Status: pre-registered 2026-09-06 (commit ea4e064), built (24aa2de) and run the same day; verdict: FAIL on XAUUSD (see research/2026-09-06-entry5m-v5-xau-scaleout-screen.md); the confirmation run was not triggered. Parent: v4 2026-09-06-entry5m-v4-xau-arm-prereg.md (FAIL on XAUUSD: 726 trades, net −68.3 R, PF 0.83).
"What if we take half profit at 1x profit and the rest at 3x." Decided with Daniel: on top of v4 (XAUUSD only, arm-triggered bias, running H1 range stop, 3R target), and the stop on the remaining half moves to breakeven once the first half is out.
| Question | Decision |
|---|---|
| Base rule | v4 exactly: bias_source = "arm", arm_tier = "Extreme", stop_mode = "h1_extreme", target_mode = "fixed_r", target_r = 3.0, XAUUSD only |
| Scale-out | Half the position (scale_out_frac = 0.5) exits at entry ± 1.0 x risk (scale_out_r = 1.0) as a limit, no slippage; the remaining half keeps the 3R target |
| Stop after the partial | Moves to breakeven = the actual fill price, from the bar after the partial (the partial bar itself keeps the original stop, which was checked first on that bar). A breakeven exit is a stop exit with slippage, reason breakeven |
| Everything else | v4: bias ends at the H1 oscillator range end (remaining position closed at that bar's close), pending-order rules, costs 18 ticks per side |
Per bar for an open position, in order: (1) stop at the effective stop (original, or breakeven after the partial); (2) if not yet scaled and the bar touches the 1R level, the first half exits there and the effective stop moves to the fill price; (3) target on the remaining size; (4) time stop. On the fill bar the same sequence follows the fill. A bar that hits both the stop and the 1R level is a stop (full loss), as with targets. Trade R is size-weighted: 0.5 x 1.0 + 0.5 x (exit − fill) x side / risk when scaled, else as before. MFE/MAE are unchanged (full-position excursions).
Config: scale_out_r = 1.0, scale_out_frac = 0.5, breakeven_after_scale = True; scale_out_r = 0 reproduces v1-v4. New counter partial_exits.
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, bias_source = "arm", arm_tier = "Extreme", scale_out_r = 1.0, scale_out_frac = 0.5, breakeven_after_scale = True, XAUUSD.
Single-instrument criteria as in v4 (all five section-8 sub-criteria on XAUUSD; the per-symbol verdict column is the judged line). A pass is provisional and triggers the pre-declared confirmation on EURUSD, GBPUSD, EURJPY under the group rule. A fail means the variant is dead as specified; no parameter search follows. Sensitivity grid printed, NOT judged.
Expected side effects, stated now: the distribution of trade R compresses (worst case after the partial is about +0.5 R, best case +2.0 R); more trades end at breakeven than at the stop; win rate (R > 0) rises. The criteria are unchanged.