CONTRA BULL / CONF BEAR
Three-way vote across 1h EMA200, 0809 bias, and London sweep. One dissenter (if 0809 or LON) → CONTRA BULL 58–63%. All-bear aligned → CONF BEAR 57%.
updated 2026-04-21 tier 2 both confluence-vote confluenceemavoteclose-direction
The setup
Three independent directional filters vote at the RTH open:
- 1h EMA200 — HTF regime (price above/below)
- 0809 zone bias — 08:00–09:00 ET bias direction
- LON sweep — which Asia extreme London swept
Two regimes fire lines on our chart HUD:
CONTRA BULL— 2 of 3 vote BEAR, 1 dissents BULL, and the dissenter is 0809 or LON (not HTF). The 1-bull-vote cohort closes BULL 56–63% despite the bear majority — a clean contrarian edge on specific dissenters.CONF BEAR— all 3 vote BEAR. Alignment edge, 57% close BEAR.
Entry rules
- Window: 09:30+ ET
- CONTRA BULL: exactly 1 bull vote, dissenter is 0809 or LON
- Base prob: 63 if 0809 dissents / 58 if LON dissents
- +5 if open gap > +20pt
- +3 if Asia range > 70pt
- Target: RTH open + 25pt
- CONF BEAR: all 3 bear
- Base prob: 57
- +4 if open gap < −20pt
- Target: RTH open − 30pt
- TTL: through RTH close (390 min)
Backtest
backtest_contrarian_1vote.py, 1-bull-vote cohort profiled with 1h EMA200 as HTF filter:
| Dissenter | n | Close BULL % | Avg close pts |
|---|---|---|---|
| 0809 | 96 | 63.5% | +60 |
| LON | 116 | 58.6% | +29 |
| HTF | — | 52.3% | noise — filtered out |
All-BEAR cohort (3 votes): 57% BEAR, −51pt avg.
All-BULL and 2-bull cohorts: flat. No edge; don’t fire.
The dead-end research that led here
Before landing on this setup we falsified a long list of confluence hypotheses:
- EMA21 on 5m/15m — no edge (triple-align 50–52%)
- Daily EMA21, Daily EMA50, 1h EMA50 — also no edge
- Triple-alignment as a compound bull/bear signal — degrades to baseline
- Daily EMA21 vs 1h EMA200 — 85.8% collinear. No independent HTF regime signal exists once you have the 1h.
The only edge that survived was the 1-vote contrarian pattern — and only when the dissenter is 0809 or LON, never HTF.
Why Tier 2
- Smaller edge (+5 to +10pt close-direction) than Tier 1 flagships.
- Smaller sample than VWAP REGIME / SB CONT.
- Still directionally meaningful and independent of most other lines.
- Exempt from OB confluence boost and regime skip — already a confluence product.
History
- 2026-04-21 — Shipped as Model 19. Five falsification scripts preceded it; the surviving edge is the 1-vote contrarian with 0809/LON dissenter.