Sophronepsis · research · 5 August 2026
It's true, we measured it, and for one afternoon it looked like a solid finding. Here is the whole path from that afternoon to the moment we knocked it down ourselves, and the little that was left standing. Which is the part that's really worth something.
From a sentence said out loud on 4 August 2026: «what's real is the halving and the regulation of the miners' workload, roughly ten minutes. That's what varies over time; everything else hangs from it.»
And behind it there's a genuinely good idea. Almost everything measured in Bitcoin measures people — how many are in profit, how many sell, how long they've been sitting still. People change. Two things don't, because they're the law of the program: every 210,000 blocks issuance is cut in half, and every 2,016 blocks the network recalibrates the work. If the cycle has a real clock, it should be that one.
Our accumulation strategy counted months since the halving. The month is a human unit: it comes from the Moon and the Romans. Bitcoin doesn't know what a month is. So the question was clean:
If we measure the cycle in blocks instead of months, does the noise disappear?
We measured it the same day. And from here on this is a chronicle, not a result. It goes in order, with the stumbles where they happened.
To count blocks you need to know what height the chain was at each day. We rebuilt it from how many bitcoins were in circulation and what each block paid in each era. The proof that the reconstruction holds: the three halving dates it produces —9 July 2016, 11 May 2020, 20 April 2024— are the real ones.
746,026 blocks in 4,961 days. That's 9.58 minutes per block. Not ten.
It looks trivial. It isn't. At exactly ten minutes, those blocks would have taken 5,181 days. They took 4,961.
The chain is 220 days ahead of its own nominal clock. More than seven months gained since 2013.
The cause explains itself and it's not a flaw: between two recalibrations, the network's power almost always rises, so blocks come out a little faster than planned and the correction always arrives two weeks late. A few seconds every fortnight. Seven months in thirteen years.
Direct consequence, and it's the one that gave the idea wings: «month 27 after the halving» doesn't mean the same in 2015 as in 2026. In blocks, it does.
We took the three market bottoms already closed and measured how long it took to reach them from the previous halving, both ways.
| Bottom | Measured in months | Measured in blocks |
|---|---|---|
| 14 January 2015 | 25,5 | 128.984 |
| 15 December 2018 | 29,2 | 134.001 |
| 9 November 2022 | 30,0 | 132.481 |
| How far apart they are from each other | 6,9 % | 1,6 % |
Measured in blocks, the three bottoms fall four times closer together than measured in months.
This is where people publish. The idea was attractive, the number backed it up, and on top of that the explanation made physical sense. If this were a channel, the video would already be up.
We were measuring from the halving. And the halving falls every 210,000 blocks by definition, not by chance.
In other words: by measuring in blocks from a point that is already fixed in blocks, part of the regularity was being put there by us. We weren't discovering it. It's like measuring the distance between two kilometre markers and being surprised that it comes out at exactly one kilometre.
The objection was written in the document itself before celebrating anything. That's the procedure, and that's why it worked.
If the reference is the bottom and not the halving, the trap disappears. We measured from bottom to bottom.
| Cycle | Date of the bottom | Block height | Price |
|---|---|---|---|
| 1 | 14 January 2015 | 338.984 | 176 $ |
| 2 | 15 December 2018 | 554.001 | 3.185 $ |
| 3 | 9 November 2022 | 762.481 | 15.758 $ |
| 4 in progress | 30 June 2026 low so far | 956.104 | 58.525 $ |
| From one bottom to the next | Blocks | Days |
|---|---|---|
| 2015 → 2018 | 215.017 | 1.431 |
| 2018 → 2022 | 208.480 | 1.425 |
| Difference | 6.537 | 6 |
| How far apart they are | 3,09 % | 0,42 % |
Measured from bottom to bottom, the wall calendar is seven times more regular than blocks. One thousand four hundred and thirty-one against one thousand four hundred and twenty-five days: six days of difference in almost four years.
Exactly the opposite of the hypothesis. And with a reasonable reading behind it: what governs the length of the cycle isn't block production, but something with a calendar rhythm — the money available in the world, economic cycles, human behaviour. The protocol sets issuance. The calendar sets the cycle.
Six days in four years. The beautiful number again. The urge to publish again.
«The bottom» is the day of the lowest price. But which day that was is only known years later. Picking it after the fact and then boasting about the precision is cheating without meaning to.
The objection, once again, had been written beforehand. So we repeated the measurement defining the bottom three different ways — the lowest day, the lowest week and the lowest month — and checked whether the 0.42 % survived.
| How we define «the bottom» | Dispersion in days | Dispersion in blocks |
|---|---|---|
| The day of the lowest price | 0,42 % | 3,09 % |
| The lowest 7-day average | 1,05 % | 1,69 % |
| The lowest 30-day average | 5,18 % | 8,02 % |
Two conclusions, and they're different from each other.
What holds: the calendar beats blocks with all three definitions. 0.42 against 3.09; 1.05 against 1.69; 5.18 against 8.02. The direction of the finding is solid.
What doesn't hold: the precision. Smoothing over a week, the number doubles. Smoothing over a month, it multiplies by twelve. The regularity of «six days in four years» doesn't exist. There is a much coarser regularity.
And it shows mercilessly in the projection. Depending on how you define the bottom, the fourth would fall on 7 October, 29 October or 5 November 2026. The honest window isn't six days: it's more than a month. It's still information. It's not the same as what we said half an hour earlier.
A measurement that only works with the most favourable definition isn't a finding. It's a well-chosen coincidence.
Up to here it was all describing the past, which is easy. The real test is a different one: if I change the strategy's unit of time, do I make more?
We simulated putting in 100 dollars a month from January 2013 to 2 August 2026 —16,400 dollars contributed in total— with the rules of our accumulation strategy. The only thing that changes between variants is the clock used to count the window. Everything else, identical.
| Variant | On what was contributed |
|---|---|
| Buy every month and never sell (the benchmark) | 127× |
| Counting months since the halving (what we were doing) | 291× |
| Counting blocks since the halving (the new idea) | 189× |
| Counting days since the previous bottom | 185× |
Counting blocks doesn't improve the strategy: it makes it 35 % worse.
The hypothesis is rejected in the only test that mattered. The idea was good, the physics was good, the first measurement was good, and it doesn't work.
A single number says nothing, so we moved the window's starting point across ten different positions, with all three clocks. We have the full table, but the summary is this:
| Clock | Worst point | Best point | How much it varies |
|---|---|---|---|
| Months since the halving | 1.0 M$ | 8.0 M$ | 51 % |
| Blocks since the halving | 0.8 M$ | 8.0 M$ | 55 % |
| Days since the bottom | 0.8 M$ | 6.1 M$ | 49 % |
Read it slowly, because it's the most important thing in the whole document:
The result multiplies by eight depending on where you put the start of the window. Changing the clock moved it by a third. Changing that point moves it eightfold.
That starting point, in our strategy, isn't measured. It's chosen. Someone put it there months ago because it seemed reasonable, and nobody ever checked it.
There's some good news inside: the point we use isn't the best in the sweep, not by a long way. That shows nobody tuned it to the historical maximum, which is the classic trick of anyone showing a pretty backtest. But the bad news wins out: what decides that strategy's result isn't any of the things we were discussing. It's a number nobody has justified with data. And it had been there since before this research.
We were looking into whether the clock was the right one. We found that the clock hardly matters and that the problem was somewhere else. The mistake was in the question, not in the answer.
Bitcoin doesn't know what day it is. All it counts is blocks, and the network tries to produce one every ten minutes by adjusting the difficulty every 2,016 blocks —about two weeks—. If more power comes in to mine, blocks come out ahead of time until the readjustment arrives; if power leaves, they come out more slowly.
Since between two readjustments more power almost always comes in than there was before, the chain runs chronically ahead of the calendar. That's where the idea we measured came from: if the real clock is blocks and not months, maybe cycles are better understood by counting them that way.
Even if the idea falls apart —and it does, as told above—, what remains is worth more than the idea: this report is the map of how a pretty finding turns into nothing when you check it.
And the pattern is recognisable. A round number came out, it fitted too well, and that was exactly the warning sign. Almost every time a result comes out that clean, the trap isn't in the market: it's in how we measured it.
Next time you see a chart where everything fits suspiciously well —yours or someone else's—, you already have the three questions that knock it down: has this been counted twice without noticing? does it hold if I move a parameter? what would have happened by ignoring it?
This describes what has happened, not a forecast of what will happen.
This document has three mistakes of our own and nobody from outside found any of them. We found all three ourselves, and always for the same reason: the objection was written before looking at the result.
That's the whole technique. There's nothing more.
Write down the objection before you see the number. After you see it, it no longer occurs to you — because you like the number.
It's a bias of the mind, not of the sector. It happens just the same to a doctor with a favourite diagnosis and to a mechanic with the usual fault. The difference between someone who measures and someone who believes isn't intelligence: it's whether they wrote down in advance what would make them change their mind.
We had a good idea at seven in the evening, measured it at eight, it looked brilliant at nine and by eleven it was dead. We tell it in full because the process is worth more than the finding.
We're people who measure. We got it wrong three times within this very document —we measured from a point that already imposed the result, we boasted of a precision that was the low day chosen after the fact, and we discovered that the parameter with the most say in our own strategy isn't measured— and all three are told above, in full detail.
We'd like you to check these numbers. Seriously. Block heights are public, closing prices are public and the calculation is described step by step so anyone can repeat it. If you find a mistake, a period 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
If you've made it this far without knowing what a block is or why the halving isn't every four years, start with the first in the series. And if what caught your interest is the calendar part, the piece where we measure the distances between the milestones of the three cycles is the one that goes furthest. All published, free and without signing up: sophronepsis.com/informes.html
And if what you want is to learn to do this yourself —write down the objection before looking, and hold firm when you like the number—, the six-day walkthrough is open and without 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