Research · Dataset
Gold versus Bitcoin, 2014–2026
Every gold desk gets asked why hold gold when Bitcoin has gone up more. This page answers with numbers instead of adjectives: over the 143 months where both are observed daily, what actually happened to each, how they moved together, and — the part most pages skip — whether any of it survives a statistical check.
Data through 2026-11-10 · gold fixes from 1968-04-01, Bitcoin from 2014-12-01 · updated 2026-11-13
Source and method
FRED notes that reproduction of Coinbase data is prohibited without prior written permission. This page therefore publishes derived statistics only and no raw Bitcoin or merged price series.
Portfolios are monthly-rebalanced, cost-free and illustrative. No financing, custody, spread or slippage is modelled — and in truth Bitcoin would trade with a much wider spread than the $0.30 per ounce used elsewhere on this site. These numbers are a direction check, not a trading plan. No transaction cost, spread, financing, custody or tax is modelled.
The headline numbers
| Measure | Gold | Bitcoin |
|---|---|---|
| CAGR (monthly averages) | 10.68% | 56.55% |
| CAGR, inflation-adjusted | 8.13% | 53.19% |
| Annualised volatility (monthly) | 12.45% | 66.88% |
| Annualised volatility (daily) | 16.05% | 70.02% |
| Sharpe (4% risk-free) | 0.564 | 0.915 |
| Maximum drawdown | -19.57% | -77.09% |
| Total return over the sample | +235.0% | +20,770.9% |
Bitcoin compounded at 56.5% a year against gold's 10.7% on this sample. It also carried roughly 5.4× the volatility and a -77.1% drawdown against gold's -19.6%.
Inflation adjustment uses CPI-U through 2026-09. The source CPI file is missing October 2025, so the real-return series steps over that month rather than dropping it, and the annualisation uses the true elapsed span.
Do they move together?
The most common claim on both sides is that gold and Bitcoin are substitutes — that investors sell one to buy the other. The data does not support it in either direction.
A monthly correlation of -0.0606 is indistinguishable from no relationship at all: the p-value is 0.472, and the Fisher interval runs from -0.223 to 0.105. It straddles zero.
The block bootstrap agrees, and it is the more trustworthy of the two because monthly returns are autocorrelated and a naive interval pretends otherwise: 95% CI -0.211 to 0.130 from 2,000 resamples in 12-month blocks.
A robustness check: recomputing on month-end prices instead of monthly averages gives 0.0737. The conclusion does not depend on which convention you pick.
Is the return gap even real?
The gap is large, and it is also not separable from zero.
Bitcoin's CAGR exceeds gold's by 45.87 percentage points (10.68% vs 56.55%). Adjusted for inflation the gap is 45.06 points.
But the moving-block bootstrap on that difference gives a 95% interval of -155.18 to 11.33 percentage points. It includes zero, so a single path through one period cannot tell you whether the gap would have repeated.
This is not a way of saying the gap does not exist — the realised numbers are the realised numbers. It is a way of saying that one sample cannot be extrapolated into a forward expectation, which is exactly what a forecast on this topic would need to do.
Does Bitcoin cushion gold's bad months?
If the two are substitutes, Bitcoin should rise when gold falls. Here is what happened across gold's 12 worst months.
Bitcoin rose in 67% of gold's worst months, and in 58% of gold's best months. It is not a reliable offset in either direction — it went up alongside gold as often as it did against it.
| Gold drawdown episode | Depth | Months | Bitcoin over the same window |
|---|---|---|---|
| 2026-02 → 2026-11 | -19.57% | 9 | +11.40% |
| 2015-01 → 2015-12 | -15.89% | 11 | +83.64% |
| 2020-08 → 2022-10 | -15.45% | 26 | +69.04% |
| 2016-08 → 2016-12 | -14.14% | 4 | +38.62% |
Gold drawdown episodes of 5% or deeper, measured on the monthly-average index. Bitcoin's return over each window is the cumulative return over the same months.
Blends
Monthly-rebalanced, cost-free mixes of the two. These are illustrative arithmetic, not recommendations.
| Portfolio | CAGR | Volatility | Sharpe | Max drawdown | Total return |
|---|---|---|---|---|---|
| Gold only | 10.68% | 12.45% | 0.564 | -19.57% | +235.0% |
| 80% gold / 20% Bitcoin | 22.56% | 16.18% | 1.103 | -24.44% | +1,029.0% |
| 50% gold / 50% Bitcoin | 38.27% | 33.64% | 1.014 | -49.69% | +4,652.1% |
| Bitcoin only | 56.55% | 66.88% | 0.915 | -77.09% | +20,770.9% |
Note what the drawdown column does. Going from 100% gold to 50/50 took the worst drawdown from -19.6% to -49.7%. A 50/50 mix was worse in the worst moment than gold alone, because the two assets draw down at the same time often enough to matter.
The Sharpe ratios here (1.103 for 80/20, 1.014 for 50/50) are computed on a single historical path with a 4% risk-free rate. Do not read them as forward expectations; the same statistical caveat as above applies.
Is the relationship stable?
A correlation is an average. If the relationship changes sign from year to year, the average is not describing anything you can trade.
By year the correlation was positive in 5 years and negative in 7. It reached -0.763 in 2015 — when gold fell 11.4% and Bitcoin still rose — and 0.092 in 2019, when both rose together.
| Year | Correlation | Gold CAGR | Bitcoin CAGR |
|---|---|---|---|
| 2015 | -0.763 | -11.4% | +15.1% |
| 2016 | -0.221 | +7.8% | +90.7% |
| 2017 | -0.288 | +9.5% | +1,853.3% |
| 2018 | -0.550 | -1.1% | -77.1% |
| 2019 | 0.092 | +18.3% | +100.7% |
| 2020 | 0.051 | +25.7% | +192.3% |
| 2021 | -0.220 | -3.7% | +134.8% |
| 2022 | 0.476 | +0.6% | -65.6% |
| 2023 | 0.622 | +12.9% | +150.6% |
| 2024 | -0.117 | +30.3% | +131.6% |
| 2025 | -0.136 | +62.2% | -9.8% |
| 2026 | 0.026 | -6.4% | -15.5% |
2026 covers 11 complete months, so its figures are partial-year and annualised from a shorter window.
Month-of-year buckets
The seasonality question, applied to both assets at once. Each bucket averages 12 observations.
| Month | Gold mean | Gold up | Bitcoin mean | Bitcoin up |
|---|---|---|---|---|
| Jan | +3.98% | 100% | +3.85% | 67% |
| Feb | +1.90% | 58% | +3.05% | 67% |
| Mar | +0.59% | 42% | +5.02% | 75% |
| Apr | +2.40% | 67% | +3.78% | 67% |
| May | -0.04% | 50% | +9.28% | 67% |
| Jun | -0.37% | 42% | +2.19% | 67% |
| Jul | -0.19% | 42% | +3.26% | 67% |
| Aug | +1.98% | 67% | +6.37% | 50% |
| Sep | +0.51% | 42% | -1.26% | 42% |
| Oct | +0.34% | 33% | +7.90% | 67% |
| Nov | -0.35% | 50% | +11.88% | 58% |
| Dec | +0.13% | 45% | +10.19% | 55% |
Twelve monthly buckets, six sub-periods and a rolling correlation window are all inspected. At a 5% level, roughly one in twenty buckets looks significant by chance alone; no single bucket should be read as a signal.
What this dataset does and does not support
Supported by the data: Bitcoin's realised return over this overlap was far higher than gold's, and its realised risk was several times larger — 66.9% annualised against 12.4%, with a -77.1% peak-to-trough drawdown. Gold and Bitcoin did not behave as substitutes: the return correlation was -0.061 with an interval that includes zero, and Bitcoin was positive in 67% of gold's worst months.
Not supported: any claim that one of the two should be held in preference to the other on the strength of these numbers. The return gap's bootstrap interval (-155.2 to 11.3 points) includes zero, and a single sample cannot forecast the next one. Any forward-looking statement built on this page would be an extrapolation, and the page does not make one.
The defensible use of these numbers is sizing and risk framing, not selection. If you size by volatility, the position size calculator on this site does it directly: Bitcoin would need roughly 5.4× less notional than gold for the same risk contribution at these measured volatilities.
Frequently asked questions
Over what period is gold compared with Bitcoin here?
Every headline figure uses the 2976 dates on which both series published an observation: 2014-12-01 to 2026-11-10, which is 143 complete months of monthly-average prices and 2974 daily returns. The gold leg is the LBMA Gold Price PM benchmark fix in US dollars per troy ounce; the Bitcoin leg is the Coinbase close in US dollars as carried by FRED series CBBTCUSD.
What is the correlation between gold and Bitcoin returns?
Monthly returns correlate -0.0606 over 143 months, with a p-value of 0.472 and a Fisher 95% interval from -0.2225 to 0.1046. A moving-block bootstrap (12-month blocks, 2,000 resamples) gives -0.2108 to 0.1304. On daily returns the figure is 0.0908. In plain terms: gold and Bitcoin do not move together in any reliable way, in either direction.
Does Bitcoin beat gold, and by how much?
On this sample, yes, and by a wide margin. Bitcoin compounded at 56.55% a year against gold's 10.68%, a gap of -45.87 percentage points. Adjusted for inflation the figures are 53.19% and 8.13%. The honest caveat is that the gap is not statistically separable from zero once autocorrelation is allowed for: the block-bootstrap interval on the difference runs from -155.18 to 11.33 points, which includes zero. A single path cannot tell you what the next path does.
Is the risk different as well as the return?
Yes, and much more so. Annualised monthly volatility is 12.45% for gold against 66.88% for Bitcoin, a ratio of 5.37 to one. The worst drawdown on the sample was -77.09% for Bitcoin (2018-12) against -19.57% for gold (2026-11), and Bitcoin took until 2020-11 to get back to its previous peak after the 2017-18 collapse. Volatility on daily returns is similarly lopsided: 16% for gold against 70% for Bitcoin.
Do gold and Bitcoin act as a hedge for each other?
Not on this evidence. Over gold's 12 worst months, Bitcoin averaged 3.14% with a median of 4%, and was positive in 66.7% of them. Bitcoin is not a dependable offset when gold falls. The blend also shows why the question matters: a 50/50 monthly-rebalanced mix produced a -49.69% drawdown against gold's -19.57%, so adding Bitcoin widened the drawdown rather than cushioning it.
Is the relationship stable over time?
No. The 12-month rolling correlation ranges from -0.7628 to 0.801 with a median of -0.0733, and it was negative in 55.3% of the 132 windows. Year by year the sign flips repeatedly: 5 of the 12 testable years show a positive correlation and 7 a negative one. A relationship that changes sign by year is not a relationship you can size against.
Can this data tell me which one to hold?
No, and it is worth being explicit about why. The dataset measures what happened, not what will happen, and one path through one period cannot separate skill from luck: the block-bootstrap interval on the return gap includes zero. The defensible use of the numbers is sizing and risk framing, not selection. Gold's realised volatility here was 12.45% annualised; if you size positions by volatility, as the position size calculator on this site does, Bitcoin at 66.88% needs roughly 5.4 times less notional for the same risk contribution.
Sources
-
Gold: LBMA Gold Price PM benchmark fix, USD per troy ounce —
scripts/daily_gold_lbma_pm.csv(14723 business-day fixes, 1968-04-01 to 2026-11-10). -
Bitcoin: FRED series CBBTCUSD (Coinbase Bitcoin close, USD) —
scripts/fred_CBBTCUSD.csv(2014-12-01 to 2026-11-12). - Inflation: FRED series CPIAUCSL, through 2026-09.
-
Analysis script:
scripts/gold_vs_bitcoin_analysis.py. Independent recomputation with numpy:scripts/verify_gold_vs_bitcoin.py(126 checks over the published figures). - For the correlation caveats used here, see Gold vs. real yields and the dollar and the gold/silver ratio study.
Returns are simple percentage returns on the stated basis. No transaction cost, spread, financing, custody or tax is modelled. Nothing on this page is investment advice, a recommendation or a forecast.