Decent Newsroom
Sign In

The Shallowest Winter

By Roger November 9, 2026
Three hundred ninety-eight days after its top, bitcoin is 40.1 percent lower. The two previous cycles were down 74.7 and 81.2 percent at the same age. The rhythm held; the amplitude did not.
The Shallowest Winter

Three hundred ninety-eight days after its top, bitcoin is 40.1 percent below it. At the same age the two previous cycles were down 74.7 and 81.2 percent. The rhythm held. The amplitude did not.

The tape on 9 November 2026

Bitcoin trades at $74,465. Down 1.8 percent on the day, 4.6 percent on the week, 10.0 percent on the month, 26.9 percent on the year. Market capitalisation $1.50 trillion on $26.2 billion of volume in twenty-four hours. The Fear and Greed index reads 29, which its publisher files under fear. It has been lower: 2022 logged 207 days of extreme fear and touched 6.

The high was $126,080 on 6 October 2025, an intraday print, and from that print the asset is 40.9 percent lower. From the highest daily average, one day later, it is 40.1 percent lower. Two kinds of price appear in this piece: the spot figure from CoinGecko, and daily averages, which are a price a bull and a bear can both re-check years later without arguing about the minute. The two differ by a few hundred dollars today, and that gap is the price of the argument avoided.

Gold $4,103. S&P 500 7,377. Dollar index 101.9. US ten-year yield 4.87 percent. That last number will do more work here than bitcoin’s own price.

Three cycles, one rhythm, three amplitudes

Bitcoin drawdown from each cycle top, first 398 days

The halvings are facts with block hashes behind them: height 420,000 on 9 July 2016, 630,000 on 11 May 2020, 840,000 on 20 April 2024. Count from each of those dates to the highest daily average of the following cycle and the tops arrive after 526, 547 and 535 days. Three weeks of spread across three cycles, at a target nobody sets and nobody can move.

The falls separate them. Measured at day 398:

  • 2017 cycle: -81.2 percent
  • 2021 cycle: -74.7 percent
  • 2025 cycle: -40.1 percent

The deepest point inside the first four hundred days was -83.2 percent at day 364 in the first cycle, -76.7 percent at day 378 in the second, and -53.1 percent on 1 July 2026 at day 267 in this one. All three lows came before day four hundred. In the current cycle the low is four months behind us and the price has recovered thirteen points from it.

Two measurements follow, and neither is a forecast. The time structure of the cycle repeated within three weeks. The depth did not repeat at all. This is the shallowest decline of the three, and it is shallow by roughly a factor of two.

Volatility is what a price costs to find

The same window measured in movement instead of direction. Annualised realised volatility of daily returns over the 398 days after each top: 85.5 percent for the 2017 cycle, 65.8 percent for 2021, 44.3 percent now. Over the first ninety days after the top it was 126.6, 63.4 and 41.5 percent. The last ninety days, which carried the price from $63,917 up to $86,608 and back to $74,707, ran at 41.1 percent.

No one sets this price. There is no committee, no fixing, no band. A market like that has to discover the price every day, and volatility is the bill for the discovery. Whether the bill is falling because ownership has broadened into longer horizons, or because the marginal trade has moved into perpetual futures and options where it can be expressed with less spot movement, this series cannot tell. Both readings fit 44 percent.

The top was not a mania

Weekly closes against the power-law trend

A power law fits bitcoin’s price as a straight line when price and age are both logged. Fitted through the full record it puts the trend at $142,449 today, with an R² of 0.96 and an exponent of 5.59. Price sits at $74,465. The residual gives a percentile of 16.7: 16.7 percent of days in the record were cheaper than this one. Put the other way round, 83.3 percent of all days were more expensive.

The same yardstick at the top reads differently. On 10 October 2025, four days after the high, price was 21.2 percent above trend and the percentile was 65.6. In November 2021 price ran 172 to 185 percent above trend, with the percentile at 90 and 91. The 2017 top falls outside the eight-year weekly series used here, so it stays out instead of being guessed at.

A market that tops 21 percent above its trend line does not have the same work to do as one that topped roughly 180 percent above it. That is consistent with the shallow fall, and it is also not independent evidence, because both numbers come from the same residuals and agree by construction. What they do say is that October 2025 was the least extreme peak of the two cycles that can be measured this way.

The percentile has a weakness worth naming before anyone quotes it. It depends on the start year of the fit: 21.4 if the series begins in 2011, 23.9 from 2013, 4.4 from 2015, 19.8 from 2017. Same price, almost twenty points of range. The 16.7 above comes from the full record. A reader who prefers 2015 as the start gets “nearly as cheap as it has ever been”; a reader who prefers 2013 gets “cheap, not remarkable”.

How rare is this? Of 418 weekly observations in the eight-year series, 23 sat at or below today’s 16.7. That is 5.5 percent of the sample.

The supply side, in dollars

The block subsidy is 3.125 BTC now, and roughly 144 blocks a day come out of it. That is 164,250 new coins a year, fees aside. At today’s price the flow is worth $12.2 billion.

At the 2021 peak the schedule paid 328,500 coins a year, which was $22.2 billion at that peak price. In 2017 it paid 657,000 coins a year, $12.7 billion at the top. So the halving halves the coin count every four years, and the dollar size of the flow has stayed between twelve and twenty-two billion a year across three cycles, because the price moved while the schedule did not.

That is the whole supply argument, and it is smaller than it sounds. New coins are a flow. What clears is the willingness of existing holders to absorb that flow at a price, and that willingness is set by the same balance sheets discussed below, not by the emission curve.

The unit grew while the asset fell

US M2 money stock against bitcoin price, November 2024 = 100

US M2 was $22,149.2 billion in September 2025 and $23,413.9 billion in September 2026. That is a rise of 5.7 percent in twelve months. Since December 2014 the stock has gone from $11,721.2 billion to $23,413.9 billion, a doubling in under twelve years.

Against the same money, bitcoin did this: -23.4 percent since November 2024, while M2 rose 9.3 percent. Measured from the April 2024 halving, bitcoin is up 17.0 percent and M2 is up 11.9 percent.

Both figures are nominal, and that is the point. A dollar price is an exchange rate between an asset and a unit. Most of the time the argument is about the asset. The unit has a growth rate too, and it is published every month.

Federal debt stood at $40.56 trillion on 5 November 2026, of which $32.77 trillion was held by the public. The ten-year note pays 4.87 percent, the British gilt 5.16, the German Bund 3.27, the Japanese ten-year 2.67. This is what a high time preference looks like once it is priced: the present has become expensive again.

Cantillon’s observation from 1730 was that new money does not enter everywhere at once. It enters at a point. Whoever stands at that point spends before prices adjust; everyone downstream receives the same money later, at higher prices. The transfer is not a conspiracy, it is a sequence.

Who paid

There is no register of who sold, so what follows is arithmetic and not measurement.

A holder who bought near the top with borrowed money at a positive rate sits on the full 40 percent drawdown, plus the cost of the loan. A holder who bought in 2019 sits on a multiple. Same asset, same ledger, same twenty-one million. What separates them is not intelligence and not conviction. It is the length of the balance sheet behind the position.

Böhm-Bawerk’s point about capital was that production takes time and time has to be paid for: interest is the premium on present goods over future ones. Mises turned that into a coordinating price. Bitcoin has no cash flow, so its entire value is a claim on future salability. It is the longest-duration asset in the market and the one a 4.87 percent discount rate should hit hardest. That explains why a jump in yields hurts bitcoin more than the S&P. It does not say by how much.

Here the cycle story and the money story meet, and this is the part most cycle charts leave out. The four-year rhythm is a supply schedule. The amplitude is a balance-sheet event. One is written in code, the other is set by whoever happens to be holding the coins when the discount rate moves.

Thiers, briefly

Gresham’s law describes a market with a fixed legal rate: people spend the bad money and keep the good, and the good disappears from circulation. Thiers’ law describes the other arrangement. When the rate is free, the bad money is what gets refused and the good money keeps circulating.

That law is settled in practice rather than in argument. An asset with an issuer can be frozen by the issuer, and an asset without one cannot, which is not a promise about the future but a description of where the switch is. The same property explains why the price can be 40 percent off its high while the asset itself is unchanged: the emission schedule is 164,250 coins a year, it was the same number a year ago, and the only route to changing it runs through consensus among the people running the software.

What would prove this wrong

Three conditions, each measurable, none of them a price target.

First, the drawdown at day 500. At day 500 the 2017 cycle was down 72.7 percent and the 2021 cycle 58.1 percent, both measured the same way as the numbers above. If this cycle is deeper than 58.1 percent at that age, then “shallower” was only “slower”.

Second, the fifth percentile band, which sits at $62,119 today. A full month below it would mean the band structure is being repriced, not tested.

Third, another year of M2 growth near five to six percent with bitcoin making lower highs. That would mean the debasement bid failed. Money growth is not a bid by itself; someone has to want the asset on the other side of it.

If somebody hands you a number with a dollar sign and a date attached, ask what they are selling.

What does not change

Twenty-one million coins, with the issuance schedule written in code. No issuer, no compliance desk, no freeze function, no one to email. The price of that object is set in a market that is 40 percent off its high and still roughly twenty-three times its December 2018 low of $3,232, and both of those are true on the same morning.

A price is a measurement, and measurements get revised. The supply never has.

Sources

Write a comment