feat(eurjpy): raise minimum stars threshold to 3

[cfournel]

Finding (EURJPY Forensics 2022-2026)

SHAP analysis identified stars as a top loss driver. The best single filter found by Module F:

require stars >= 3 → removes 12% of losses, costs only 6% of wins

Current EURJPY config has no minimum stars gate (all qualified OBs are taken).

Recommendation

Test inpMinStars=3 via cross-year sweep on EURJPY:

• Baseline: PF=2.10, T=32
• Expected: fewer trades, higher win rate, similar or better PF

Implementation

inpMinStars input already exists in inputs.mqh.
Add to EURJPY sweep: inpMinStars=3 override vs baseline inpMinStars=0.

Related

• Linked to EURJPY forensics findings (2026-03-28)
• H4 trend alignment: losses trend toward bearish H4 at entry — may correlate with low-star setups

[cfournel]

Tested inpMinStars=0/2/3 cross-year on EURJPY (2022-2026). Finding: MinStars=3 Full period = T=32 PF=2.08 vs baseline T=32 PF=2.10 — identical trade count. All 32 EURJPY trades already carry ≥3 stars, so the filter is a no-op for this symbol. The inpMinStars input was still added to the EA (useful for future symbol testing where low-star trades actually occur). No deployment needed for EURJPY.