France vs. Morocco — Consensus Ensemble (QF, World Cup 2026)

Thu July 9, 2026 · 22:00 CEST (20:00 UTC / 16:00 EDT) · Gillette Stadium, Foxborough MA (OPEN-AIR)


TL;DR

🇫🇷 France advance ~60% (FRA win 90' ~47% / Draw ~25% / MAR ~28%). Modal 1-1 or FRA 2-1. Mbappé #1 scorer (30-36%), Ounahi #1 MAR (15-28%). FRA's elite defense (3 GA in 5 games) vs MAR's false-9 pivot (no specialist finisher) is the defining matchup.

(2-agent consensus — g4 [glm-5-turbo] failed 3× to rate-limit and could not complete. mm3 + g5 agree tightly: FRA advance 59-60%, FRA win 46-48%, draw 24-25%. Narrow 1pt spread = both anchored on the same WDM + the same MAR regression thesis. g4's scoreline-precision strength is absent — modal scoreline split is less robust than a full trio.)


Ensemble Reconciliation

Agent Model 90' FRA 90' Draw 90' MAR Advance FRA Modal 90'
mm3 minimax/MiniMax-M3 46% 25% 29% 59% 1-1 (16%)
g5 zai/glm-5.2 (xhigh) 48% 24% 28% 60% FRA 2-1 (13%)
g4 glm-5-turbo (xhigh) FAILED (rate-limit)
CONSENSUS ~47% ~25% ~28% ~60% 1-1 / FRA 2-1

⚠️ 2-agent ensemble. g4 (glm-5-turbo, xhigh) failed 3× to API rate-limit and could not produce output. The consensus is mm3 + g5 only. Agent-strength caveat: g4's validated strength is scoreline precision on clear favorites (FRA fav_share 0.66 qualifies) — its absence makes the modal scoreline less robust. The 1pt spread between mm3/g5 is tight; flag as potential shared-anchoring (both keyed on MAR's R16 xG regression), not necessarily high confidence.

Key factors (both agents agree):


Consensus reconciled by lead (2-agent ensemble — g4 rate-limited). Spread 1pt (tight — flag shared anchoring on MAR xG regression). Modal: 1-1 / FRA 2-1. Watch: 60' sub window + MAR bench cavalry + Bounou shootout pedigree if level.