Our 7-day Bitcoin ranges held 13 times out of 13. That is not the good news it sounds like

A range that says 80% should miss about one week in five. Ours has not missed once, and even the narrower 50% range caught every close. Here is what that tells us, what it does not, and why we are not changing anything yet.

By Andres Gayá · October 28, 2026

Every day we publish where Bitcoin is likely to close seven days later: an 80% range, a narrower 50% range inside it, and the chance of closing above four price levels. Seven days later, each range is graded against the real close, and the result stays on the page whether it held or missed.

As of October 28, thirteen of those 7-day ranges had been graded. All thirteen held.

That sounds like a win. It is closer to a warning.

What “80%” promises

An 80% range is not a promise to be right. It is a promise about how often it will be wrong: about one time in five, the close should land outside it. A range that never misses is either lucky or too wide, and a range that is too wide is useless in a different way. It tells you less than it could.

The 50% range is a stricter test. By construction, the close should land inside it about half the time. Bitcoin’s close landed inside our 7-day 50% range in all thirteen graded forecasts.

The record, as of October 28

So far the 1-day ranges are close to what the percentages promise, a little over eight in ten inside the wide range and a little over half inside the narrow one, though 19 is far too few to judge. The most recent of their three misses is on the page in red: on October 22 we gave Bitcoin $82,456–$85,764 for the next day’s close, and it closed at $81,416.

The 7-day ranges are not close, at least not yet.

Why thirteen out of thirteen proves less than it seems

If the 7-day ranges were exactly right and every forecast were a separate test, thirteen out of thirteen inside the 80% range would happen about one time in eighteen. That alone would not mean much. The 50% range is the surprise: thirteen in a row inside it would happen about once in 8,000 tries.

But they are not separate tests. We publish a new 7-day range almost every day (two days are missing from the record, October 10 and 11), so a forecast made on Monday and one made on Tuesday cover six of the same seven days. Thirteen daily forecasts made between October 7 and 21, and graded between October 14 and 28, are far fewer than thirteen separate tests. Taking the overlap into account, a perfectly calibrated model would put all thirteen closes inside its 80% range about one time in four, which is unremarkable. It would put all thirteen inside its 50% range about one time in ninety. That is unlikely, but it is three quiet weeks, and we only looked closely because the number stood out.

So the honest reading is: in these three weeks the 7-day ranges were probably too wide, especially the 50% range. Whether that holds beyond a quiet stretch, the record is not yet long enough to say.

What actually happened to Bitcoin

Part of the answer is the market. Over the last ten graded dates, Bitcoin closed between $80,151 and $84,093. Our 7-day 80% ranges in that stretch were about $8,700 to $10,100 wide.

Bitcoin’s daily moves were about as large as the model expected, but they kept reversing, so no week added up to much: the biggest 7-day move in the graded stretch was 3.0%. A model that scales one day’s volatility to seven days assumes the moves add up. In these weeks they cancelled out.

Ethereum and Solana did miss. Over the same thirteen 7-day forecasts, Ethereum’s close landed inside its 80% range 12 times and inside its 50% range 9 times. Solana’s: 11 and 9. Those are closer to what the percentages promise.

Why we are not changing anything

It would be easy to narrow the Bitcoin ranges today and make the record look sharper. We will not, for two reasons.

First, three weeks is not enough. We have said that we will not claim accuracy, in either direction, before roughly 90 graded forecasts per horizon.

Second, the rule for changing the model is written down and fixed in code (we wrote it on October 24, before any recalibration), and it is on our methodology page. The earliest a recalibration can happen is January 15, 2027, and it follows that rule, not a judgment made after a good or bad month. Any change gets a new version number and a dated note. Old forecasts keep the version that made them, and nothing is recomputed after the fact.

If more than 86% of 7-day closes still land inside the 80% range when at least 90 are graded, the rule will narrow it by 5%, at most one step per quarter, on the record, with a date.

This week’s 7-day ranges

For reference, here is how far each of the seven US spot-ETF coins could move by November 4, at 80%, according to the ranges published on October 28. None of this says which way.

Each of these will be graded on November 4, hit or miss, on its coin’s page.