Sophronepsis · research · 5 August 2026
We measured the time distances between the milestones of Bitcoin's three complete cycles. They look much more alike than the prices do. We also tell you about the number that lied to us along the way, and why our own strategy comes out badly.
“Bitcoin moves in four-year cycles. After the halving it goes up, then it falls, and in the end it starts again.”
You've heard that sentence a hundred times. And it has a problem you can see without any data: the four years aren't a rule, they're a consequence. The protocol doesn't count years. It counts blocks: every 210,000, the reward is cut in half. That this lands close to four years is the result, not the cause.
So the useful question isn't whether the cycle lasts four years. It's a different one: do the distances between events repeat? Not the prices. The timings.
And that question has an advantage almost no indicator has. You'll see it now.
Imagine I tell you to buy bitcoin when it falls 70 % from its high. It sounds precise. But that 70 % didn't come from anywhere: I moved it until the result looked nice. If 70 didn't work, I tried 65. And if that didn't either, 78.
A threshold can be pushed until you like it. A distance between two dates can't.
From 9 July 2016 to 16 December 2017, 525 days went by. That number is what it is. I can't move it to make anything fit.
A threshold gets adjusted. A date gets counted.
Bitcoin has had four halvings. For the first there isn't enough price data, so what can be measured in full is three cycles. Three. It's not much, and we'll come back to that at the end.
| Cycle | Halving | Top | From halving to peak | Lowest point | From peak to low |
|---|---|---|---|---|---|
| 2012 | 28-Nov-2012 | no data | — | 14-Jan-2015 | — |
| 2016 | 09-Jul-2016 | 16-Dec-2017 | 525 days | 15-Dec-2018 | 364 days |
| 2020 | 11-May-2020 | 08-Nov-2021 | 546 days | 09-Nov-2022 | 366 days |
| 2024 | 19-Apr-2024 | 06-Oct-2025 | 535 days | open | 303 and counting |
From halving to peak: 525, 546 and 535 days. Twenty-one days between the shortest and the longest, across three cycles that look nothing alike in price.
From peak to lowest point: 364 and 366 days. Two days apart across two cycles that look nothing alike.
And from the lowest point to the next halving: 542, 513 and 527 days.
Two days' difference over four years is a beautiful number. And that's exactly why you should distrust it.
“The lowest point” is the day of the minimum price. But which day that was is only known years later. Picking it by looking back and then boasting about the precision is cheating without meaning to. So we repeated the measurement defining the bottom in three different ways:
| How we define “the lowest point” | 2017-18 cycle | 2021-22 cycle | Difference |
|---|---|---|---|
| The day of the lowest price | 364 d | 366 d | 2 days |
| The lowest week | 362 d | 377 d | 15 days |
| The lowest month | 406 d | 388 d | 18 days |
The two days don't exist. Smoothing over a week multiplies the margin by seven, and smoothing over a month by nine. What does exist, and holds up under all three definitions, is this: from peak to bottom about 370 days go by, with a margin of two and a half weeks. That's still useful information. It isn't a Swiss watch, and anyone who sells you an exact date is selling you smoke — us included, if we hadn't done this check.
Compare that with what prices did in those same three cycles. The rises shrank each time: first it multiplied by a hundred, then by thirty, then by seven and a bit. The prices don't look alike. The calendars do.
Before looking at the calendars we spent weeks measuring something else: in which month of the cycle it was best to start buying back. And we found a beautiful curve. The later you bought, the better it came out. It went as far as multiplying by 593.
We almost published it.
And it was a lie. Not because of the data, but because of where we're standing. The current cycle is halfway through its fall. Buying later comes out better for now, simply because the price keeps falling while we write this. We hadn't discovered any pattern: we had measured the present and mistaken it for a law.
It's the classic mistake and we fell right into it: if a number comes out perfect on the first try, be suspicious. Especially if the result depends on where you're looking from.
We're publishing it because it's exactly what a reader can't know unless we tell them.
The first version of this report said that 364 and 378 days went by from peak to lowest point. The 378 weren't made up: they're that same distance measured with the seven-day average, which gives 377. The 364 were measured with the closing price, without smoothing.
A raw number and a smoothed one, in the same row, subtracted from each other. That's where “fourteen days of difference” came from, which meant nothing: half of that difference was the method, not the market.
We caught it when checking why two of our documents gave different dates for the same bottom. The rule we take away, and it applies to anyone who measures anything: two figures can only be compared if they were measured the same way. If one is smoothed and the other isn't, the difference between them is partly your own.
Our own accumulation strategy starts buying back several months before where the cheap zone has historically been.
The three lowest points we measured came more than half a year after the moment it's already buying.
Put simply: our strategy empties the magazine too early. It spends the money ahead of time. We haven't fixed it yet, and we're not going to wait until we do to tell you.
This is where this stops being history and takes a position. Two paths, and neither looks at the price.
| Path | How it's calculated | Where it falls |
|---|---|---|
| 1 | This cycle's peak (6 October 2025) plus the days the two complete cycles took, with the three definitions of bottom | From 3 October to 16 November 2026 and from 3 to 18 October if you stick with the two finer definitions |
| 2 | The next halving minus the 513 to 527 days there were between the lowest point and the halving. And the halving doesn't fall in April 2028: at the 9.58 minutes per block the chain averages, it falls around 17 March | From 7 to 21 October 2026 |
They overlap, and that's more than we expected. They're two calculations that don't talk to each other — one starts from the peak, the other from the coming halving — and both fall in October. When two independent paths point to the same place, that zone deserves attention. It isn't a date: it's a window from October to early November 2026.
The first version assumed the 2028 halving on 15 April, which is the theoretical figure: 210,000 blocks at ten minutes. But the chain doesn't run at ten minutes, it runs at 9.58, and it has built up 220 days of lead since 2013. At the real pace the halving falls a month earlier, and path 2 moves from November to October. In other words: our own mistake was what kept the two paths from matching.
And what that means today. This cycle's lowest point so far is from 30 June 2026, 267 days after the peak. The two complete cycles took 364 and 366 days. Today we're at 303.
If the geometry repeats, the lowest point hasn't arrived yet.
At no point have we written “the bottom was 30 June”. And it isn't decorative caution: a bottom isn't a data point, it's a label you put on looking back. You only know a low was the low once enough time has passed without it breaking. For this cycle we have no bottom. We'll know where it was once it's behind us, as happened with the three before.
It's in the question you ask yourself. “How far will it fall?” has no answer — nobody knows, and whoever gives you a number is making it up. “Where are we along the way?” can be located, because the time elapsed is a fact, not an opinion.
And there's a technical reason to trust time more than price: a percentage can be adjusted until it looks nice. You change a threshold, move an average, pick another window, and the past looks beautiful. With dates you can't do that: the days that went by between two milestones are what they are, and there's no parameter that changes them.
What you take away is a change of question. You stop trying to guess the number — which is impossible — and start placing yourself on the map, which is possible. Placing yourself isn't predicting, and that's why it works.
It's the question you have to ask before believing any measurement, and we answer it before anyone asks.
The sample is three complete cycles, and each one contains a rise and a fall. That's different from other studies where the whole period was bullish: here, by construction, there's some of both.
What we haven't done: we haven't measured what exact percentage of those days was a rising market and what was a falling one, as we did in our report on leveraged products. We don't say it because we haven't calculated it, and we're not going to put down a number we haven't measured.
And where we are today: falling, 303 days after the peak. That is, inside the stretch where this measurement says the bottom hasn't been reached yet. That makes the check come soon, which is exactly what we want.
All of this is a description of what has happened, not a forecast of what will happen.
This stays dated 5 August 2026:
We expect this cycle's lowest point to fall between October and early November 2026. If it falls there, the geometry held. If it falls in February 2027, or if it turns out it already happened in June and we don't go lower, it broke — and we'll publish it all the same, in the same detail as this.
And we say “a window” and not “a day” on purpose. With the three ways of defining the bottom, the honest margin is a good month. Promising the day would be exactly the mistake this report calls out.
A method is only worth something if it gets right what it hasn't seen yet. Everything else is explaining the past, which is easy.
Three cycles are three data points. With three observations nothing is proven; at most, something worth looking at is pointed out. Anyone who presents three cases to you as a law is selling you something.
And the 2026 market isn't the 2018 one. Now there are listed funds buying spot bitcoin, companies holding it on their balance sheets, governments accumulating, and a mining industry competing for electricity with artificial-intelligence data centres. None of that existed when the two cycles we measured took shape.
What would knock it down, specifically: this cycle's lowest point falling clearly outside October–November 2026. If it comes in January, or in March, or if it already happened in June and doesn't go lower again, the geometry describes nothing. It's a clean test with a date.
That's why this isn't adopted: it's published and put to the test.
The underlying bias isn't specific to this industry. It's in anyone's head.
We look for the pattern that can be adjusted, because it's the one that proves us right. A threshold, a percentage, an indicator with three parameters: there's always a combination that makes the past look beautiful. And the more you tweak it, the more convinced you become that you've found something.
What's uncomfortable about a distance between two dates is exactly what makes it valuable: it won't let you cheat. What comes out comes out, and if you don't like it, there's nothing to tweak.
We're people who measure. We got it wrong three times inside this very report — the number that multiplied by 593 and was a mirage of where we're standing; a table that subtracted a raw number from a smoothed one; and the 2028 halving placed on the theoretical date instead of at the chain's real pace, which was exactly what kept our two paths from matching — and all three are told above, in full detail. And we've put in writing that our own strategy buys too early according to this very measurement.
We'd like you to check these numbers. Really. The dates are public and anyone can count them. If you find a mistake, a cycle that doesn't fit or an interpretation we missed, write to us and we'll correct it right here, in public and with your name on it if you want it to appear.
We don't publish to be right. We publish to be right more often, and for that we need more eyes than ours. A method nobody can break isn't a method: it's a belief.
hola@sophronepsis.com — everything is read and answered.
If you want to keep pulling the thread
We have another measured report on the signals people use to guess bitcoin's bottom: we took four and three didn't hold up. It's published, free and with no sign-up, along with the rest: sophronepsis.com/informes.html
And if what interests you is learning to look at this on your own instead of having it told to you, we have a six-day course, also open and with no sign-up, at app.sophronepsis.com/empieza.
And if you don't want this to stay a snapshot. A report is a measurement from one day. Inside the platform the follow-up does not stop: measurements are redone when new data arrives, each strategy comes with its test and its published failures alongside, and the mentor answers whatever you ask about this at any hour. Subscribe and follow it from the inside: sophronepsis.com/mentoria.html