The 60-minute window · 09:30–10:30
The longer window waits for the whole first hour to complete before it commits. It gets a bigger, better-defined range to work from and, usually, a target that sits further away in points.
A clean Atlas-mode morning
🧲✅ favoured side reached first · ~10:51- IB range371.75 pt · avg5 492.50 · ⚖️ Normal
- Split▲ 69% · ▼ 13% · ∅ 19% (n 620)
- Shorter horizons▲ 2h 54% · 4h 63%
- Target distance≈93 pt (0.25 × range)
The window opened with a sharp drop and spent the hour climbing out of it, closing at the top of its own range. Comparable windows went on to reach the upper target first 69% of the time, clear of the 62% threshold, so the box tints, the upper rail takes the magnet and the morning is in Atlas mode.
Price left the box on the upside almost immediately and tagged the target at about 10:51, twenty minutes after the window closed. The table had said 54% of comparable windows resolved that way inside two hours; this one took twenty-one minutes.
Abstain, and the abstain was the useful part
✅ larger share correct · ~12:47- IB range356.25 pt · avg5 458.55 · ⚖️ Normal
- Split▲ 49% · ▼ 11% · ∅ 40% (n 620)
- Shorter horizons▲ 2h 27% · 4h 40%
- Target distance≈178 pt (0.5 × range)
A visually similar morning to Monday, an upward hour closing near its high, but the numbers are not similar at all. The upside gets 49%, well short of the threshold, and the largest single share on the chart belongs to ∅ 40%: neither target reached at all. Both rails stay grey and the badge says abstain.
The upside did eventually win the race, at about 12:47, more than two hours after the window closed. That is the shape the shorter-horizon figures were describing before it happened: only 27% of comparable windows resolved up inside two hours, rising to 40% by four. A lean you have to wait two and a quarter hours for is a different trade from Monday's, and the chart said so at 10:30.
The quietest hour of the week
✅ larger share correct · ~11:57- IB range172.25 pt · avg5 451.85 · 💰 Accumulation
- Split▲ 4% · ▼ 25% · ∅ 71% (n 404)
- Shorter horizons▼ 2h 9% · 4h 17%
- Target distance≈172 pt (1.0 × range)
172 points against a five-session average of 452, roughly a third of recent behaviour, which is what earns the 💰 ACCUMULATION tag. Price spent the hour rolling over inside the box and closed the window on its low.
Note what the compression does to the geometry: because the window is small, the automatic ladder widens the multiple to 1.0 × the range, so the target still sits 172 points away, and ∅ jumps to 71%. The single most likely outcome, by a distance, was that neither target would be reached at all. The downside got there in the end, at about 11:57, and because 25% was the larger of the two shares the mark prints as a tick, but it prints without a magnet, because this window never offered a lean worth the name.
The week's most instructive pair: a lean that was right about direction and wrong about reach
∅ neither target reached, both sessions- Thu · IB range425.50 pt · avg5 394.60 · ⚖️ Normal
- Thu · Split▲ 65% · ▼ 18% · ∅ 17% (n 570)
- Thu · Horizons▲ 2h 50% · 4h 58%
- Thu · Target≈106 pt (0.25 × range)
- Fri · IB range257.50 pt · avg5 394.35 · 💰 Accumulation
- Fri · Split▲ 19% · ▼ 14% · ∅ 67% (n 470)
- Fri · Horizons▲ 2h 7% · 4h 12%
- Fri · Target≈258 pt (1.0 × range)
Thursday produced the week's widest window at 425.50 points and a genuine Atlas-mode lean: ▲ 65%, magnet on the upper rail, target 106 points above the window's high. Worth noticing that the two range tags disagree here: against the last five sessions this hour reads ⚖️ NORMAL, but against seventeen years it is large enough to count as expanded, which is why the automatic ladder chose the narrowest multiple it has, 0.25 ×, rather than widening the target further.
It still was not enough. For the rest of the session price never traded more than about thirty-five points above the window's high. Neither rail was reached, the race closed unresolved at the session boundary, and no outcome mark prints. This is the ∅ 17% branch of the split, the least likely of the three, and it happened.
Then it went. Follow the chart across the overnight session: price grinds higher through the evening and Asia, and at 08:38 on Friday, eight minutes after the payrolls release, it trades clean through the level Thursday's rail had been sitting at. Directionally the 65% was right. It was right roughly twenty-two hours late, which for a window whose race is scored to the session close is the same thing as being wrong. That gap is exactly what the ∅ column is there to warn you about.
Thursday was also the session before Non-Farm Payrolls, and the desk convention that pre-payrolls tape stalls (that price is reluctant to commit to a direction until the number is out) is widely held. We have not measured that as part of the Atlas tables, so treat it as context rather than a statistic; but it is a reasonable description of a session that leaned up, went nowhere, and then moved the moment the release cleared.
Friday, the payrolls session itself, is the mirror image. The move happened at 08:30, an hour before the window even opened; by 09:30 the tape had already spent itself, and the cash-open hour printed just 257.50 points against a 394-point five-session average, 💰 ACCUMULATION. The split is flat and weak in both directions, ▲ 19% against ▼ 14%, so both rails go grey: abstain. And ∅ 67%, two thirds of comparable windows reached neither target. Neither target was reached. The single most likely outcome on the chart was "nothing happens here", and nothing happened here.
