Bitcoin has spent 43% of its life above its long-run power law trend, and a few months at a time far above it. What has each spike meant for HODLers?
A spike here means price running far above its power law trend, measured as a multiple of trend, not as a price. The two can differ. Because the trend keeps rising, a later price can be higher and still sit lower against trend. In 2021 the biggest move above trend came in February at 3.2×; the highest price came in November at 2.6×. The floor works the same way: a level that rises with the trend, not a fixed price.
Few have called a bitcoin top in advance, and fewer still have done it twice. No published forecast I found identified the 2021 or 2025 tops. In hindsight there was a pattern: a falling ceiling drawn through the 2011, 2013 and 2017 spikes pointed to about 3.5× trend for 2021, close to the 3.2× reached. The same line, extended, pointed to about 2.0× for this cycle; the actual spike was 1.19×.
What the record does show is how spikes have behaved, how long they lasted, and what HODLers who acted on them ended up with. That is enough to decide what you would do, before you need to.
below the price high by the time price fell back to trend.
“Just HODL” was the right default when nobody knew how far bitcoin could run or where its price should sit. The power law changed that. In public use since 2018–19, it gives HODLers a reference point that earlier HODLers lacked: how far price is above its trend, and how every past spike that far up ended. Today’s informed HODLer is guessing less than they were in 2017.
The record favors small, planned actions sized so that being wrong is bearable. It does not favor selling the stack.
What would you want a spike to do for you?
Selling at a spike means fewer coins for the same dollars, and there’s no rebuy to get wrong.
Next: Spend and Replace, a separate exploration that models spending bitcoin regularly and buying it back over time, for spending that isn’t a single bill.
Open Spend and Replace →It has worked. It has also cost HODLers most of their stack. The difference was where they sold, where they bought back, and which account they did it in.
Judged on the spike in front of you. To commit to the same trade for every spike, .
Next: How Much Cash, a separate exploration where you set your own sell and rebuy points and see every past entry near them, in bitcoin and after tax. Your trigger, share and account carry over.
Open How Much Cash →A standing rule, such as selling a slice above one multiple of trend and buying back below another, takes the decision out of the moment.
Judged across every cycle, including the ones it gets wrong. Deciding on a single spike instead? .
Next: Disciplined Rebalancing, a separate exploration where you write your own sell and buy-back rule and see it run through every cycle and forward. Your account type carries over.
Open Disciplined Rebalancing →In a taxable account, selling at a spike and buying back lower only grows your stack if price falls by more than the tax you paid. With a low cost basis, that’s roughly your tax rate: 24% at the top federal long-term rate, more with state tax, 41% short-term. In an IRA or other tax-advantaged account there is no hurdle. Any fall in price, after small trading costs, buys back more than you sold.
Two costs the hurdle leaves out. A HODLer who never sells can defer tax indefinitely, and US heirs may receive a stepped-up cost basis; selling turns that deferred tax into tax paid now. And bitcoin ETFs in an IRA trade only on weekdays, so a spike or crash over a weekend can’t be acted on until Monday.
Price must fall this far before you buy back as many coins as you sold.
A loan taken when price is far above trend is sized on collateral at its peak value. The decline that has followed every spike, 53–86% by the time price returned to trend, is what triggers margin calls and liquidations. If you’re going to borrow against bitcoin, the record favors doing it when price is near the floor, not near a high.
Every cycle since 2011 has carried bitcoin above its trend, and every time it came back to trend, and in four of five cycles fell on to within about 10% of the floor. Here is each one, measured on the same model the rest of this site uses.
Trend, floor and upper band are the site’s Power Law channel; see the Power Law page for the model.
Through 2017, bitcoin’s biggest move above trend and its highest price arrived together. In the last two cycles they didn’t.
Filled dot: the spike, the largest multiple of trend. Hollow dot: the price high, the highest price. Labels give the multiple of trend at each.
Price and the Power Law channel, 2020 to today (log scale). The price high of November 2021 was above the February spike in dollars and lower against trend, because the trend kept rising. The same happened in 2024–25.
Days above a multiple count the unbroken run around each spike. Prices are the site’s ~12-day closes, so price highs read slightly below intraday highs.
The tops everyone recognized came at the end of parabolic run-ups. The cycle indicators that flagged them, Pi Cycle among them, are built to detect that acceleration. The two final price highs nobody flagged came after slow climbs. Pi Cycle never fired in 2025, and seven of eleven classic top indicators missed (Galaxy Research, June 2026). Their trigger lines are fixed; the waves they measure are getting smaller.
Price rise in the 90 days before each top (log scale). Orange: parabolic run-ups, over 200% in 90 days. Grey: slow climbs, including the two final price highs nobody flagged.
Buyers at each cycle’s price high waited 1.7, 3.3, 2.9 and 2.3 years to get back to even. Buyers at the March 2000 Nasdaq peak waited 15.1 years, or 17.9 after inflation. Buyers at the December 1989 Nikkei peak waited 34.1 years. Gold bought in January 1980 took 26.3 years to recover in dollars and 45.1 years after inflation, until February 2025. Bitcoin’s difference is that the asset is the network. The railroads and fiber built in past manias lifted the economy, not the shares bought at the top.
Years from the price high to the first close at or above it. Bitcoin from the site’s price series; Nasdaq and Nikkei from FRED daily closes; gold from World Bank monthly averages, which understate the $850 intraday high of January 1980. “Real” uses US CPI.
In 2013 bitcoin’s trend was rising about 220% a year; today it rises about 37%. That changes the arithmetic of selling into a spike.
Annual growth of the trend price, by year (log scale). Past the dashed line is the model’s own arithmetic carried forward, not data.
Bitcoin’s trend is a power law: price rises with time since 2009 raised to a fixed power. Other growth laws make good comparisons. Two of the ones below are power laws too; Moore’s law is exponential, a fixed doubling time. Some held for a century. Some broke, and the way they broke is instructive.
Solid: the period the law described well. Faded: after it bent or broke. Population’s law fits records back millennia; shown here from 1900. The population curve was fitted to records up to 1960 and pointed to infinite population on 13 November 2026. Growth slowed in the 1960s instead.
Why the population law broke: birth rates fell as incomes rose and people moved to cities, so growth stopped accelerating. What each law said, and how it broke, is in the case files below.
Across the cases I examined, spikes above a long-running law came back to it. Earthquake aftershocks fade on a schedule. Manias in credit, commodities and paper assets returned to where they started. The exceptions, where a spike marked a new and higher path, involved a change in the mechanism underneath: a new technology, or a currency losing value.
The best predictor of whether a law holds is whether something drives it. Wright’s law, which ties falling costs to how much has been produced, has held for 90 years. Moore’s law, tied to the calendar, bent when transistors hit physical limits. A 1960 paper in Science fitted world population to a curve that reached infinity on Friday, 13 November 2026. Within a few years of publication, birth rates fell and the law broke.
Bitcoin’s law is fitted to time, the weaker kind. Its proposed mechanism is a fixed supply meeting a growing network. The trend is credible to the extent that mechanism holds.
What I looked at, graded by how well the record supports it. Open a card for the detail.
Many past manias ended when high prices brought new supply: mines reopened, railroad track was overbuilt, new shares flooded the market. Bitcoin’s issuance follows a fixed schedule written into the protocol. The difficulty adjustment keeps blocks, and so new coins, arriving on that schedule whatever the price. Since the 2024 halving that is about 450 coins a day, against roughly 20 million in circulation. Issuance halves about every four years and ends around 2140. The supply that answers a bitcoin spike is existing HODLers selling. Which means the readers of this page are part of how spikes end.
One change could tighten supply further. As lending against bitcoin matures, more HODLers may borrow against their coins rather than sell them. Fewer sellers at a spike would leave even less supply to meet demand. That is a possibility, not yet a measured effect, and it does not change the warning on the first tab: borrowing near a spike is the risky moment.
A breakaway here means a lasting move to a steeper trend, not a spike that comes back.
Someone says it at every peak. Sometimes they’re right about the technology and wrong about the price and the timing. Here is how to tell the difference, and what I’d need to see.
Old metrics don’t apply. Prices only go up. The cycle is dead.
Wrong at essentially every peak on record, from 1720 to 2021.
Railroads will reshape commerce. The internet will dominate business.
Often right, but early. Investors misjudged the timing, not the societal impact.
You can believe bitcoin will ultimately matter enormously and still judge that a spike has priced in too much, too soon.
Not price rising further above trend. The trend itself steepening, because the things that drive it, such as HODLers, hashrate, wallet addresses holding a balance, and network use, are growing faster than before. And it would show against everything, not just the dollar.
In Weimar Germany, Venezuela and Zimbabwe, local stock markets soared while losing value in real terms. A rise that shows only in dollars may be the dollar falling. So I priced bitcoin in inflation-adjusted dollars and in ounces of gold and fitted the trend again. It holds in both. The exponent falls only from 5.5 in dollars to 5.3 after inflation and 5.2 in gold, and the fit stays nearly as tight. Bitcoin’s trend is bitcoin’s, not the dollar’s.
Each line is bitcoin’s price divided by its own trend in that unit (log scale), 2011 onward. These are a 2011-onward cross-check fitted by ordinary least squares, computed 2026-10-04; they differ from the site’s canonical model, which is fitted from genesis with an exponent of 5.77.
One surprise. Priced in gold, bitcoin sits at 0.28× its trend, against 0.60× in dollars on the same fit, because gold has risen so far since 2024. In gold terms the October 2025 price high was not a spike at all: it sat below trend.
Gold’s price history is broken by policy. Americans were barred from owning it from 1933 to 1974, and the dollar was fixed to it at $35 an ounce until 1971, when the link was cut. Bitcoin’s largest test so far, China’s 2021 ban on bitcoin mining and trading, cut bitcoin’s price by about half and pushed hashrate out of the country. Both recovered within the year and the trend held. Self-custody makes bitcoin hard to seize; bitcoin held through ETFs and exchanges is as reachable as gold in a vault.
The power law’s strongest critics argue the fit is weaker than it looks. Regressing a trending series on time inflates the fit, the exponent shifts with the start date, and a stack of three adoption S-curves fits the history better. The same researchers found that the simple power law forecasts 12 to 24 months ahead better than any alternative they tested (Baquero & Menezes, 2026). Weak structure, strong forecasts.
My reading: the critics and the model agree on what matters for this page. The power law is not a law of nature, and it has still been the best guide to where price sits a year or two out. That is all this page asks of it: a yardstick for how far above trend price has run, and what followed.
Every bitcoin figure is computed in your browser from the site’s shared price series and the Power Law model: trend 1.6×10−17 · days5.77 since genesis, floor 0.42×, upper band 3×. Prices are ~12-day closes, measured on 2011 onward as on The Bitcoin Floor. The live price, when it loads, moves only the readout at the top of the first tab and the “never bought back” figures.
Not in time to act on it. No published forecast I found identified the 2021 or 2025 tops in advance, and by the time price fell back to trend it was 53–86% below the price high.
It depends what for. Funding a planned expense at a spike has historically used fewer coins; trimming to buy back has worked in some cycles and cost HODLers coins in others, especially in taxable accounts.
No tax is due on the sale, so any fall in price lets you buy back more than you sold. In a taxable account the price must fall by roughly your tax rate first.
Yes. The largest spike above trend fell from 12× in 2013 to 1.19× in 2024, and the October 2025 price high sat close to trend.
It is the riskiest time to borrow: the loan is sized on peak value, and the declines that followed every spike are what trigger liquidations.
Get the essays and tool updates
A few emails a month when something new ships — an essay, a new tool, a meaningful update. No funnel, nothing for sale, unsubscribe anytime.
Every page on this site has been improved by someone pushing on it. Ask a question, flag an error, or suggest what’s missing — it goes straight to the author, never published.
I’m also available for one-on-one conversations and open to collaborations — Work with me →