feat(EURUSD): deploy DailyBias config — cross-year validation then set file update

[cfournel]

Summary

EURUSD 20-test improvement sweep + 10-test combo sweep completed.

Winner: DailyBias alone — nothing else improves upon it.

Combo sweep results (2022–2026, full period)

Config	PF	Trades	DD%	vs Baseline
DailyBias alone	1.21	35	2.30	+0.35
DailyBias+Skip09	1.21	35	2.30	0.00
DailyBias+NoMonFri	1.09	19	1.56	-0.12
DailyBias+OB40	1.00	32	1.71	-0.21
OB40+Skip09	1.13	48	4.13	+0.27
Baseline (no DailyBias)	0.86	40	5.90	—

Key insights:

• DailyBias dramatically reduces DD (5.90% → 2.30%) while improving PF
• OB40 hurts when combined with DailyBias (cancels too many OBs)
• NoMonFri kills trade volume (35→19) — too aggressive with DailyBias
• Skip09 is redundant with DailyBias (DailyBias already avoids those entries)

Remaining tasks

• Cross-year validate DailyBias on EURUSD (5 windows: 2022/2023/2024/2025/Full)
• If 3+/5 years profitable + PF ≥ 1.1 overall → update OBInclude/SetFiles/eurusd.set
• Update MQL5/Profiles/Charts/OB_Deploy/chart05.chr with EURUSD chart (magic 1234581)
• Add EURUSD to crossyear_winners.csv and portfolio analysis

Cross-year validation script

python3 scripts/eurusd_crossyear.py
# Already queued in run_all_sweeps.sh Phase 4

Success criteria

PF ≥ 1.1 full period, 3+/5 years profitable (PF ≥ 1.0)

[cfournel]

Phase A Cross-Year Validation — FAIL ✗

Year	PF	Trades	Pass?
2022	5.58	8	✓
2023	0.68	14	✗
2024	1.29	6	✓
2025	0.58	7	✗
Full	1.21	35	✓

Result: 2/4 profitable years — FAIL (need 3+/4)

The DailyBias filter reduces EURUSD to 6-14 trades/year (too low for stable results). 2023 and 2025 are loss years.

Full-period PF=1.21 looks decent but cross-year instability makes this config unreliable for live trading.

Decision: Do NOT deploy EURUSD with DailyBias filter.

[cfournel]

Closing: cross-year validation failed (2/4 profitable years). EURUSD not deployable with current config.