Sessions backtests → w/c 14 September 2026

NY AM, week commencing Monday 14 September 2026

The side the indicator called went first on all five sessions, which has happened once before in seven published weeks. Four of the five went to the high; Friday, the only session that opened below London EQ, went to the low. The modal branch landed three times, and six of the ten London levels were taken, which is the joint lowest of the record. Monday is the reason to read the rest of this page rather than the scoreline: it scored on the first sweep and on the branch after falling 146 points the wrong way and stopping ten and a half points short of the level that would have reversed both.

At a glance

Three scores, because the indicator makes three separable claims, and this week they separated.

5 / 5
first sweep called correctly, the second such week of seven
3 / 5
sessions that landed the four-way branch rated highest
6 / 10
levels taken, against a mean stated 64%
0
inside sessions, after the record's first one last week
SessionLondonPrior sweep High onlyLow onlyBothInside First sweep calledActualOutcome
Mon 14 Sep219 ptSwept low 37%35%18%10% ▲ high 53%▲ high 09:45 ✓high only ✓
Tue 15 Sep196 ptSwept low 52%8%38%2% ▲ high 88%▲ high 08:00 ✓both sides
Wed 16 Sep94 ptSwept high 34%25%41%0% ▲ high 58%▲ high 08:00 ✓high only
Thu 17 Sep172 ptSwept high 53%20%24%3% ▲ high 74%▲ high 07:55 ✓high only ✓
Fri 18 Sep144 ptSwept high 10%52%38%0% ▼ low 85%▼ low 07:45 ✓low only ✓

The four share columns are the indicator's own figures, read off the status table at the session open, and they sum to 100 on every row. Gold marks the branch the tables rated highest. The London column is a whole number because the study rounds it, half up. The exact ranges are 218.50, 196.25, 93.50, 172.25 and 143.50 points, so the two opening distances quoted in each session below sum to the exact range and will sit up to half a point away from the figure in this column: Monday's 62.00 above and 156.50 below make 218.50, shown here as 219. London printed above its own average range on Monday and Tuesday and below it on the other three days. No red folder USD release fell inside the 07:30 to 11:30 window on any of the five days. The week's only one was the FOMC decision at 14:00 ET on Wednesday with the press conference at 14:30, two and a half hours after Wednesday's window had closed. Times are the 5-minute bar in which the level was crossed, and every time on this page is read off the 5-minute bar data rather than off the pictures.

None of this week was in the sample. The Atlas Sessions tables were measured on NQ history ending 27 July 2026. Every session below happened afterwards, and the percentages are fixed lookups, so the figures shown are exactly what a live user saw at 07:30.
Every chart here is a 5-minute one, chosen for legibility and nothing else. The levels and the percentages are identical on any timeframe, because both are defined by clock time and price rather than by bars. Only the precision of the timestamp follows the bar size: a take stamped 08:00 here happened somewhere in the five minutes from 08:00.
The number on a level is an exclusive share, not the chance that level gets taken. London Low | 25% means 25% of comparable sessions took the low and left the high alone. The chance the low is taken at all is low only plus both sides, which on Wednesday was 25 + 41 = 66%, and on Wednesday it was not taken, by 38.50 points. The full explanation is on the first week's page.
The week's score, and what it is worth

Five from five on the narrowest of the three claims

The first sweep call was correct on all five sessions. That has happened once before, in the week of 31 August, and across the thirty-five published sessions the call now stands at twenty-six from thirty-five. Four things are worth saying before anyone reads a skill estimate into it.

The first sweep is the smallest claim the indicator makes. It names which of the two London levels trades through first, conditional on one of them being taken at all. It says nothing about whether the other side follows, which is what the four-way branch scores, and nothing about how far price goes, which is what the penetration medians describe. A week can be five from five on it and still be six from ten on levels taken, and this is that week.

Monday's call was worth ten and a half points. The table said 53% to the high, the thinnest call of the five. Price then fell 146.00 points from the open to 28826.00 on the 09:05 bar, which is 10.50 points above the London low, turned there, and took the high at 09:45 instead. Had that bar traded ten and a half points lower, the first sweep would read low, the branch would read low only or both sides rather than high only, and two of the week's three scores would be different. The call was right. It was not comfortable, and the page would be dishonest to print the tick without the distance.

The one below-EQ open called the low, and got it. Monday to Thursday all opened above London EQ and all called the high; Friday opened below EQ, 6.50 points above the London low, called the low at 85% and had it inside four bars. Friday's open and Friday's call are not independent observations: where the open sits relative to EQ is part of the context the lookup reads before it prints anything. It is not a rule either. Monday's own bin, 37 / 35 / 18 / 10 from 215 sessions, has been served to three above-EQ opens and two below-EQ ones, and it called the high on all five. What the week adds is five observations of the outcome agreeing with the call, which is five observations.

Five is five. If each of the five calls landed at exactly the rate the table printed beside it, 53, 88, 58, 74 and 85, then all five landing in the same week has a probability of about 17%, or one week in six. That multiplication also treats the five mornings as independent, which five consecutive sessions on one instrument are not, and dependence between them raises the chance of a clean sweep, so 17% is a floor rather than an estimate. It has now happened twice in seven. A clean week is what a well calibrated table is supposed to produce sometimes, and it is not evidence of anything on its own. The number to watch is the running one: twenty-six from thirty-five, at a mean stated 72% this week, on a record still far too short to separate a good call from a lucky one.

Session by session

The five NY AM sessions

Left, the session opening at 07:30. Right, the same chart at 11:30 with the outcome latched.

Monday 14 September 2026, NY AM opening at 07:30 on NQ: a London range printed as 219 points, 218.5 exactly, that swept its own low, price above London EQ and 62 points under the London high which is rated 37 percent, the London low 156.5 points below at 35 percent, both sides 18 percent, inside 10 percent, first sweep 53 to 47 favouring the high, on a sample of 215 sessions.
07:30 open
London high 29034.00 at 37%, low 28815.50 at 35%. Price opened 62.00 points under the high, above EQ. n 215, the week's only Medium sample.
Monday 14 September 2026, NY AM at the close: a fall of 146 points from the open to a session low of 28826 on the 09:05 bar, 10.5 points above the London low and not latching it, then a turn and a rally that latched the London high taken at 09:45 and carried to a session high of 29089 on the 11:25 bar, the last bar of the window, with the close at 29078.75 above the London high. High only.
11:30 close
High taken 09:45, low missed by 10.50 points. High only, the 37% favourite, and the session high printed on the final bar.
Monday 14 September

A 146-point fall that stopped 10.50 points short of the low, then the high at 09:45

First sweep ✓ · high only, the 37% favourite ✓

The fifth draw of 37 / 35 / 18 / 10 from 215 sessions, the only Medium bin of the week against 479 to 840 on the other four days. London had run 219 points, above its own average, and price opened at 28972.00, 62.00 points under the London high and 156.50 above the London low, above EQ. First sweep 53% to the high, the thinnest call of the week.

Price went the other way first. The session fell 146.00 points from the open to 28826.00 on the 09:05 bar, which is 10.50 points above the London low, and turned there without latching it. The London high was crossed at 09:45 and the session ran on to 29089.00 on the 11:25 bar, the last bar of the window, 55.00 points through the level against a median up penetration of 74. The close at 29078.75 was outside the London range.

First sweep correct at 53% and the modal branch landed, high only at 37%, with the 55% high taken and the 53% low not. It is worth being plain about how close that was: ten and a half points lower and the low goes first, the branch is low only or both sides, and both scores read differently. That is what a 53% call looks like when it scores.

  • London range219 pt, above average
  • Prior sweepSwept low
  • Open vs London EQAbove EQ
  • Four-way37 / 35 / 18 / 10
  • First sweep▲ 53% · ▼ 47%
  • P(high taken)55% → taken 09:45
  • P(low taken)53% → not taken, by 10.50 pt
  • Session high29089.00, 55.00 pt through
  • Session low28826.00, 10.50 pt short
  • Samplen 215, Medium
Tuesday 15 September 2026, NY AM opening at 07:30 on NQ: a London range printed as 196 points, 196.25 exactly, that swept its own low, price above London EQ and 22.25 points under the London high which is rated 52 percent, the London low 174 points below at 8 percent, both sides 38 percent, inside 2 percent, first sweep 88 to 12 favouring the high, on a sample of 766 sessions.
07:30 open
London high 29427.50 at 52%, low 29231.25 at 8%. Price opened 22.25 points under the high. First sweep 88 to 12, the week's strongest call. n 766.
Tuesday 15 September 2026, NY AM at the close: the London high latched taken at 08:00 and a session high of 29484.25 on the 08:55 bar, then a fall of 268.5 points over the next two hours that latched the London low taken at 10:50 and bottomed at 29215.75 on the same bar, then a recovery to 29297.75 by 11:25 and a close at 29291.25, back inside the London range. Both levels taken.
11:30 close
High taken 08:00, low taken 10:50. Both sides, the 38% branch, on a table whose favourite was high only at 52%.
Tuesday 15 September

The high in the seventh bar, then 268.50 points down to the low at 10:50

First sweep ✓ · both sides, the 38% branch

The third draw of 52 / 8 / 38 / 2 from 766 sessions. Price opened at 29405.25, 22.25 points under the London high and 174.00 above the low, with 88% on the high going first, the strongest first-sweep call of the week and one of the strongest the tables produce.

The high went at 08:00, in the seventh bar, and the session topped at 29484.25 on the 08:55 bar, 56.75 points through against a median of 77. From there price fell 268.50 points over the next two hours, latching the London low at 10:50 and bottoming at 29215.75 on the same bar, 15.50 points through against a median down penetration of 83. The last half hour recovered to 29297.75 and the session closed at 29291.25, back inside the range it had been outside on both sides.

First sweep correct at 88% and both marginals taken, the 90% high and the 46% low. The branch was both sides at 38% against a 52% high-only favourite. Three draws of this bin have produced one high only and two both sides, against a table that says 52 and 38, and its 2% inside branch has not appeared.

  • London range196 pt, above average
  • Prior sweepSwept low
  • Open vs London EQAbove EQ
  • Four-way52 / 8 / 38 / 2
  • First sweep▲ 88% · ▼ 12%
  • P(high taken)90% → taken 08:00
  • P(low taken)46% → taken 10:50
  • Session high29484.25, 56.75 pt through
  • Session low29215.75, 15.50 pt through
  • Samplen 766, High
Wednesday 16 September 2026, NY AM opening at 07:30 on NQ: a London range printed as 94 points, 93.5 exactly, the narrowest prior session of the week, that swept its own high, price above London EQ and 39.5 points under the London high which is rated 34 percent, the London low 54 points below at 25 percent, both sides 41 percent, inside 0 percent, first sweep 58 to 42 favouring the high, on a sample of 528 sessions.
07:30 open
London high 29419.00 at 34%, low 29325.50 at 25%, both sides 41%. A 94-point London, the narrowest prior session of the week. n 528.
Wednesday 16 September 2026, NY AM at the close: the London high latched taken at 08:00 with a push to 29439 by 08:05, then a fall back inside the London range to a session low of 29364 on the 08:45 bar, 38.5 points above the London low, then a climb through the rest of the morning to a session high of 29536.25 on the 11:05 bar and a close at 29506.75. High only.
11:30 close
High taken 08:00, low missed by 38.50 points. High only, on a table whose favourite was both sides at 41%.
Wednesday 16 September

Through the high at 08:00, back inside by 08:45, and the session high on the 11:05 bar

First sweep ✓ · high only, against a 41% both-sides favourite

The fifth draw of 34 / 25 / 41 / 0 from 528 sessions. London had run 94 points, the narrowest prior session of the week, and price opened at 29379.50, 39.50 points under the London high and 54.00 above the low, just above EQ. First sweep 58% to the high.

The high went at 08:00, but only barely at first: price reached 29439.00 by 08:05, twenty points through, then fell back inside the London range to a session low of 29364.00 on the 08:45 bar, which is 38.50 points above the London low. That is as close as the 41% both-sides branch came. Price then climbed for the rest of the morning to 29536.25 on the 11:05 bar, 117.25 points through the high against a median of 76, and closed at 29506.75.

First sweep correct at 58%, the 75% high taken and the 66% low not. The branch was high only at 34% against a both-sides favourite of 41%, and it is the first high only this bin has produced in five draws: the other four went both sides three times and low only once. Its 0% inside branch has still not appeared.

  • London range94 pt, below average
  • Prior sweepSwept high
  • Open vs London EQAbove EQ
  • Four-way34 / 25 / 41 / 0
  • First sweep▲ 58% · ▼ 42%
  • P(high taken)75% → taken 08:00
  • P(low taken)66% → not taken, by 38.50 pt
  • Session high29536.25, 117.25 pt through
  • Session low29364.00, 38.50 pt short
  • Samplen 528, High
Thursday 17 September 2026, NY AM opening at 07:30 on NQ: a London range printed as 172 points, 172.25 exactly, that swept its own high, price above London EQ and 35.5 points under the London high which is rated 53 percent, the London low 136.75 points below at 20 percent, both sides 24 percent, inside 3 percent, first sweep 74 to 26 favouring the high, on a sample of 479 sessions.
07:30 open
London high 29613.00 at 53%, low 29440.75 at 20%. The table that drew on the last two Fridays, back for a third time on a Thursday. n 479.
Thursday 17 September 2026, NY AM at the close: a session low of 29555 on the 07:50 bar that stayed 114.25 points above the London low, the London high latched taken at 07:55, a run to a session high of 29753.25 on the 08:30 bar, a pullback of 147.5 points to 29605.75 by 09:35 which is 7.25 points back inside the London range, then a recovery and a close at 29706.5 above the high. High only.
11:30 close
High taken 07:55, 140.25 points through. The low, 114.25 points under the session low, was never approached. High only, the 53% favourite.
Thursday 17 September

The high in the sixth bar, 140.25 points through it, then back inside by 09:35

First sweep ✓ · high only, the 53% favourite ✓

The third draw of 53 / 20 / 24 / 3 from 479 sessions, the table that drew on each of the last two Fridays. Price opened at 29577.50, 35.50 points under the London high and 136.75 above the low, with 74% on the high going first.

The session low came first in time, 29555.00 on the 07:50 bar, and it was still 114.25 points above the London low, so nothing was latched by it. The high went at 07:55, in the sixth bar, and price ran to 29753.25 on the 08:30 bar, 140.25 points through the level against a median up penetration of 79. It then handed 147.50 of that back by 09:35, to 29605.75, which is 7.25 points back inside the London range, before recovering to close at 29706.50.

First sweep correct at 74%, high only at 53% landed, the 77% high taken and the 44% low not. This is the bin's first high only in three draws; the two before it, on 4 and 11 September, both went both sides. On both of those the table's 53% favourite was the miss, and this time it was the hit, on the same four numbers.

  • London range172 pt, below average
  • Prior sweepSwept high
  • Open vs London EQAbove EQ
  • Four-way53 / 20 / 24 / 3
  • First sweep▲ 74% · ▼ 26%
  • P(high taken)77% → taken 07:55
  • P(low taken)44% → not taken, by 114.25 pt
  • Session high29753.25, 140.25 pt through
  • Session low29555.00, 114.25 pt short
  • Samplen 479, High
Friday 18 September 2026, NY AM opening at 07:30 on NQ: a London range printed as 144 points, 143.5 exactly, that swept its own high, price below London EQ and 6.5 points above the London low which is rated 52 percent, the London high 137 points above at 10 percent, both sides 38 percent, inside 0 percent, first sweep 15 to 85 favouring the low, on a sample of 840 sessions.
07:30 open
London high 29967.25 at 10%, low 29823.75 at 52%. Price opened 6.50 points above the low, the only below-EQ open of the week. The record's largest bin, n 840.
Friday 18 September 2026, NY AM at the close: a session high of 29859 on the 07:35 bar, 108.25 points under the London high, then the London low latched taken at 07:45 and a fall to a session low of 29694 by 10:15, then a recovery to 29771.5 by 11:05 and a close at 29749.5, still below the London low. Low only.
11:30 close
Low taken 07:45, then 129.75 points through it. The high was never approached. Low only, the 52% favourite.
Friday 18 September

The week's only open below London EQ, and the low went in the fourth bar

First sweep ✓ · low only, the 52% favourite ✓

The seventh draw of 10 / 52 / 38 / 0 from 840 sessions, the largest bin in the record, and the only session of the week to open below London EQ. Price opened at 29830.25, 6.50 points above the London low and 137.00 under the high, with 85% on the low going first.

The session high printed almost immediately, 29859.00 on the 07:35 bar and 108.25 points under the London high, and that was the whole of the upside. The low went at 07:45, in the fourth bar, and price carried to 29694.00 by 10:15, 129.75 points through it against a median down penetration of 88. The recovery reached 29771.50 by 11:05 and the session closed at 29749.50, still below the London low.

First sweep correct at 85%, low only at 52% landed, the 90% low taken and the 48% high not. Seven draws of this bin have now produced five low onlys, one both sides and one inside, against a stated 52 / 38 / 0, and the 10% high-only branch has not appeared in any of them. Seven draws cannot tell you whether 10 is the right number for it.

  • London range144 pt, below average
  • Prior sweepSwept high
  • Open vs London EQBelow EQ
  • Four-way10 / 52 / 38 / 0
  • First sweep▲ 15% · ▼ 85%
  • P(high taken)48% → not taken, by 108.25 pt
  • P(low taken)90% → taken 07:45
  • Session high29859.00, 108.25 pt short
  • Session low29694.00, 129.75 pt through
  • Samplen 840, High
The most useful thing in the record

The same forecast, more than once

Across the thirty-five sessions published so far, eight context bins have each come up more than once, and thirty-four of the thirty-five sessions sit in one of them. The only session still on its own is Monday 10 August. Same context bin, same lookup, same numbers to the percentage point, on days weeks apart with different London ranges and different price levels. Every one of this week's five forecasts was a repeat, as every one of last week's was.

ForecastSessionsFirst sweepOutcome
10 / 52 / 38 / 0 · sweep 15 / 85 · n 840Mon 3 Aug▼ low 85%both sides
Thu 20 Aug▼ low 85%low only
Mon 31 Aug▼ low 85%low only
Tue 1 Sep▼ low 85%low only
Mon 7 Sepnone takeninside
Thu 10 Sep▼ low 85%low only
Fri 18 Sep▼ low 85%low only
38 / 11 / 50 / 1 · sweep 80 / 20 · n 524Fri 7 Aug▲ high 80%high only
Wed 12 Aug▲ high 80%high only
Fri 14 Aug▲ high 80%both sides
Wed 26 Aug▲ high 80%both sides
Fri 28 Aug▲ high 80%both sides
34 / 25 / 41 / 0 · sweep 58 / 42 · n 528Wed 5 Aug▲ high 58%both sides
Thu 6 Aug▲ high 58%both sides
Tue 11 Aug▲ high 58%low only
Mon 24 Aug▲ high 58%both sides
Wed 16 Sep▲ high 58%high only
24 / 45 / 25 / 6 · sweep 38 / 62 · n 406Tue 18 Aug▼ low 62%low only
Wed 19 Aug▼ low 62%both sides
Thu 3 Sep▼ low 62%both sides
Wed 9 Sep▼ low 62%low only
37 / 35 / 18 / 10 · sweep 53 / 47 · n 215Tue 25 Aug▲ high 53%low only
Thu 27 Aug▲ high 53%low only
Wed 2 Sep▲ high 53%high only
Tue 8 Sep▲ high 53%both sides
Mon 14 Sep▲ high 53%high only
52 / 8 / 38 / 2 · sweep 88 / 12 · n 766Tue 4 Aug▲ high 88%high only
Fri 21 Aug▲ high 88%both sides
Tue 15 Sep▲ high 88%both sides
53 / 20 / 24 / 3 · sweep 74 / 26 · n 479Fri 4 Sep▲ high 74%both sides
Fri 11 Sep▲ high 74%both sides
Thu 17 Sep▲ high 74%high only
28 / 15 / 56 / 1 · sweep 58 / 42 · n 173Thu 13 Aug▲ high 58%both sides
Mon 17 Aug▲ high 58%low only

Two things follow, and they matter more than this week's score.

The forecast is a lookup, not a fit. Nothing about Thursday was used to produce Thursday's numbers. They are the numbers Friday 4 September and Friday 11 September got, because all three sessions landed in the same measured bin, on London ranges of 125, 125 and 172 points at price levels three hundred points apart. Two of the three went both sides and the third went high only.

Identical forecasts resolving differently is the system working, not failing. Every one of the eight repeated bins has now produced more than one outcome. The largest, 10 / 52 / 38 / 0 from 840 sessions, has drawn seven times: five low onlys, one both sides and one inside, against a stated 52 / 38 / 0, and its 10% high-only branch has appeared in none of them. Two bins produced an outcome this week they had not produced before, Wednesday's first high only in five draws and Thursday's first in three. Seven draws cannot tell you whether 52 is the right number, and neither can thirty-five sessions; what they can show is whether the table keeps returning the same answer for the same context bin, which it does.

What this week actually shows

Five things, from a week that scored well on its narrowest claim and poorly on its broadest.

  1. Five first sweeps from five, and the record moves to twenty-six from thirty-five. The mean stated confidence across the five calls was 72%, ranging from Monday's 53% to Tuesday's 88%. One clean week is one clean week: the seven published weeks now read four, three, three, two, five, four and five from five.
  2. Monday scored on both claims and came within 10.50 points of scoring on neither. Price fell 146.00 points from the open before it did anything else and stopped ten and a half points above the London low, then turned and took the high at 09:45. The scoreline cannot express that, which is why the session pages exist.
  3. The high was taken on four of the five days, the low on two. Three sessions finished high only, one both sides and one low only. Of the four levels not taken, one was a 66% low on Wednesday and one a 53% low on Monday; the two lows that did go were rated 46% and 90%. The per-side view is the one with the most events behind it and the least drama in it.
  4. Six of ten levels taken, against a mean stated 64%. A realised 60%, which is the joint lowest of the seven weeks with last week. Across the record forty-nine of seventy levels have been taken against a mean stated 67%, a realised 70%. That gap is the thing to keep watching, and seventy level events cannot close it in either direction.
  5. Thirty-five sessions settle nothing, and the indicator's own row is still behind. The realised versus expected row on the status table read 51 to 52% against 56% expected across the week, on 262 to 277 setups loaded, which is behind for the seventh week running. The record has a shape, not a verdict.
How to check this yourself. Load Atlas Sessions on your own NQ chart, set it to any timeframe you like, and scroll back to the week of 14 September 2026. The levels and percentages will be the ones printed above, because they are fixed table lookups rather than anything computed from what happened next. Put Monday next to Wednesday 2 September while you are there: same four numbers, same sample of 215, same outcome, twelve days apart.