We measured 129 ways of starting to buy bitcoin little by little. And we discovered something we weren't looking for: changing only the day you take the snapshot, the best entry moment becomes the worst.
The same measurement, with ten different time windows. In nine, starting in the bottom zone is among the best. In one, it's the worst of all. And that one was the one we had chosen.
We took 129 different moments over the last thirteen years. In each one we did the same thing: 100 dollars a month for two years, without looking at the price or skipping any. Then we grouped them by the point in the cycle where they started, counting months since the last halving.
The question was the usual one: is it worth waiting until it's cheap?
To know how each start turned out you have to value it at some point. And that point is chosen by whoever does the study. We had chosen four years.
So we repeated it changing only that: the same purchase, the same series of starts, changing only the day of the snapshot.
| Valued at… | 0-12 months | 12-24 | the zone · 24-33 | 33-48 | Rank of the zone |
|---|---|---|---|---|---|
| 30 months | 30 % | 113 % | 422 % | 154 % | 1.ª |
| 36 months | 101 % | 257 % | 528 % | 66 % | 1.ª |
| 42 months | 201 % | 605 % | 321 % | 109 % | 2.ª |
| 48 months | 559 % | 598 % | 200 % | 247 % | 4.ª |
| 54 months | 1.187 % | 352 % | 853 % | 406 % | 2.ª |
| 60 months | 1.109 % | 347 % | 1.088 % | 523 % | 2.ª |
| 66 months | 806 % | 1.236 % | 1.716 % | 759 % | 1.ª |
| 72 months | 914 % | 1.446 % | 3.342 % | 835 % | 1.ª |
Medians. Each row is the same purchase —$100/month for 24 months— valued at a different moment. The shaded row is the one we had chosen. Each window uses the starts that have that full window of data ahead of them, so they aren't the same in all of them: they range from 134 in the 30-month row to 92 in the 72-month one.
Correction — 17 August 2026
We redid the whole table to check it and the figures have changed slightly: no cell moved much, but several didn't match to the decimal what we had published. They've been replaced with the measured ones, and the calculation is saved in VENTANAS-CUANDO-MIRAS.json so it can be repeated. Two more things change. One: we said «the same 129 starts» in every row, and that can't be — each window can only use the starts that have that full window of data ahead of them, so they range from 134 in the 30-month row to 80 in the 84-month one. Two: we said the zone comes first in six of the ten windows, and it's five. What doesn't change is the conclusion: the zone's rank comes out the same as we published in all eight rows of the table, and it still goes from first to fourth just by moving the day of the snapshot.
A bitcoin cycle, from bottom to bottom, has lasted about four years. And four years was exactly our window.
That means that if you start near the bottom and look four years later, you're looking right at the bottom of the next winter. We weren't measuring whether entering low was good or bad: we were measuring where the snapshot fell. Move the snapshot six months and the result flips.
The window was synchronised with what it was measuring. And when that happens, the study no longer talks about the market: it talks about itself.
A few days ago we published that starting in the bottom zone came out worse than starting early, and at the end we wrote what would overturn it: «repeat it with three- and six-year windows; if the order flips, what we've found is a coincidence of choosing four».
We've repeated it. The order flips. That conclusion doesn't hold, and this page replaces it.
Because this mistake isn't rare: it's in almost every study you'll be shown, and no bad faith is needed to make it. Choosing a round time frame is enough.
The check takes a minute and requires no knowledge: «what if you look a year earlier or a year later?». If the answer changes, the study doesn't measure what it claims to measure.
And there's an especially treacherous case, which is ours: when the chosen time frame resembles the length of the cycle of what you're measuring. Four years in bitcoin. One year in something seasonal. A quarter in something paid quarterly. There, the result is almost always a mirage.
It doesn't say entering in the bottom zone is better. It comes first in five of the ten windows, but we don't claim that either: it's the same mistake in reverse. What we're saying is that with this sample the question has no answer, because the result is decided by a choice of ours and not by the market.
And it's four cycles. Four. And the starts overlap with each other —the one that began in January and the one that began in February share twenty-three of their twenty-four purchases—, so the hundred-plus in each row aren't as many independent tests.
A one-minute question that takes apart half of what you'll see: what if you look a year earlier or a year later? Always ask it, starting with what we publish.
Suspicion of the round time frame. When a study chooses four years, or one year, or a quarter, ask why that one. Often the answer is «because it looks good», even if whoever did it doesn't know it.
And something that takes the pressure off: if the moment of looking moves the result that much, the moment of entering weighs less than it's costing you in sleep. What you do decide, and what does matter, is how long you can go without touching that money.
If you want to keep pulling the thread
There's another report where the same trap shows up from a different angle: bitcoin's most repeated rule gives opposite results depending on what you compare it with. It's with the rest at sophronepsis.com/informes.html
And if you want to learn to look at this on your own, the six-day walkthrough is at app.sophronepsis.com/empieza.
A report is a snapshot of one day. Inside the platform the measurements are redone when new data arrives —this is an example of why that's needed—, every strategy comes with its test alongside it, and the mentor answers whatever you ask, at any hour. sophronepsis.com/mentoria.html
From 92 to 134 starts per window —each one uses those that have that full window of data ahead of them—, measured on bitcoin daily closing prices, from January 2013 to August 2026. Each start: 100 dollars a month for 24 months, valued at ten different moments between 30 and 84 months. Four cycles; the starts overlap with each other. Educational content, not financial advice. We don't sell signals or manage third-party capital. Familia FVR · Sophronepsis.