Each week we publish every NY AM session Atlas Sessions read, captured twice: once at the open, when the tables state what they expect, and once at the close, when the outcome is settled. Right calls, wrong calls, and the sessions that split the difference.
One row per week, newest first. Three scores, because the indicator makes three separable claims about a session. Click a week to read it in full.
| Week | Sessions | First sweep | Modal branch | Levels taken | Stated P(take) | Realised | Inside sessions |
|---|---|---|---|---|---|---|---|
| w/c 14 September 2026 | 5 | 5 / 5 | 3 / 5 | 6 / 10 | 64% | 60% | 0 |
| w/c 7 September 2026 | 5 | 4 / 5 | 2 / 5 | 6 / 10 | 62% | 60% | 1 |
| w/c 31 August 2026 | 5 | 5 / 5 | 3 / 5 | 7 / 10 | 62% | 70% | 0 |
| w/c 24 August 2026 | 5 | 2 / 5 | 3 / 5 | 8 / 10 | 66% | 80% | 0 |
| w/c 17 August 2026 | 5 | 3 / 5 | 2 / 5 | 7 / 10 | 67% | 70% | 0 |
| w/c 10 August 2026 | 5 | 3 / 5 | 2 / 5 | 7 / 10 | 74% | 70% | 0 |
| w/c 3 August 2026 | 5 | 4 / 5 | 3 / 5 | 8 / 10 | 70% | 80% | 0 |
| Since 3 August 2026 | 35 | 26 / 35 | 18 / 35 | 49 / 70 | 67% | 70% | 1 |
This is a running tally over a small number of sessions, published as it accumulates. It is not a performance record and not a return: no entries, exits, position sizing or costs are modelled anywhere on these pages. The place to judge calibration remains the indicator's own realised versus expected row, running on your own chart over your own history.
Atlas Sessions reads all four sessions. These reviews fix on one, and the reason is measured rather than assumed.
| Session | Clock (ET) | Share of day's minutes | Share of day's range | Range concentration | Volume density |
|---|---|---|---|---|---|
| Asia | 18:00 to 02:30 | 37.0% | 37.1% | 1.00× | 0.17× |
| London | 02:30 to 07:30 | 21.7% | 36.5% | 1.68× | 0.33× |
| NY AM | 07:30 to 11:30 | 17.4% | 66.2% | 3.80× | 2.45× |
| NY PM | 11:30 to 17:00 | 23.9% | 61.1% | 2.56× | 1.94× |
Measured over 15,214 session instances of NQ 1-minute data from 2009 to 2026, with roll sessions excluded. Range concentration is a session's share of the day's range divided by its share of the day's minutes, so 1.00 means a session covers exactly its proportional amount of ground and nothing more. Asia landing on 1.00 is a useful check that the method is not manufacturing the result. We use ratios rather than points per minute deliberately: raw points cannot be compared across a period in which NQ went from 1,200 to 30,000, whereas the concentration figure is stable to two decimal places between the full history and the last three years alone. Range shares do not sum to 100 because each session traverses an overlapping slice of the same daily envelope.
NY PM is the busier session, not NY AM. It carries more total volume across the day, 46% against 42%. NY AM is the denser one, which is the fair comparison given NY PM runs an hour and a half longer. We would rather state both numbers than pick the flattering one.
NY AM takes at least one London level in about 49 sessions out of 50, and that is mostly arithmetic rather than skill. Take rates track the size of a session relative to the one before it almost exactly: the two sessions that are 0.94 times their predecessor both take a level about 84% of the time, and NY AM, at 1.50 times London, takes one 98% of the time. So the high base rate is what a session that size should produce. What it means in practice is that at NY AM the question is almost never whether a level goes, but which one, and when. That is what these reviews score.
Every chart on these pages is the indicator as it renders live, with nothing added and nothing removed.
Newest first.
The called side went first on every session, four of them to the high, and Friday, the only morning that opened below London EQ, went to the low. Monday scored on both the sweep and the branch after falling 146 points the wrong way and stopping 10.50 points above the London low. The modal branch landed three times and six of ten levels were taken, the joint lowest of the record.
Read the week →Monday 7 September finished between both London levels without touching either, the first time in thirty published sessions, on a table that rated the inside branch at 0%. The low stopped 2.00 points above the London low and the high 37.75 points short. With no side taken there is no first sweep to call, so the scorer records a miss by construction; on the four days that did sweep, the called side went first every time. Four from five on the first sweep, two from five on the branch, six of ten levels taken against a mean stated 62%.
Read the week →The first clean week on the first-sweep axis, at a mean stated 72%, though on three of the five mornings the called level was within 35 points of the open. Monday and Tuesday drew 10 / 52 / 38 / 0 from a 105-point London and a 379-point one and both took the low inside the first half hour. Last week's new table came back on Wednesday and went high only after two low-only draws, so every repeated bin has now split. Friday's payrolls morning took the high at 08:25 and the low on the 08:30 bar, five minutes apart.
Read the week →Two first sweeps from five, the weakest week on that axis so far, with an 80% call on the high missing on Wednesday's 08:30 Core PCE bar. Three modal branches landed, all of them the both-sides favourite, which takes it to seven from twelve across the record. Tuesday and Thursday drew a table the record had not seen, 37 / 35 / 18 / 10 on 215 sessions, and both took the low against the 37% favourite, missing the high by 4 and 27 points. Eight of ten levels resolved as their marginals implied.
Read the week →All five of this week's tables had appeared before, three from earlier weeks and one pair within the week, and four of the four repeated bins resolved differently from their earlier draw. Three first sweeps from five, two modal branches, seven of ten levels taken, and the both-sides favourite appearing once and losing, which leaves it four from nine.
Read the week →Both sides was the highest-rated branch on all five mornings and was paid on two of them. Tuesday landed the smallest share on its table. Wednesday and Friday returned identical forecasts, to the percentage point, and resolved two different ways, which is the clearest demonstration yet that these are lookups rather than anything fitted after the fact.
Read the week →Ten level events across five NY AM sessions, eight of them taken. Wednesday and Thursday returned identical forecasts from different days and both landed the same branch, in opposite orders. The one miss came on a call tied for the week's least confident, and payrolls Friday took London's high fifty-five minutes before the release.
Read the week →Published at the end of the trading week.