NQ · model #18

LON BOTH BULL

London sweeps Asia low first, then reclaims and sweeps Asia high by 9:30 — a bull takeover pattern. Tight-Asia subset hits 64% UP close; with inside/below gap, 70%.

updated 2026-04-21
tier 2 long bull-takeover londonasia-sweeptakeoverbull-regime

The setup

London session sweeps both Asia extremes — but in a specific order:

  1. Asia low is taken out first (early manipulation to the downside)
  2. Price reverses, reclaims the Asia range, and then sweeps Asia high

This L→H ordering is the bull-takeover flag. The mirror pattern (H→L) is a bear takeover but shows no close-direction edge — we don’t trade it.

Entry rules

Probability table:

Asia widthGap class (9:30 open vs Asia)Probability
tight (<40pt)inside or below70
tightnormal65
normalinside or below65
normalnormal58
wide / xwideanyskip

Backtest

n = 1,364 MNQ sessions. Distribution of London sweep types:

London sweep typen%NY close UPNY close DN
ONLY_H58242.7%52.7%47.3%
ONLY_L46634.2%53.0%47.0%
BOTH H→L (bear takeover)1188.7%53.4%46.6%
BOTH L→H (bull takeover)1158.4%58.3%41.7%
Neither836.1%54.2%45.8%

Both-swept ≈ 17.1% of all sessions (consistent with Herman’s 18% observation). But only L→H prints a tradeable close-direction edge.

Sub-buckets that tighten the edge

FilternUP rate
+ tight Asia (<40pt)6564.6%
+ INSIDE Asia gap2070.0%
+ BELOW Asia gap3268.8%

Excursion signature

The directionality shows up in the intraday excursion, not just the close:

So the bear version actually has more intraday follow-through — but it doesn’t translate into close-direction edge. Only the bull version does. Another case of “touch ≠ close.”

Kill-switch interaction

Because L→H satisfies the “both Asia sides swept” condition, lon_both_swept=True also fires. The kill-switch would normally penalize this signal −10 prob. However, the L→H subset is the one carve-out where the both-swept pattern is productive — so the Python detection emits lon_both_active separately from the lon_both_swept chop flag, and the PredictionModel line for LON BOTH BULL is added from its own block, not from the chop-day branch.

Why Tier 2

The sample is thin (~19 fires/year) and the edge is modest (+5 to +12pt above baseline). Worth adding to the board, not worth sizing up on alone.

History