PULSEHAWK
Does gold rise when Bitcoin falls?
Three times Bitcoin has lost more than three quarters of its value. Gold finished every one of those windows within 5.1% of where it started. It did not rise. It also did not follow Bitcoin down.
The three windows, with the S&P 500 alongside
| Bitcoin bear market | Bitcoin | Gold (GLD) | S&P 500 (SPY) |
|---|---|---|---|
| 4 Dec 2013 to 14 Jan 2015 | -84.87% | -1.66% | +11.76% |
| 16 Dec 2017 to 15 Dec 2018 | -83.24% | -1.78% | -2.27% |
| 8 Nov 2021 to 21 Nov 2022 | -76.67% | -5.03% | -15.85% |
| 6 Oct 2025 to 30 Jun 2026 | -53.09% | +1.10% | +11.19% |
| 6 Oct 2025 to 25 Sep 2026 | -32.65% | +7.83% | +14.52% |
The first three rows are finished and will not change. The last two are the current cycle: one measured to Bitcoin's lowest price so far, one measured to the latest day in our record.
The current one is not behaving like the other three
This is the part a three-row table hides. Since the 6 October 2025 top gold is +1.10% measured to Bitcoin's low of 30 June 2026, and +7.83% measured to 25 September 2026. In the three completed windows gold ended slightly down every time. This time it is up.
Three completed cases are three data points. They are not a rule, they do not make a fourth case, and the fourth case is currently disagreeing with them. Anyone who tells you what gold does during a Bitcoin bear market is working from the same three numbers we are.
Gold had its own fall in the same period
A number that sits still against Bitcoin is not a number that only goes up, and this window shows it plainly. Gold set the highest close in our record on 29 January 2026, at 495.90. On 23 September 2026 it closed at 392.92, which is 20.77% under that mark.
So the same asset that barely moved through three Bitcoin bear markets has, inside the fourth, first run to the top of its own record and then given back a fifth of it. Both things are in the same series and on the same page on purpose.
we count what moved, not what it means.
The steady one is not the stock market
The obvious comparison is the S&P 500, and it goes the other way. Over the same three windows SPY moved +11.76%, -2.27% and -15.85%. Those three readings span 27.61 percentage points. The three gold readings span 3.37.
That is the one genuinely surprising number on this page: measured across Bitcoin's three completed declines, gold's outcomes sat about eight times closer together than the stock market's. Whether that says something about gold or something about three arbitrary windows is not a question three cases can settle.
How we count
Bitcoin is one price per UTC day from our own record, built from the blockchain.com market price series. Gold and the S&P 500 are the GLD and SPY daily closes from Twelve Data. Our gold and stock series begin on 3 September 2010, and "highest close in our record" means exactly that: the highest close since that date, not a claim about gold's whole history.
Bitcoin trades every calendar day, GLD and SPY only on exchange days. Of the 36 window edges in this comparison, 8 fell on a Saturday. For those we took the last exchange day before, which was exactly one day earlier every time, never more. Only two calendar dates are affected: 16 December 2017 and 15 December 2018. We would rather print that than quietly shift a date, because a one-day shift on a window of roughly a year moves the result by less than the rounding on this page, and you can now check it yourself.
Every figure above is stamped into the page source from the record, with the date it belongs to. Nothing here is copied from another site's comparison table.
Questions people ask
In the three completed cases it did not. It finished at -1.66%, -1.78% and -5.03%. It did not rise, and it did not follow Bitcoin down either.
+7.83% since 6 October 2025, which is the first time it has gone the other way. That case is not finished.
+11.76%, -2.27% and -15.85% over the same three windows, a span of 27.61 points against gold's 3.37.
Yes, inside the current window. It is 20.77% under its highest close in our record, set on 29 January 2026.
No. Three finished cases are three data points, and the fourth is already disagreeing with them. This page counts, it does not forecast.
Our own record of daily closes, with every date printed next to the number it belongs to.