«Buy when the average holder is at a loss.» We tested it with two different comparisons, both reasonable. One says it's among the best decisions possible. The other says it's useless. Both are true.
Same data. Two ways of comparing it. Opposite results. What decides isn't in the market: it's in what you compare it against — and that is almost never mentioned.
The measure compares what bitcoin is worth today with what the coins in circulation cost on average. Below 1, the average holder is at a loss. The rule says: that's where you buy.
In thirteen years that has happened on 623 days, 13 % of the time, spread across five real episodes: late 2014, summer 2015, winter 2018-19, the March 2020 crash and the winter of 2022. We took each entry and looked at what happened at three, six and twelve months.
Against buying on any day in those thirteen years, the rule comes out worse:
| Horizon | With the rule | Any day | Difference |
|---|---|---|---|
| 3 months | −13,6 % | +8,0 % | −21.6 pts |
| 6 months | −5,1 % | +29,1 % | −34.2 pts |
| 12 months | +37,2 % | +75,8 % | −38.6 pts |
Now the same rule, but compared with buying on any day in the year before each signal —that is, against someone who was already buying while the market was falling:
| Horizon | Times it wins | Difference |
|---|---|---|
| 3 months | 16 of 18 | +7.6 pts |
| 6 months | 16 of 18 | +33.5 pts |
| 12 months | 18 of 18 | +83.3 pts |
Eighteen out of eighteen. With the first comparison the rule is a disaster; with the second, almost perfect.
Because they aren't the same question, even though they look like it.
Here the rival is the whole period, including the years of rising prices. And waiting turns out badly, because while you waited the price went up without you.
Here the rival is the fall itself. And waiting turns out well, because that threshold is only reached near the end of the crash.
The rule, then, is neither good nor bad: it's good for one question and bad for the other. Whoever sells it doesn't usually say which.
This is the part that serves you for the rest of your life and has nothing to do with bitcoin.
When someone shows you that a signal works, the question isn't «how much did it make?». It's «compared with what?». There are almost always two or three possible comparisons, they give different results, and the one they show you is the one that looks best.
You don't need to know statistics to ask for it. One sentence is enough: «show me what you compared it against, and what comes out with the other comparison». If there's no answer, you know what the number is worth.
We're not saying the rule works or doesn't: we're saying that the answer depends on the question. There are five real episodes —five— and our own yardstick asks for thirty. Also, the result shifts if you change the threshold: at 0.8 the sign flips at three months. It doesn't go into our trading, and the numbers are published in full in case anyone wants to redo them.
A question worth more than any indicator: compared with what? Always ask it. It's free and it disarms almost everything that gets published.
And the uncomfortable one, to use on us as well: in this very report we could have shown you only the second table —18 of 18— and you'd have come away convinced. Both are here because both exist.
If you want to keep pulling the thread
This same question —what it's compared against— has three new pieces applied to a system with entries and exits: how much history you need, why a high win rate isn't good news and the same test with twenty-five results.
The other reports are at sophronepsis.com/informes.html. The one on bottoms is there too, where you can see why that threshold of 1 might not come back.
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, every strategy comes with its test alongside it, and the mentor answers whatever you ask, at any hour. sophronepsis.com/mentoria.html
Measured over 4,963 days of public bitcoin price and on-chain data, from 1-Jan-2013 to 3-Aug-2026. Five episodes grouped; eighteen entries counting the separate stretches. Medians. Educational content, not financial advice. We don't sell signals or manage third-party capital. Familia FVR · Sophronepsis.