Sessions most like 2026-09-15, as of 11:30 ET
What would separate them: Whether the morning low holds on its initial retest, and whether the retrace pushes past its current high-water mark or rolls over from it; whether price reclaims VWAP and whether participation picks up into the afternoon or stays as thin as it has been.
The cohort is mornings that set the session high at or right after the open, drove lower from there, and by late morning had retraced a moderate portion of that drive — roughly in line with the corpus, but with a bounce that had not yet extended into a full reclaim. Within it today's daily range sits at the quiet end and its initial balance at the wider end, while the retrace itself is close to the cohort's typical level. The distinctive placement is that today's retrace is at its maximum right now — the bounce has not yet rolled over — whereas a fair number of the analogs had already bounced further and faded back by the cutoff. Today is also outside the sample on volume: relative volume is far below the lightest analog and a fraction of the corpus norm, and the matcher did not select on it, so the comparison rests on shape rather than participation.
Broke the morning low and pressed well beyond it; most closed well below the cutoff. Some bounced hard toward the morning high before collapsing; others rolled straight over from the retrace.
Pushed through the morning low by a modest margin, found no follow-through, and closed near the cutoff level with the afternoon range contained on either side.
Never revisited the morning low. The retrace kept extending in a slow drift and these sessions closed modestly above the cutoff without reclaiming the morning high.
Did not fit the others: took out the morning low, then reversed through the entire morning range to print a new high and closed above the cutoff.
Written from the computed figures only, by a model that is not permitted to write a number — every count above is rendered from the same data as the table. Camps are one reading of these sessions, not a forecast: the matched features have been measured to add nothing over price position alone, so a cohort rate at this sample size is descriptive.
What kind of session these became — expansion one way, both ways, or balance — on the same four-way split the corpus study uses, where balance is a third of all sessions. Deliberately not a hit rate: which extreme gets taken largely restates where price already sat at the cutoff, and nobody holds to the close anyway. How far it ran, what it cost to sit through, and when it got there are the parts a trade is built on. n=12.
| plot | session | dist | ATR | IB /ATR | RVOL | retr now | retr max | to close | max up | max down | new high | new low |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2026-09-15 — today | 0.00 | 39.2 | 0.82 | 0.15 | 33% | 34% | in progress | |||||
| 2024-06-13 | 0.32 | 38.2 | 0.73 | 0.96 | 32% | 33% | +0.32 | +0.56 | -0.41 | YES | YES | |
| 2024-11-15 | 0.38 | 45.8 | 0.78 | 1.36 | 33% | 35% | -0.32 | +0.04 | -0.76 | — | YES | |
| 2024-12-09 | 0.42 | 41.0 | 0.42 | 0.99 | 38% | 42% | -0.27 | +0.16 | -0.41 | — | YES | |
| 2024-10-15 | 0.44 | 48.2 | 0.35 | 0.96 | 32% | 47% | -0.52 | +0.07 | -0.79 | — | YES | |
| 2024-09-03 | 0.48 | 59.2 | 0.76 | 1.22 | 24% | 32% | -0.71 | +0.11 | -1.16 | — | YES | |
| 2025-03-12 | 0.49 | 101.2 | 0.52 | 1.35 | 38% | 42% | +0.20 | +0.50 | -0.03 | — | — | |
| 2024-01-31 | 0.56 | 36.2 | 0.53 | 1.09 | 29% | 48% | -1.13 | +0.55 | -1.14 | — | YES | |
| 2025-02-21 | 0.60 | 50.8 | 0.74 | 1.14 | 27% | 39% | -1.36 | +0.11 | -1.48 | — | YES | |
| 2025-10-09 | 0.60 | 47.8 | 0.61 | 0.97 | 22% | 53% | +0.09 | +0.18 | -0.34 | — | YES | |
| 2023-05-30 | 0.63 | 38.0 | 0.64 | 1.44 | 33% | 34% | -0.06 | +0.14 | -0.45 | — | YES | |
| 2026-07-07 | 0.65 | 92.2 | 0.53 | 0.91 | 22% | 33% | +0.14 | +0.33 | -0.10 | — | — | |
| 2025-08-29 | 0.66 | 44.0 | 0.97 | 1.13 | 29% | 38% | +0.16 | +0.18 | -0.22 | — | — | |
| cohort low | 36.2 | 0.35 | 0.91 | 22% | 32% | -1.36 | +0.04 | -1.48 | ||||
| cohort median | 47.8 | 0.64 | 1.13 | 32% | 39% | -0.06 | +0.18 | -0.41 | ||||
| cohort high | 101.2 | 0.97 | 1.44 | 38% | 53% | +0.32 | +0.56 | -0.03 | ||||
| cohort hit rate | 1/12 | 9/12 | ||||||||||
| CORPUS median (n=967) | 50.5 | 0.51 | 1.00 | 34% | 60% | +0.05 | +0.30 | -0.28 | ||||
| CORPUS base rate | 56% | 50% | ||||||||||
All ATR-normalised and measured from the 11:30 cutoff. max up / max down are forward excursions after the cutoff — not the session extremes, which in most of these were already set. new high / new low ask whether the post-cutoff move exceeded the pre-cutoff extreme. Click any row to expand that session's chart to full width and jump to it; click again to collapse.
Matched on seven measures, all ATR-normalised and z-scored: distance from VWAP, magnitude of the high off the open, the clock time of that high, expansion of the price-VWAP gap across IB creation, how far VWAP itself was dragged, and how much of the drive has been retraced — both currently and at its deepest. The IB-gap and VWAP terms separate a day that trended from one that chopped to the same place; the retracement terms separate a day still pinned near its high from one that has round-tripped to the open. Ordered nearest first. Solid = up to 11:30, the part matched on. Faded = the rest of the day, the part to read. Shaded band left of the 09:30 marker is premarket from 06:00 ET. The y-axis is a shared ATR ruler with the session open at 0, identical on every tile, so magnitudes are directly comparable.
n=12. This is a case-inspection tool for rare behaviour — sort the tiles into camps by eye. It is not an inference tool, and a percentage taken off this many charts would not mean anything.
Charts rendered with Lightweight Charts™, Copyright © 2026 TradingView, Inc. (Apache-2.0).