NQ · model #10

OB RANGE (BULL / BEAR / MID / STRETCH)

NY opening-range box (09:30–10:30 ET). Where price sits relative to the box midpoint at 10:30 projects the rest of the session. OB BULL/BEAR carry real +18–21pt close-direction edge.

updated 2026-04-21

tier 1 both opening-range opening-rangeconfluenceclose-directionmagnet

The setup

The first hour of RTH (09:30–10:30 ET) forms an opening-range box. At the 10:30 close we check where price sits relative to the box midpoint and fire a family of lines on our chart HUD:

OB BULL / OB BEAR — direction-setting line toward the box extreme
OB MID MAGNET — pullback target to the midpoint
OB STRETCH B — ±25pt extension target past the box extreme

Entry rules

Trigger: 10:30 ET close
Box: high / low / midpoint of 09:30–10:30
ABOVE MID (close > midpoint): fire OB BULL long + OB STRETCH B up target
BELOW MID (close < midpoint): fire OB BEAR short + OB STRETCH B down target
Magnet: OB MID MAGNET always fires toward box midpoint
TTL: through RTH close

Probabilities

Line	Prob	Nature	Tier
OB BULL	83%	close-direction	1
OB BEAR	82%	close-direction	1
OB MID MAGNET	70%	touch	1
OB STRETCH B (±25)	65%	touch	2

Backtest

audit_ob_range_close.py, n = 1,360 RTH sessions:

Cohort	n	Touch	Close-direction	Edge vs baseline
ABOVE MID → OB BULL	731	81.4%	71.4% BULL	+17.8pt
BELOW MID → OB BEAR	629	80.3%	67.4% BEAR	+21.0pt
ABOVE MID → STRETCH up	731	64.7%	—	matches claim
BELOW MID → STRETCH dn	629	62.6%	—	matches claim

Baseline: BULL 53.6% / BEAR 46.4%.

The close-direction dividend

Most magnet signals we’ve audited carry zero close-direction edge — the magnet just predicts a touch. OB BULL/BEAR is the exception. The 10:30 positioning rule that generates the signal happens to also filter for days with genuine directional bias: a market that’s above its OR midpoint at 10:30 closes bull 71% of the time, not just touches the OR high 81%.

This is why the OB family earned Tier 1 while AMD, LON SWEEP, and HRR got demoted to Tier 2 magnets when audited against close-direction rather than touch.

The recalibration

OB MID MAGNET used to fire at 92%. That was the unconditional “did price ever touch the midpoint during RTH” rate — not conditional on our ABOVE/BELOW positioning rule. Under proper conditioning the rate is 68.9% (from above) / 70.1% (from below). Probability pulled back to 70.

The OB MID kill rule

Separately from the OB MID MAGNET touch line, there’s a regime kill: when 10:30 closes inside the OB midpoint zone (neither clearly above nor below), every other directional model on the HUD gets a −15 probability penalty. The data:

PO XTND BULL + OB MID: 49% → 24% (−25)
PO RNG BEAR + OB MID: 72% → 58% (−14)
PDH TOUCH + OB MID: 48% → 34% (−14)
MN OPEN FILL + OB MID: 67% → 54% (−13)

OB MID days are chop days for directional thesis. The kill rule keeps the HUD from over-confident calls on them.

Confluence amplifier

The flip side: when OB BULL or OB BEAR agrees with an earlier 09:30 directional signal, hit rates jump +19 to +27pt. Examples from the 10:30 audit:

Earlier signal + OB same-dir	Lift
PO XTND BEAR + OB BEAR	44% → 71% (+27)
MON HIGH + OB BULL	47% → 73% (+26)
PO XTND BULL + OB BULL	49% → 73% (+24)
GAP FILL→PDH + OB BEAR	66% → 88% (+22)
DOW Friday LOW + OB BEAR	75% → 97% (+22)

PredictionModel.cs applies a +15 probability boost to non-OB lines that share direction with an active OB confirmer. A few models are exempt from the boost (VWAP REGIME, PRELON SWEEP, SB CONT) because their own edge already subsumes the OB signal and double-counting would inflate.

OB midpoint reversion — the 10:30–11:30 trade on range days — full backtest of the MidpointReversion.cs strategy layered on this box. Produces the V3 ship config: MidLong + FlipShort only, MN_INACTIVE only, cutoff 11:30, OB-extreme stop, TOWARD-VWAP gate. +19.4 pts/trade across ~154 trades/year.
The OB MID and OB MID MAGNET lines now carry MN HOLD confluence adjustments (−5 aligned, −20/−18 opposed) — mean-reversion at the OR midpoint dies on trend days regardless of direction agreement.

History

2026-04-20 — Original ship with 92% OB MID MAGNET.
2026-04-21 — Audit round 2. OB BULL/BEAR validated (+18–21pt real close-direction edge, surprise). OB MID MAGNET recalibrated 92→70. OB STRETCH B validated. OB MID kill rule and +15 confluence boost quantified.
2026-04-24 — OB MID / OB MID MAGNET gained MN HOLD confluence penalties ({−5, −20, 0} and {−5, −18, 0}) after the 1,370-session midpoint-reversion replay. See the OB midpoint reversion research for the underlying data and the MidpointReversion.cs V3 upgrade.