Crypto Market Recap Sep 28: Bitcoin Slips to $83,100 USD, Gold Falls 3.85%
Bitcoin (BTC) was changing hands at $83,149.48 USD at 6:05 pm ET Monday, down 1.55% from Sunday’s close. Gold had a far worse day: the December futures contract fell 3.85% to $4,154.90 USD, and the gold miners fell further still, with the TSX-listed iShares S&P/TSX Global Gold Index ETF (XGD.TO) down 5.14%. Behind both sits the same repricing, documented first hand: a US Treasury curve that closed at its highest level of 2026 at the 2-year, the 5-year, the 10-year and the 30-year. Crypto trades around the clock, so the coin prices here are a snapshot taken at 6:05 pm ET, not a close, and should be read that way.
Prices at the snapshot
| Last | Prior close | Change | |
|---|---|---|---|
| Bitcoin (BTC) USD | $83,149.48 USD | $84,458.09 USD | -1.55% |
| Bitcoin (BTC) CAD | $117,827.09 CAD | $119,549.74 CAD | -1.44% |
| Ethereum (ETH) USD | $2,670.96 USD | $2,686.93 USD | -0.59% |
| Ethereum (ETH) CAD | $3,785.28 CAD | $3,803.32 CAD | -0.47% |
Ethereum (ETH) fell about two fifths as far as Bitcoin, so it picked up close to a point on it. The Canadian dollar slipped 0.19% to 70.56 US cents, and that is why the CAD losses above are slightly smaller than the USD ones: a weaker loonie absorbs a little of the fall for a Canadian holder.
On hourly bars, a different basis from the snapshot above, Bitcoin ranged from a low of $82,624 USD at 5:00 am ET to a high of $84,017 USD at 2:00 pm ET before giving that recovery back into the evening.
One number needs stating before gold and Bitcoin get compared anywhere below. Bitcoin’s -1.55% is a move from Sunday’s close, because Bitcoin traded through the weekend. Gold’s -3.85% is a move from Friday’s close, because gold did not. Measured over gold’s own window, Friday’s close to now, Bitcoin is down 1.05%. That is the figure used everywhere the two assets are set side by side, because it covers the same clock time.
What gold and the wider tape did
The selling was broad across the sector. Alongside that 5.14% drop in the gold miners, the iShares Materials ETF (XMA.TO) fell 3.70%. Equity indexes were softer too: the S&P/TSX Composite closed at 35,489.86, down 0.87%, the S&P 500 at 7,683.69, down 0.77%, and the NASDAQ at 26,820.38, down 0.92%. Silver went with gold, the December contract at $61.11 USD, down 5.70% from Friday’s $64.80 USD. Energy was the one outlier, the November WTI contract (CLX26.NYM) up 0.60% to $92.96 USD. XGD is a global fund, so the Canadian cut is worth stating separately: the 42 gold and silver names inside the S&P/TSX Composite fell 4.74% as a block. That is a different basket from the ETF and a different number. For the individual names, Canadian gold stocks is where to see which ones carried the weight.
The cause: a Treasury curve at its 2026 high
The belly and the long end of the US Treasury curve both moved up on Monday, and each of the four readings below is the maximum of all 186 published sessions of 2026. The very front end did not move: the 1-month held at 4.04% and the 2-month at 4.20%.
| Friday Sep 25 | Monday Sep 28 | Change | |
|---|---|---|---|
| 2-year | 4.81% | 4.92% | +11 bp |
| 5-year | 4.98% | 5.06% | +8 bp |
| 10-year | 5.17% | 5.24% | +7 bp |
| 30-year | 5.49% | 5.56% | +7 bp |
For scale, the 2-year opened 2026 at 3.47% and the 10-year at 4.19%, so Monday’s levels are the top of a year-long climb rather than a one-day spike. The full curve is published on the US Treasury daily yield curve page.
The mechanism is simple. Neither gold nor Bitcoin pays interest. A government bond does. When yields rise, the income an investor gives up by holding either asset instead of a bond rises with them.
Worth keeping in proportion: a 7 basis point move on the 10-year is a small input for a 3.85% fall in gold, so the day is better read as a repricing that had been building for a week than as a reaction to Monday alone. The 10-year has added 28 basis points since September 22.
Canada’s rate market moved the same way. One-month CORRA futures settlements price 13.56 basis points of tightening into the Bank of Canada’s October 28 decision, an implied rate of 2.39% against the current 2.25%, which works out to a 54% chance of a quarter-point hike. The Bank of Canada’s key interest rate has held at 2.25% since October 30, 2025.
Our research: correlation says direction, not distance
The obvious read of Monday is that Bitcoin decoupled from gold. Gold fell nearly 4% and Bitcoin fell about 1%. Our own arithmetic on eleven years of daily data says the answer takes two numbers rather than one, because correlation and sensitivity measure different things. Correlation is scale-free: it tells you whether two assets move in the same direction, not how far each one travels.
We measure gold with the SPDR Gold Shares ETF on adjusted closes, since it gives an eleven-year daily series with no futures roll in it, on its own trading calendar. Bitcoin is measured over the same window as gold, gold session to gold session, so a Monday window spans the weekend for both assets and the two returns cover the same span of days. The endpoints are offset by a few hours, because Bitcoin’s daily close is Yahoo’s 00:00 UTC print and gold’s is the 4pm New York equity close. That gives 2,949 matched windows from January 5, 2015 to September 25, 2026, with today excluded because it is still in progress.
On direction, the two are moving together about as closely as they ever have. The rolling 90-window correlation of their returns is 0.546, against a full-sample mean of 0.099. Sixteen readings in the series have been higher, and every one of the sixteen was set between August 28 and September 23 of this year. Not a single window in the prior eleven years reached the level the pair is at now. The record, 0.595, was set on September 11.
The path there was not a straight line, and the reading has already turned down. Month by month through 2026: 0.127 in January, 0.209 in February, 0.174 in March, 0.190 in April, 0.232 in May, 0.285 in June, 0.459 in July, 0.555 in August, and 0.546 now, which is below where August ended and well below the September 11 peak. These windows overlap heavily, sharing 89 of their 90 observations with the reading next to them, so there are on the order of 25 independent windows behind that series and it should be read as one broad regime shift rather than as thousands of separate measurements. For a portfolio the implication is direct: an investor who holds gold miners and Bitcoin, or who bought Bitcoin as a gold substitute, has been getting materially less separation between the two through 2026 than the long-run numbers imply.
On distance, Bitcoin now travels much closer to gold’s distance than it used to. The regression slope of Bitcoin’s returns on gold’s over the same 90 windows is 0.881, meaning Bitcoin has moved 0.881% for each 1% move in gold. The full-sample average is 0.310. A year ago it was 0.072, which is to say effectively nothing. That slope also stalled in the spring rather than climbing cleanly: 0.177 in January, 0.324 in February, then back to 0.263 in March and 0.263 in April, 0.316 in May, 0.393 in June, 0.612 in July, 0.824 in August, 0.881 now. February’s level was not regained until June.
Two cautions on that slope, both of which the numbers behind it require. Gold explains under 30% of Bitcoin’s variance over the window, so the slope is an estimate with real width: the 95% interval runs from roughly 0.60 to 1.16. And it is not only Bitcoin that changed. A slope is the correlation multiplied by the ratio of the two assets’ volatility, and that ratio has fallen from about 3.1 on the full-sample averages to about 1.6 today. Part of what happened in 2026 is that Bitcoin came to move with gold. The other part is that the gap in their volatility narrowed, and the evidence we have points to gold’s side of it: 13 of the 28 gold drawdowns in eleven years fall in this year alone.
Applied to Monday, on gold’s own series rather than the futures, a 3.94% fall in the gold ETF implies a Bitcoin move somewhere between 2.4% and 4.6%, centred near 3.5%. Bitcoin fell 1.05%.
Our research: what Bitcoin has done on gold’s worst days
The eleven-year record gives the other half of the answer, and it is observed rather than extrapolated. Of those 2,949 matched windows, gold fell 3% or more in 28 of them. Bitcoin’s median move over the same 28 windows was -1.39%, and its mean was -1.77%, dragged down by the single worst case, a 37.17% fall on March 12, 2020, which was followed the very next session by an 11.93% rebound. Both belong to the same COVID crash week. Bitcoin fell in 22 of the 28 windows, or 79%, and fell 3% or more in 7 of them. For comparison, across all 2,949 windows Bitcoin fell in 47.3% with a median of +0.17%, so gold’s worst days are genuinely worse than average for Bitcoin without being disasters.
Monday’s matched-window -1.05% is a little milder than that -1.39% median.
The sharper comparison is with the four gold drawdowns that fall inside the very window the 0.881 slope is fitted to, because those are what the current regime has actually delivered rather than what a line through it predicts. Gold is measured on the ETF here, the series the slope is fitted to, so Monday’s figure is its 3.94% fall rather than the 3.85% in the futures contract.
| Gold session | Gold | Slope implies | Bitcoin |
|---|---|---|---|
| June 5, 2026 | -3.65% | -3.22% | -4.51% |
| June 10, 2026 | -4.15% | -3.66% | -0.32% |
| June 24, 2026 | -3.02% | -2.66% | -2.67% |
| August 28, 2026 | -3.24% | -2.85% | -3.02% |
| Monday | -3.94% | -3.47% | -1.05% |
Monday was milder than three of those four, and the median Bitcoin response across them was -2.84%. But the middle column is the reason to be careful with the slope in both directions. Three of those four events matched or exceeded what it predicted, and only June 10 came in far below. Across all 13 of 2026’s gold drawdowns the ratio of Bitcoin’s move to gold’s runs from -0.28 to 1.62, with a median of 0.63, and six of the thirteen land at or above 0.881. So the typical tail day has cost Bitcoin somewhat less than its everyday sensitivity implies, while individual tail days have been close to a coin flip. Monday sits at the mild end of a wide distribution, not inside a reliable pattern.
Two honesty notes on the sample. Twenty-eight is a small number of events. And it is concentrated in the present: 13 of the 28 fall in 2026 and 7 more in 2020, with the remainder scattered across 2016, 2021, 2024 and 2025. Gold has been far more volatile in 2026 than for most of the series, so this year’s events carry most of the weight in the median.
The equities amplified on both legs of the trade
The operating companies levered to a hard asset moved further than the asset itself on both sides, though by very different multiples.
| Close | Change | |
|---|---|---|
| Hut 8 (HUT.TO) | $131.51 CAD | -4.01% |
| Hut 8 (HUT, Nasdaq) | $92.71 USD | -4.24% |
| Riot Platforms (RIOT) | $21.62 USD | -6.00% |
| MARA Holdings (MARA) | $12.11 USD | -3.51% |
| Coinbase (COIN) | $191.79 USD | -1.70% |
| Strategy (MSTR) | $157.14 USD | -0.93% |
Gold’s -3.85% against the miners at -5.14% is 1.33 times the move. Bitcoin’s -1.05% against crypto miners down between 3.51% and 6.00% is three to six times, a much steeper multiple on a much smaller underlying move. Two names in the table did not amplify at all: Strategy at -0.93%, which is roughly Bitcoin itself, and Coinbase at -1.70%, which earns fees on exchange volume rather than mining margin and does not carry the same leverage. For the Canadian-listed miners behind those Hut 8 numbers, our Canadian crypto stocks page covers the group. The gold side of Monday broke down the same way across the TSX, and our report on Monday’s TSX metals selloff attributes the index decline name by name.
What the day argues, in both directions
Monday supports two readings and the numbers above carry both. On one side, Bitcoin lost a little over a quarter of what gold lost over the same window, and against eleven years of gold drawdowns that is an unremarkable outcome. On the other, the direction-matching between the two assets has no precedent outside the last month of this series, so the case that Bitcoin trades independently of the metal is weaker now than at any point in eleven years.
Neither reading wins outright on one session, and the causal chain is looser than a single day makes it look. A 7 basis point move in the 10-year did not itself take 3.85% out of gold. What the rising curve offers is a mechanism that has been pressuring both assets for a week, and a reason to expect them to keep responding to the same input. That, rather than Monday’s prices, is the thing worth watching, with one caveat attached: the correlation has already eased from its September 11 high, and a regime four months old is not yet a fact about the asset.
Disclaimer: The content on bestcanadianstocks.ca is for informational and entertainment purposes only and does not constitute financial advice. Past performance is not indicative of future results. Always consult a qualified financial advisor before making investment decisions. Bitcoin and Ethereum quotes are a snapshot taken at 6:05 pm ET on September 28, 2026, from Yahoo Finance, not a close, because crypto trades continuously, and the day-over-day change is measured against Sunday September 27. Where Bitcoin is compared with gold it is measured over gold’s own window instead, Friday September 25’s close to the snapshot, so both cover identical clock time. Bitcoin’s intraday high and low are Yahoo Finance hourly bars, a third basis again. Index, commodity and equity figures are September 28 closes from Yahoo Finance, with gold and crude quoted by contract symbol, GCZ26.CMX and CLX26.NYM. US Treasury yields are the Treasury’s own daily yield curve rates for 2026, 186 published sessions, fetched September 28, 2026. Bank of Canada rate expectations come from Montreal Exchange one-month CORRA futures settlement with spot CORRA from the Bank of Canada Valet API. The correlation, the regression slope and the drawdown study are our own arithmetic on SPDR Gold Shares adjusted closes against BTC-USD, over 2,949 matched windows from January 5, 2015 to September 25, 2026, measured gold-session close to gold-session close on gold’s own trading calendar; the rolling 90-window readings overlap and are not independent observations, roughly 25 independent windows sit behind the series, and the 28 gold drawdown events are a small sample concentrated in 2026.
