Status: pre-registered 2026-09-06 (commit bd9034e), built (151e0e1) and run the same day; verdict: FAIL, 1 of 3 groups (EURUSD passes alone; see research/2026-09-06-entry5m-v3-fixed3r-screen.md). Parents: v1 spec 2026-09-06-entry5m-design.md (FAIL 0/3), v2 prereg 2026-09-06-entry5m-v2-h1stop-prereg.md (FAIL 0/3).
"try 1:3 risk reward" — clarified with Daniel: a fixed take-profit at 3 x risk, paired with the H1-extreme stop of v2.
| Question | Decision |
|---|---|
| Target | target = entry - 3 * risk (short) / entry + 3 * risk (long), where `risk = |
| Stop | v2: one tick beyond the H1 range-2 divergence extreme (Bias.ext), fixed at placement; pending orders cancelled at that level |
| Bias lifecycle | Unchanged from the v1 spec section 4: the bias still ends when price touches the zero-touch extreme, on contradiction, on timeout, or when replaced. Consequence, stated now: when the 3R target lies beyond the zero-touch extreme, the trade never reaches it and is closed at the bias end (bar close, reason bias_end). The run counts these placements as target_beyond_zero_touch (reported, not judged) |
| RR floor | min_rr = 1.0 is automatically satisfied (reward/risk = 3); no trades are rejected on RR except a na or wrong-side stop |
Entry5mConfig.target_mode = "fixed_r" with target_r = 3.0. In PullbackScanner.order_for, after entry, stop and risk are known, target = entry - target_r * risk (short) / entry + target_r * risk (long); the already-broken check and everything else are unchanged. target_mode = "zero_touch" reproduces v1/v2 exactly. The broker exits at the target price with no slippage, as before.
pivot_n = 3, min_pb_bars = 3, timeout_h1 = 96, min_rr = 1.0, pending_max_bars = 48, max_hold_bars = 0, stop_mode = "h1_extreme", target_mode = "fixed_r", target_r = 3.0, ZTD defaults. Same instruments and data as v1/v2.
Criteria: the v1 spec's section 8, verbatim (at least 200 trades; PF > 1.15 and ex-outlier PF > 1.0; expectancy > +0.05 R; best month <= 35% of net R; both calendar halves net-positive; at least two of the three groups; GBPUSD net R not negative). A fail means this variant is dead as specified; no parameter search follows. The sensitivity grid (pivot_n in {2, 3, 5} x min_rr in {1.0, 1.5}; min_rr is not binding under a fixed target) is printed and NOT judged.
Expected side effects, stated now: fewer target exits and more bias-end exits than v2 (many 3R targets sit beyond the zero-touch level), a lower win rate than v2, larger average wins. None of these change the criteria.