Bitcoin week in review: $77,184 close misses the published range

The week, from open to close

Bitcoin traded between a weekly low of $62,751 and a high of $79,500, and finished +22.7% from Monday's open. The number that settles the range is the close, and this week the close was $77,184.

Measured against the centre of the projected band, that landed +19.0% away, with official resolution still pending while the checker validates the close inside its window. The verdict: outside the band.

Reading volatility off the range itself

The band is 8.3% wide between its edges. That single figure is the model stating, in advance, how much weekly movement it considered ordinary for Bitcoin right now. Wider bands mean the engine sees a jumpier market; narrower ones mean it is willing to commit.

It is a more useful reading than it looks, because it is the part of a forecast that people usually skip. A narrow band that holds is a real result. A wide band that holds proves very little, and the width is published precisely so nobody has to take the hit rate at face value.

Four models, one band

Set the price aside for a moment. The band is not one forecast dressed up as a range. It comes out of a combined engine of four models (a power law over a decade of price history, a Monte Carlo simulation driven by GARCH volatility, an Ornstein-Uhlenbeck residual model and an implied density read from the Deribit options surface), pooled and reweighted every week against real history. Each one is wrong in its own way, which is the point: pooling them cancels part of the error that any single Bitcoin model carries alone.

Around fourteen live data sources feed the bias layer on top, among them MVRV-Z, SOPR, NUPL, the Puell multiple, Reserve Risk, the Bitcoin realized price, funding rates and global liquidity. What comes out is a distribution of thousands of simulated price paths, and the published band is its middle half, from the 25th to the 75th percentile. Roughly one path in two ends inside it.

Forecast, not prediction, and the difference matters

Search for a Bitcoin price prediction and you will find a single number with a date attached and no way to check it later. This is the other thing. The band from $62,143 to $67,529 says where the model thinks price is more likely than not to land, it says so before the fact, and it gets marked against the real close afterwards.

The distinction is the same one a weather forecast makes. Nobody promises rain at four in the afternoon; they give you a probability and you decide what to do with the umbrella. A Bitcoin price target promises certainty that nobody has. A forecast hands you the odds and keeps the receipt.

Hits and misses, published the same size

Context first, verdict after. 6 of the 7 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 range and the person buying every month

Now the part most readers skip. Most people reading this are not trading the week. They are buying Bitcoin gradually, dollar cost averaging into a position they intend to hold for years, and a seven day band is not a signal to act on.

What it is good for is calibration. A band of $62,143 to $67,529 tells you what the engine currently considers an ordinary week, so when price moves inside that span you know you are watching noise rather than a change of story. The decision of whether to buy stays yours, and it should be driven by a plan written when you were calm, not by a week.

When the market gets loud

A 22.7% move in seven days is the kind of week that fills timelines. On the way up it arrives with talk of buying the dip too late and of the next all time high. Both conversations are older than most of the accounts having them.

The band was sized before any of that happened, which is the only reason it is worth anything now. A model that reacts to the mood is just the mood with a chart attached.

Where Bitcoin sits in its four year cycle

One layer deeper. Counting days since the April 2024 halving puts this piece on day 856. The next halving is estimated by block height for April 10, 2028, a date that drifts with the real cadence of mined blocks rather than sitting on a calendar.

That count is the honest way to compare cycles, because what matters is not the date but how far in we are. It is also worth saying plainly that four halvings means four cases. Enough to see a shape, nowhere near enough to prove one, and whether the four year cycle still holds is an argument that is very much open. The cycle is context here, never the reason for a weekly number.

Frozen in advance, on purpose

Zoom out for a second. This range was emitted on August 16, 2026 and resolves against the weekly close of August 23, 2026. Those two dates are not decoration. They are what separates a forecast from a comment, because once the first one passes the number cannot be edited.

Almost nobody in this corner of the internet publishes their errors. That is not a moral failing, it is an incentive: an unmarked forecast is always right in hindsight. The full record, including the ranges that missed, is public and dated, and it can be read as data rather than taken on trust.

What happens next

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.