&P 500 1.85; realised vol over the window is 54% annualised for BTC against 20% for gold and 10% for the S&P. Look at BTC's six biggest days since 18 Aug: 21 Aug +7.0% (gold +2.4%), 19 Aug +6.8% (gold +2.8%, DXY -0.8%), 21 Sept +6.5% (gold -0.9%, Nasdaq-100 +2.8%), 18 Sept +5.7% (gold +0.6%), 20 Aug +5.3% (gold +0.6%), 3 Sept +5.0% (gold +2.8%). Five of the six were gold up-days; the one exception, 21 Sept, was the Nasdaq's best session of the stretch and the $999M ETF inflow day. An asset that rises roughly three times as much as gold on shared up-days compounds a large performance gap while remaining tightly correlated. The price-level correlation over the last 25 sessions is -0.17 for gold and +0.05 for the S&P, and that is precisely what a price chart shows: trend divergence. Trend divergence at high return-correlation is a beta and drift story, not decoupling.

What did change, and when. The genuine shift is a summer de-rating of the equity link, not a September event. The 90-day BTC/Nasdaq-100 correlation is +0.30, the 9th percentile of two years, down from +0.40 on 1 July; 90-day BTC/S&P is +0.34 (12th percentile, from +0.46 on 1 July). The Kobeissi Letter flagged the same numbers on 5 Sept (90-day BTC/gold +0.50, near the 2020 pandemic high; BTC/Nasdaq-100 ~0.30, a one-year low) and tied the acceleration to the Treasury's 19 Aug announcement. That announcement is real and primary: Treasury press release sb0607 raises the maximum size of long-end liquidity-support buybacks (10-20y and 20-30y sectors) from $2B to at least $4B per operation, effective 9 Sept through the 4 Nov refunding. On 19 Aug BTC rose 6.85%, gold 2.78%, and DXY fell 0.83%; that is a liquidity-and-debasement day on which the two "hedges" moved together. The early-September "digital gold" coverage landed after the gold correlation had already risen, and the 30-day reading kept climbing to its 17 Sept peak and is still +0.65. CryptoPotato's "reversal in a matter of weeks" does not exist in return data.

The two stress tests confirm co-movement, not independence. The FOMC raised the target range 25bp to 3.75-4.00% on 16 Sept (12-0 vote, "inflation remains elevated"): BTC +0.7%, gold +1.25%, S&P -0.45%. CLARITY cloture failed 49-50 on 15 Sept: BTC -3.3%, gold -0.4%, S&P -0.45%, Nasdaq-100 -0.65%. On both days gold and BTC shared a sign; BTC simply moved more. Context on levels: gold's own peak was $5,318 (GC=F close, 29 Jan 2026); at $4,355 on 23 Sept it is 18% below that and flat since 1 Sept ($4,396). BTC at roughly $85,500 is 31% below its 6 Oct 2025 high of $124,753 and 46% above the 30 June 2026 low of $58,559. Both are drawdown-recovery assets responding to the same rates and liquidity impulses, which is exactly what a high correlation looks like.

Numbers in circulation to distrust. 247wallst (15 Sept) states BTC's "QQQ correlation flipped from +0.80 to -0.30". No 30-day window of daily returns in 2026 shows BTC/Nasdaq-100 below +0.08 (13 May) or above +0.70 (9 Mar); no 90-day window is below +0.24. Santiment's +36.0% market-cap gain versus my close-to-close +33.2% ($64,681 on 18 Aug to $86,172 on 22 Sept) is intraday timing (BTC printed above $87,000), not an error. The parent card's "improved diversification value" is backwards for gold: on the current data BTC's marginal diversification against gold is the lowest in two years. Against equities it is fair: a 90-day reading of +0.34 is the low end of the 2-year range (+0.21 to +0.59).

Base rate and what it means for positioning. Rolling correlations mean-revert: the 2-year median for 30-day BTC/gold is +0.11, so the +0.65 reading is more likely to fall than rise. The tradeable implication is the reverse of the headline. If BTC is at peak correlation with gold with a 1.45 beta, and gold is being repriced for "tighter for longer" after a 12-0 hike, then a gold drawdown is a BTC risk rather than something BTC is immune to. The bullish reading survives only in the narrower form: BTC's link to equities is at a 2-year low while its link to the debasement trade is at a 2-year high, which is a change in what BTC is correlated with, not a break from correlation. Method note: all correlations here are reproducible from the Yahoo chart endpoint cited below by swapping the ticker; the same endpoint supplied GC=F, ^GSPC, ^NDX and DX-Y.NYB.

Related Feeds

Brent Below $100 Did Not Move Bitcoin: 30-Day BTC-Brent Return Correlation Is -0.20 (Gold +0.65, DXY -0.42), BTC Rose on Only 27 of the Last 50 Oil-Down Days, and at 42.5% Realized Vol Bitcoin Is Now Calmer Than Crude (46.2%) While DVOL 37.7 Sits at Its 21st Percentile

CoinDesk's 23 September 2026 market wrap (Knight/Godbole) paired two facts: bitcoin consolidating at $86,379 in the European morning (+0.24% since midnight UTC, +1.3% over 24 hours, daily volume down 36% to $38 billion) and Brent crude slipping below $100 for the first time since 9 September, to $99.13 after touching $108 in mid-September. The thin parent card turns that pairing into a causal story ("increased sensitivity to macroeconomic signals", oil "did not provide a clear catalyst"). Measured against price data, the oil link is close to zero; the parent card is accurate on its numbers but the framing implies a relationship the data does not show. The correlation, measured. Using Yahoo Finance daily closes for BTC-USD and the front-month Brent future (BZ=F), daily log-return correlation to 23 September is -0.20 over 30 trading days, -0.22 over 60, -0.17 over 90 and -0.06 over 250. An r of -0.2 explains about 4% of bitcoin's daily variance; sign-wise it is mildly inverse, not the positive "risk-on with oil" or the "oil down, inflation down, BTC up" story implied by the headline. Rolling 30-day BTC-Brent correlation has wandered from -0.38 (to 30 June) to -0.14 (31 July) to +0.02 (31 August) to -0.20 (23 September): noise around zero. A cruder test gives the same answer: of the 50 sessions in the past year on which Brent fell more than 2%, bitcoin closed higher on 27, a 54% hit rate, i.e. a coin flip. What bitcoin is actually tracking is gold and the dollar. Over the same 30 days the BTC-gold (GC=F) return correlation is +0.65 (60d +0.55, 90d +0.51, 250d +0.23) and BTC-DXY is -0.42 (60d -0.34, 90d -0.36). Gold at r=0.65 explains roughly ten times as much daily variance as Brent. That is consistent with today's tape as CoinDesk describes it: gold down 0.85% to $4,321, silver down 2.2% to $65.53, the dollar index up 0.21% to 100.76 (Yahoo DX-Y.NYB shows 100.90 intraday on 23 September) and bitcoin softening under $86,000 with futures open interest flat near 710,000 BTC. Bitcoin is behaving as a dollar-denominated monetary asset on the day, not as an energy-inflation hedge. Notably the S&P 500 correlation (+0.42 over 30d, +0.49 over 250d) is the only relationship that has been stable across horizons, and even that is weaker than gold's over the past three months. The oil move in context. Brent's September path on Yahoo's front-month closes: $97.92 on 8 September, $101.21 on 9 September (first close above $100), $108.75 on 15 September (the September high), $105.83 on 16 September (Fed day), $103.87 on 18 September, $100.34 on 21 September, $99.25 on 22 September and $96.29 intraday on 23 September. The 12-month range is $58.92 (16 December 2025) to $118.35 (31 March 2026), and Brent is still up 42% on its 23 September 2025 close of $67.63. The driver is the US-Israel war on Iran and the Strait of Hormuz: Al Jazeera reported on 23 September that the US and Iran held mediated talks at the UN General Assembly on ending the war and reopening the strait, with US envoy Steve Witkoff saying mediators had "shuttled between" the two delegations throughout Tuesday. One nuance to CoinDesk's "first time since 9 September": on the continuous front-month contract the first sub-$100 close was 22 September ($99.25), not 23 September, so the break is a day older than the headline suggests. That is a contract-basis detail, not an error worth flagging. Bitcoin's own move dwarfs anything oil did. Yahoo BTC-USD closes: $75,613 on 15 September, $76,150 on 16 September (the Fed hike day), $80,901 on 18 September, $86,603 on 21 September, $86,172 on 22 September; Coinbase spot printed $85,472 at 12:25 UTC on 23 September. That is +14.5% in four sessions, and the 21 September leg coincided with the $999 million US spot ETF inflow day, not with any single oil print. The 12-month range is $58,559 (30 June 2026) to $124,753 (6 October 2025), so $86,000 sits 31% below the high and 47% above the low. Volatility: bitcoin is now the calmer asset. From the same daily series, bitcoin's annualised realised volatility is 59.0% over 7 days, 42.5% over 30 days and 38.8% over 90 days. Brent's 30-day realised vol is 46.2%, above bitcoin's. Deribit's DVOL daily close for 23 September (00:00 UTC) is 37.68, the 21st percentile of the last 371 daily closes (range 33.59 to 82.62); CoinDesk quotes "near 38%, around the 23rd percentile", which matches. So implied vol at 37.7 sits below both the 30-day (42.5%) and the 7-day (59.0%) realised, which is the precise sense in which Deribit calls implied "cheap relative to historical spot momentum": options are pricing that the four-session 14.5% burst is over and a range resumes. Options positioning confirms the range read. Deribit's public book summary at ~12:25 UTC on 23 September shows 498,903 BTC of total BTC option open interest, 183,039 BTC of it (about $15.6 billion at $85,500) in the 25 September expiry. Largest 25 September strikes by OI: 70,000 call 8,792 BTC, 70,000 put 8,133, 90,000 call 7,286, 100,000 call 6,950, 85,000 call 6,734, 80,000 call 6,230, 60,000 put 5,571, 95,000 call 4,949. Across all expiries call OI is 30,280 BTC at 90,000, 31,624 at 95,000 and 27,693 at 100,000, against put OI of 13,360 at 70,000, 9,410 at 75,000 and 8,405 at 80,000. That is exactly CoinDesk's description (calls building at $90K-$100K, a put floor at $75K and below); the point is that this structure was built before oil broke $100 and is unchanged by it. Where oil does matter is second-order, through the Fed. The FOMC statement of 16 September raised the target range 25 basis points to 3.75%-4.00% on a 12-0 vote, said "inflation remains elevated" and noted "uncertainty remains elevated owing, in part, to geopolitical developments". Brent at $108 was part of that inflation picture; Brent at $96 removes some of the case for a further hike, which would weigh on the dollar, and the dollar is the variable bitcoin actually responds to (r = -0.42). But that chain runs through DXY and gold, and today DXY is up and gold is down, so the same channel that could help bitcoin is currently pointing the other way. The honest read on 23 September is that bitcoin's consolidation is post-breakout digestion after a 14.5% four-session move, with the 25 September expiry and the gold/dollar tape as the live variables, and Brent's slip under $100 is a coincident headline, not a catalyst.