NeuPortal | ADA - dip below the core, then a three-week climbCardano / TetherUSBINANCE:ADAUSDTneuportalThe view: a drop through the lower edge of the measured zone first, then a sustained climb that runs well past the horizon this band actually covers. Both halves of that sentence matter. MEASURED, 24-bar horizon (about four days on this 4H chart): Median 0.1551 Core 50% 0.1486 - 0.1625 Wide 80% 0.1416 - 0.1701 125 independent windows, not the 3,001 overlapping ones FIRST, THE HORIZON MISMATCH, BECAUSE IT IS VISIBLE AND I WOULD RATHER NAME IT THAN LET SOMEONE ELSE: the shaded band ends around Jul 29. The drawn path runs to mid-August. So most of that arrow travels through territory the printed band says nothing about. The endpoint sitting above the band's ceiling is NOT a claim that price beats the 80% zone - it is a claim about a date three weeks past where the zone stops. Anyone reading the picture as "we expect a break of the band" is reading it wrong, and that is my fault for drawing across the edge. THE PART THAT IS A REAL CLAIM: the dip. The path goes below the 50% core floor at 0.1486 before turning. That is inside the band's own horizon and it is a genuine directional assertion - the model's sealed read for today is a SHORT lean with a median at 0.1569, so the down leg agrees with the engine while the recovery is my own. WHERE THIS BAND IS WEAK, AND WE FOUND OUT HOW WEAK THIS WEEK: out-of-sample coverage here reads 58.5% in the core against a 50% target and 89.3% wide against 80%. Too wide, not too narrow. That is not a footnote today. We audited our whole live board this week and found the same defect at scale: 31 of 36 resolved 24-hour forecasts landed inside a stated 50% band. Eighty-six percent. A walk-forward test over 20,058 observations showed the cause was not history length but the absence of regime conditioning - an unconditional band carries an allowance for turbulence that a quiet market has not earned. Conditioning the sample on today's volatility brings coverage to 50.8%. Our forecast engine now does that. The band you see plotted here still comes from the unconditional script, so read these zones as generous. Fixing the chart is next. Also on the panel: empirical width is 0.93x the sigma-root-t band, so the textbook formula is too wide on this instrument as well, and the tails are fat (excess kurtosis 1.95) with a mild left skew (-0.14). Invalidation for the down leg: acceptance back above the core top at 0.1625 before the dip happens. The forecast is the band. The drawn path is one illustration of a route, and a band is scored on coverage across many calls - which, as of this week, is a number we have publicly failed and publicly fixed. Educational content - not financial advice, and not a betting tip.