Measured, not assumed. Both bands below are interquartile ranges of how BTC has actually moved over the next 24 four-hour bars, read straight off this chart's own history. No distributional assumption, no fitting.

UNCONDITIONAL, all regimes, 3,001 overlapping windows:
Median 63,559
Core 50% 61,978 - 64,967 (4.7% of spot)
Wide 80% 60,234 - 66,471 (9.8%)

REGIME-MATCHED, only windows that opened in conditions like today's:
Median 63,557
Core 50% 62,164 - 64,972
Wide 80% 60,450 - 66,265

WHY TWO NUMBERS

The second band is the one we now seal. An unconditional band is read from every kind of market at once, so it carries a permanent allowance for turbulence that a calm tape has not earned. We found that in our own record last week and published the diagnosis before the fix: across resolved forecasts, our stated 50% intervals were containing 84% of outcomes. Too wide is a failure, and it is the failure nobody catches, because every result keeps landing inside and nothing ever looks broken.

Conditioning on the volatility regime observed at the moment the forecast opens - not at the close, which would leak the answer - is the correction. On BTC today the two bands are close, which is itself informative: current conditions are near the middle of the historical distribution, so filtering changes little. On a quiet day the gap is much wider.

The cost is sample size, and it is stated on the chart rather than hidden. The unconditional band rests on 125 genuinely independent windows. The regime-matched one rests on 41. Overlapping windows share 23 of their 24 bars, so quantile error bars scale with the small number, and 41 is a small number.

NO DIRECTIONAL LEAN

The regime line reads range with no trend, RSI is 48.8, and the median sits within 0.1% of spot. This is a statement about how far, not which way.

Note the tension, because it is visible on the panel and I would rather name it: ADX prints 28.5, which is nominally strong, while the regime read says range. Those disagree. When they do, the band is the part worth reading and the direction is not.

WHAT THE TABLE ADMITS

Out-of-sample coverage, fitted on older history and tested on newer it never saw: core 46.5% against a 50% target, wide 77.7% against 80%. Both slightly under. That is the closest to calibrated of anything we publish, and it still means the band is marginally too narrow rather than perfect.

Empirical width is 0.95x the textbook sigma-root-t band, so the conventional formula is slightly too wide here. Skew -0.56 and excess kurtosis 1.48: fatter left tail, fat tails generally.

Widths do not scale the way diffusion says. 4.7% at 24 bars, 8.0% at 48 (1.69x where root-t expects 1.41x), 9.6% at 72 (2.03x against 1.73x). Every step widens faster than independent increments would.

SEPARATELY SEALED TODAY

Our 24-hour forecast for BTC was sealed and Bitcoin-timestamped this morning at 07:32 UTC, before publication and on a different horizon from the four-day band plotted here: core 62,766 - 64,040, with a short lean and invalidation at 64,731. Different window, different claim, and both are on the record.

Nothing here is a signal, a target or a recommendation. It is a description of how this market has moved before, with its probability stated and its sample size printed.

Educational content - not financial advice.