Five 09:30 cash-open sessions, each read at both durations Atlas IB offers, a 60-minute window and a 30-minute one. Ten readings, seven of them leans. Eight of the ten reached neither target by the close, which equals the quietest week on the record, and the two that did resolve were the same price move, at 11:20 on a Labor Day session that shut at one o'clock. Nothing was marked against a lean, although Thursday's half hour came within a point and a half of being.
Every window that appeared, in the order the indicator produced them.
| Session | IB range | Box tag | ▲ | ▼ | ∅ | n | State | Target (×0.5) | Outcome |
|---|---|---|---|---|---|---|---|---|---|
| 60-minute window · 09:30 to 10:30 | |||||||||
| Mon 7 Sep | 44.50 | 💰 Accumulation | 18% | 41% | 41% | 410 | 🧲 Atlas ▼ | ▼ 22.3 pt | 🧲✅ ~11:20 |
| Tue 8 Sep | 262.25 | 🚀 Expansion | 24% | 35% | 41% | 347 | abstain | ▼ 131.1 pt | ∅ neither |
| Wed 9 Sep | 177.00 | ⚖️ Normal | 40% | 26% | 34% | 373 | abstain | ▲ 88.5 pt | ∅ neither |
| Thu 10 Sep | 205.25 | ⚖️ Normal | 44% | 20% | 36% | 466 | 🧲 Atlas ▲ | ▲ 102.6 pt | ∅ neither |
| Fri 11 Sep | 127.75 | ⚖️ Normal | 47% | 15% | 38% | 581 | 🧲 Atlas ▲ | ▲ 63.9 pt | ∅ neither |
| 30-minute window · 09:30 to 10:00 | |||||||||
| Mon 7 Sep | 43.25 | 💰 Accumulation | 33% | 50% | 17% | 347 | abstain | ▼ 21.6 pt | ✅ ~11:20 |
| Tue 8 Sep | 232.00 | 🚀 Expansion | 28% | 53% | 19% | 440 | 🧲 Atlas ▼ | ▼ 116.0 pt | ∅ neither |
| Wed 9 Sep | 150.00 | ⚖️ Normal | 65% | 16% | 19% | 635 | 🧲 Atlas ▲ | ▲ 75.0 pt | ∅ neither |
| Thu 10 Sep | 146.00 | ⚖️ Normal | 28% | 53% | 19% | 440 | 🧲 Atlas ▼ | ▼ 73.0 pt | ∅ neither |
| Fri 11 Sep | 127.75 | ⚖️ Normal | 65% | 16% | 19% | 635 | 🧲 Atlas ▲ | ▲ 63.9 pt | ∅ neither |
▲ / ▼ / ∅ are the measured shares of comparable historical windows in which the upper target was reached first, the lower target was reached first, or neither was reached by the session close; n is how many such windows sit behind them. They are a historical null and not a prediction of the day being read. Every reading on this page was taken with the target multiple set to ×0.5, the setting the Atlas IB handbook recommends for NQ and the default in the current build, unchanged since w/c 17 August; the first two weeks on the record ran on Auto. These captures are 5-minute charts, so outcome times are the bar in which the rail was crossed and are good to the bar. The session close is normally the 16:55 ET bar, and the eight windows that reached neither rail were scored to it. Monday is the exception: CME index futures closed at 13:00 ET for Labor Day, so Monday's last bar was 12:55 and neither Monday window could have run to 16:55. Both resolved at 11:20 regardless, so nothing turns on it.
Two comparisons appear in every capture and they answer different questions. The box tag grades the window against its own last five sessions (💰 accumulation, ⚖️ normal, 🚀 expansion), while the status table's regime row grades it against the full seventeen-year history, and the two are allowed to disagree. This week they read the same way on eight of the ten windows: Monday's pair accumulation and compressed, Tuesday's pair expansion and expanded, the rest normal on both. The two that split were Thursday's hour, ⚖️ Normal on the box against expanded in the table, and Friday's hour, ⚖️ Normal on the box against compressed. With the multiple fixed at ×0.5 the regime no longer moves the rails, so it is reported here but it is not doing anything to the targets. The Window strength row read 68% favoured-rail all week, #11 of 91 for the hour and #14 of 91 for the half hour; it describes the window rather than the day.
The red folder USD calendar was light and then heavy. Monday 7 September was a US bank holiday, Labor Day, with the early close noted above. Tuesday and Wednesday carried no high-impact releases. Core PPI and PPI printed at 08:30 ET on Thursday and CPI at 08:30 ET on Friday, in both cases an hour before the window opened. Timing is all that is claimed for them here.
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, at a fixed multiple, a target that sits further away in points.
Labor Day. CME index futures closed at 13:00 ET, so the whole session was three and a half hours long, and the first hour of it covered 44.5 points against a five-day average of 158.6: 💰 ACCUMULATION on the box, compressed in the table. It is the smallest 60-minute box on the record by more than fifty points, and the half hour beneath it is smaller still at 43.25, the smallest initial balance of either duration.
The read was 18 / 41 / 41 from 410 sessions, and the table's own directional row printed 69%, past the 62% the Atlas mode asks for, so the magnet went on the lower rail with the target 22.3 points below the box. Price held near the top of the range until about eleven, then fell away and crossed at 11:20. Magnet and tick. The capture stops at the mark; the session itself ran on to the 12:55 bar.
This is the fifth draw of this cell on the record, after 19, 20, 21 and 24 August, and its second payment. It is also the second nearest rail the record has produced, after the 21.6 points on the half hour beneath it, and the obvious question, whether a rail that close is an easier thing to reach than a rail a hundred points out, is not one that a single holiday morning can answer. The percentages are a historical null, not a prediction, and they are not adjusted for the fact that this box was a quarter the usual size.
262 points in the first hour against a 134.9 average, 🚀 EXPANSION on the box and expanded in the table, nearly six times Monday's range. The read was 24 / 35 / 41 from 347 sessions, the cell Monday 17 August and Friday 4 September both returned, and the table's directional row printed 59%, three short of 62. Both rails stayed grey.
A 262-point box at ×0.5 puts the rails 131 points out on each side, at 29,817.88 and 29,293.38. Price spent the rest of the session inside a band 156 points wide, 29,645 down to 29,489.5, and closed in the lower part of it. The nearer of the two rails was still 173 points away when the 16:55 bar printed, the widest margin of the week.
Third draw of the cell: reached on 17 August, reached on 4 September, neither today. All three abstained, and on all three ∅ was the largest of the three shares, which is the branch that landed here.
177 points against a 159.55 average, ⚖️ NORMAL on both rulers, and the first appearance of 40 / 26 / 34 from 373 sessions anywhere on the record. The table's directional row read 61%. The Atlas mode asks for 62, so this window abstained by one point and both rails printed grey.
Price broke down out of the box before noon, made 29,358.25 at 11:20, and spent the afternoon chopping between about 29,385 and 29,485. The high after the box, 29,563.25, finished 119 points under the upper rail at 29,682.50. The low was 29.75 points above the lower rail at 29,328.50, the closest approach of the day, and it was on the side the table had put 26% on.
An abstained window is graded on the larger of its two shares, which here was the upside, and the upside was never in the picture. Nothing was offered on this morning and nothing is claimed for it. A single point on a rounded percentage decided whether this window carried a magnet, which is worth knowing about the threshold rather than about the day.
Core PPI and PPI printed at 08:30 ET, an hour before the window opened. The hour that followed was 205 points, ⚖️ NORMAL against a 159.6 average and expanded in the table, and read 44 / 20 / 36 from 466 sessions with a directional row of 69%. Magnet on the upper rail, target 102.6 points above the box. The release is where it is in the timeline and nothing here is attributed to it.
The high for the rest of the session came on the 10:35 bar, minutes after the window closed, at 29,275.25, only 12.25 points above the box high and ninety points under the rail at 29,365.63. Nothing traded higher for the rest of the day, but it was not a one-way slide either. Price fell to 29,133.5 by 12:35, then put 133.5 points back on in twenty-five minutes to reach 29,267 at 13:00, within 8.25 points of the morning's high and still 98.6 short of the rail. From there it fell away to 29,123 at 15:50, 168 points above the lower rail, and closed at 29,142.25. The whole post-box range was 152.25 points, and neither target by the 16:55 bar.
This is the one morning of the five on which the two durations favoured opposite rails. The half hour closed its box 59.25 points lower at the top, from the same low, read a different cell and leaned short. Same session, same low, opposite magnets; the half hour is further down.
CPI printed at 08:30 ET, an hour before the window opened; the pre-market bars to the left of the box are that hour. The hour itself was 127.75 points, ⚖️ NORMAL against a 165.45 average and compressed in the table, and the first half hour had already set both extremes: the 09:30 to 09:55 range was the 09:30 to 10:25 range. The two windows therefore share this box, these rails and this target exactly.
The hour read 47 / 15 / 38 from 581 sessions, directional row 76%; magnet on the upper rail, target 63.9 points above the box. Price fell out of the box first, to 29,334 by 10:45, then turned and rallied to 29,499.5 at 11:30, 39 points short of the rail at 29,538.88. It drifted lower for the rest of the afternoon and closed at 29,392.75. Neither target by the 16:55 bar.
Second draw of 47 / 15 / 38: paid on 3 September, not reached today. Those 39 points are the closest any window came to its favoured rail this week without reaching it, level with the half hour, which is the same number because it is the same rail.
The shorter window commits half an hour earlier, from a smaller range, and its target is therefore usually nearer. It is a different measurement of the same morning, not a second opinion on it.
The first half hour of the holiday session covered 43.25 points, 💰 ACCUMULATION against a 132.8 average, and the second half hour added 1.25 points to it, all of them at the low. The read was 33 / 50 / 17 from 347 sessions and the table's directional row printed 61%, one point under the test, so both rails stayed grey.
The lower rail sat 21.6 points under the box and went on the 11:20 bar, the same bar that paid the hour. No lean had been offered, so the mark is a plain green tick recording that the larger of the two grey shares got there first.
Third draw of 33 / 50 / 17, all three abstains: a cross on 27 August, neither on 31 August, a tick today. Read the two ticks on this page as what they are. Both windows were settled by one price move on one morning, and a move that clears a 22-point rail clears a 21.6-point rail in the same instant.
232 points in the first half hour, twice the 115.4 average, 🚀 EXPANSION on the box and expanded in the table. The read was 28 / 53 / 19 from 440 sessions, the most frequently drawn cell on the record and its seventh appearance, with a directional row of 66%. Magnet on the lower rail, target 116 points below the box.
Price came back up into the box within the hour and stayed between 29,645 and 29,424.5 for the rest of the day. The lower rail at 29,338.75 was 86 points below the session low when the 16:55 bar printed, and the upper rail was never a question.
Eight draws of this cell now: three paid, five reached neither, none against. It comes back on Thursday from a box 86 points narrower and with the same three shares, and does the same thing; that is two entries down.
150 points, ⚖️ NORMAL against a 138.25 average. 65 / 16 / 19 from 635 sessions, directional row 81%, is the most one-sided cell the record has drawn since the multiple was fixed at ×0.5 on 17 August. Magnet on the upper rail, target 75 points above the box.
The high after the box came in the ten o'clock hour at 29,594 and finished 48 points short of the rail at 29,642. Price then sold off to 29,358.25 by 11:20 and chopped sideways for the rest of the day. That low ended up 16.25 points above the lower rail at 29,342, nearer to the rail the table had put 16% on than the day ever got to the one it favoured.
Third draw of this cell: paid on 26 August, marked against on 4 September, neither today, and neither again on Friday from an identical reading. Four draws of a cell measured over 635 sessions settle nothing about it in either direction.
146 points, ⚖️ NORMAL against a 147.9 average, and the same reading as Tuesday: 28 / 53 / 19 from 440 sessions, directional row 66%, magnet on the lower rail, target 73 points below the box. The hour was built on the same low and added 59.25 points at the top, which was enough to send it to a different cell and the opposite lean.
Price went the other way. The high came 35 minutes after the box closed, on the 10:35 bar: 29,275.25, within 1.5 points of the upper rail at 29,276.75. It turned there, came back to 29,267 at 13:00, 9.75 points under the same rail, and then fell away into the close. The favoured lower rail at 28,984.75 was never closer than 112.5 points.
That near miss deserves stating plainly rather than being filed as a blank. A point and a half further and this window would have been a cross against a lean instead of a neither, and the week would not have the clean sheet of crosses it has. The record grades rails reached, not approaches, and anyone weighing what a 53% share is worth should know how close this one came to being scored the other way.
The 09:30 to 09:55 range, 127.75 points, held for the whole first hour, so this window and the hour share a high, a low, both rails and the same 63.9-point target. What they do not share is the cell. The half hour read 65 / 16 / 19 from 635 sessions at 81% directional; the hour read 47 / 15 / 38 at 76%. Both leaned long from the same box.
One move settled both, or rather failed to. The high, on the 11:30 bar, was 29,499.5 against a rail at 29,538.88, and the afternoon went the other way. Neither target by the 16:55 bar.
Second blank in three days for this cell, and its fourth draw on the record: one paid, one against, two unreached. The table on the chart is a lookup of what comparable windows did between 2009 and 2026. It is not a forecast of this one, and four draws are four draws.
Eight of the ten windows on this page never touched a rail. The two that did were both settled by the same fall on Monday morning, on the smallest pair of boxes the record has produced, in a session that closed at one o'clock.
That gives the quietest week on the record so far, tied with w/c 10 August, which also had eight of ten reach neither rail. It is also the first week with no cross at all, and the thing to say about that is how close it came to not being true: Thursday's half hour leaned short and the day's high stopped 1.5 points under the upper rail. A point and a half the other way and the week has a cross against a lean in it. Nothing about the mechanism changed in those 1.5 points, only the label on the outcome, which is a reason to read a five-day scoreboard for what it is.
Two cells drew twice each this week, with identical shares and bins both times, so the same call was made from two different mornings. 28 / 53 / 19 from 440 on Tuesday's and Thursday's half hours, its seventh and eighth appearances on the record, both unreached. 65 / 16 / 19 from 635 on Wednesday's and Friday's half hours, its third and fourth, both unreached. The outcomes are independent, the readings were not, and across the whole record those two cells now stand at four paid, seven unreached and one against between them.
That pattern held all week. On every one of the five mornings the second half hour extended at most one side of the first half hour's range: the low on Monday, by 1.25 points, and on Tuesday by 30.25; the high on Wednesday by 27 and on Thursday by 59.25; and on Friday neither side, so the two windows shared the box outright. Five mornings is five mornings, and it is recorded here rather than explained.
Across six weeks the running count now stands at 60 windows, of which 36 were in Atlas mode: 15 paid, 19 unreached, 2 against. This week contributed one paid and six unreached, the lowest paid share of any week on the record. Sixty windows cannot confirm or refute frequencies measured over hundreds of sessions, and a quiet week is exactly the sort of result that a null with 19% to 41% on ∅ is entitled to produce.
Five things, from a week in which almost nothing resolved.