The Bitcoin week, settled: +0.5%

How the week actually went

Bitcoin opened the week at $84,472 and goes into Friday night at $84,884: up 0.5% on the week. The best day was Thursday at +1.5%, the worst was Monday at -1.2%. The full day-by-day sits in the data box above.

That is the recap's whole job: what actually happened, priced and dated, before the weekend rewrites it into a story. Whatever narrative the timeline settles on, these closes are the part nobody gets to edit.

Last week, same series: Bitcoin this week, day by day: +4.2%. That range is already settled and published, hit or miss.

The range comes out of a four-model engine, reweighted every week against the real record. How it is calculated, step by step.

The committed range, checked against tonight

Zoom out for a second. The committed weekly range said $77,588 to $81,672, frozen before the week started. Tonight's price of $84,884 sits outside it, 3.9% beyond the top edge.

The range does not settle tonight: it resolves at its own expiry, and the verdict gets published either way, like the 8 of 12 weekly ranges before it. This is the midweek-to-Friday photo, not the final whistle.

Weeks that looked like this one

Scanning roughly nine years of daily closes for weeks with a similar move and a similar range turns up 47 precedents. In the seven days that followed them, the median move was -0.1%, and 22 of the 47 closed higher.

Read that as a base rate, not a forecast. 47 weeks is a small sample, no two market regimes rhyme exactly, and the model behind the published ranges does not use this table. It is context for the weekend take, nothing more.

Which chapter this week was part of

Set the price aside for a moment. Read through the structure engine behind the market moment alerts, this week belongs to a correction inside an uptrend: the weekly degree points up and the daily one down. That is the frame the week's candles were painted inside.

The Tuesday piece digs into that structure in detail. The recap only needs the label: a week means something different inside an uptrend than inside a correction, and pretending weeks float free of context is how recaps mislead.

Hits and misses, published the same size

8 of the 12 projected weekly ranges published so far resolved inside the band. Every one of them was frozen before its week, and every result went up afterwards, including the ones that went wrong.

There is a trap in that number worth naming. A band that contains the close every single time is not a good model, it is a wide one. The published band is the middle half of the distribution, so a properly calibrated engine should land inside about half the time. A hit rate far above that means the bands are wider than they should be, which is a miscalibration too, just the flattering kind.

The next number is already live

Context first, verdict after. The next projected range is drawn over the live Bitcoin chart in the app, alongside the dated zones for the longer horizons, and it is recalculated on the server every day so that every visitor sees the same numbers. Nothing here is computed in your browser and nothing is personalised.

None of this is investment advice. It is one statistical model keeping its numbers where anybody can check them, including the weeks it gets wrong.