LON MOMENTUM BULL
Fresh, tight London sweep of Asia high — enter long near session VWAP, target +60pt. PM line on our chart HUD.
updated 2026-04-23The setup
The first 02:00–02:59 ET London bar sweeps the Asia high — but only if the sweep is fresh, the range-so-far is tight, and the close is still at or below session VWAP. That combination of conditions isolates a real London-session momentum pattern that continues rather than reversing.
It’s the live-safe residue of a much bigger (but oracle-driven) edge we found while investigating AMD Power-of-3 Early Pull. See the paired writeup: AMD Early Pull — walk-away.
Entry rules
All of these are checkable at the sweep bar — nothing peeks ahead:
- Window: 02:00–02:59 ET (London killzone, hour == 02)
- Sweep side: Asia high (not low — the short side broke OOS 2/3 and was dropped)
- Time to sweep: ≤ 15 minutes from 02:00
- Range-so-far: < 20 NQ points (tight pre-sweep range = fresh momentum)
- Entry price: ≤ session VWAP (don’t chase extended; mean-reversion gotcha below)
- Only-one-side: Asia low has not also been breached in the same window (handled by the kill-switch, see below)
Target & stops
- Target: entry + 60 points
- Probability: 62 on the HUD
- TTL: 180 minutes
- Direction: long only
Backtest
| Cohort | Sample | P&L | WR |
|---|---|---|---|
| Best live-safe composite | n = 150 (5yr) | +$3,407 | 53.3% |
| Out-of-sample subset | 3/3 years positive | +$4,011 | — |
All of the profit is in out-of-sample. The pattern is improving over time, not decaying. ~30 signals per year — thin enough that we ship it as a PM probability line, not a standalone NT8 strategy.
The VWAP mean-reversion twist
The feature that matters most is the one that’s least intuitive. In the 5-year dataset, entries that print 15+ points past session VWAP in the trade direction lost money in 2 of 3 OOS years. Entries at or below VWAP for longs printed +$3,890 OOS across all 3.
So the rule isn’t “London breaks Asia high, chase it.” It’s “London reclaims Asia high cheaply, ride it.” Completely opposite stance.
Kill-switch interaction
If both Asia high and Asia low get swept during 02:00–05:00 ET, the day is a chop day. Historically: 233 such days, 7.3% WR, −$32,620 total.
When lon_both_swept = True fires, a −10 probability penalty is applied to all London-window directional lines — including this one, PRELON IB, and AMD. Lines that drop below prob 40 are removed entirely. This runs in PredictionModel.cs right after the AMD confluence block.
Why it’s Tier 2, not Tier 1
- Sample size is thin (~30/yr). Variance matters.
- The original oracle ceiling (+$17,310 OOS with peek-ahead) is unreachable live.
- We don’t want this line crowding out higher-confidence Tier 1 signals when it fires.
History
- 2026-04-23 — Shipped. Ported from
backtest_lon_momentum_livesafe.pycomposite. Paired with thelon_both_sweptkill-switch.