Colombia vs. Switzerland β€” Consensus Ensemble (R16, World Cup 2026)

Tue July 7, 2026 Β· 22:00 CEST (20:00 UTC / 13:00 PDT) Β· BC Place, Vancouver (roof CLOSED)


TL;DR

πŸͺ™ GENUINE COIN-FLIP R16 β€” COL advance ~50% (3-agent consensus, razor-thin). 90' leans COL 35% with 29% draw band; modal 1-1. The €250M vs €282M squad gap is negligible. COL has James/DΓ­az creativity + discipline edge; SUI has Embolo counter + the tournament's best bench (13.3% sub-impact).

(Agent-strength weighting: coin-flip β†’ mm3's draw-band reasoning weighted higher. g5 uniquely leans SUI (53%) β€” its model-breaker detection flagged SUI's bench as the deciding edge. The 6pt spread (47-53%) is honest disagreement on whether bench or starters win tight KOs.)


Ensemble Reconciliation

Agent Model 90' COL 90' Draw 90' SUI Advance COL Modal 90'
mm3 minimax/MiniMax-M3 37% 27% 36% 51% COL 1-0 (17%)
g4 glm-5-turbo (xhigh) 37% 28% 35% 53% 1-1 (13%)
g5 zai/glm-5.2 (xhigh) 33% 31% 36% 47% 1-1 (13%)
CONSENSUS β€” ~35% ~29% ~36% ~50% 1-1 / COL 1-0

Agent-strength weighting applied. Coin-flip β†’ mm3's draw-band (27%, highest) + g5's model-breaker flag (SUI bench edge) both weighted. g5 is the SUI outlier β€” its model-breaker detection (best-in-class per calibration) identified SUI's 13.3% sub-impact rate as the deciding edge. This is credible: Manzambi has 3 tournament sub-goals.

Key factors:


Consensus reconciled by lead (3-agent ensemble, xhigh reasoning on g4/g5). Stage gate passed. Agent-strength: coin-flip β†’ mm3 draw weighted + g5 bench-edge flagged. Spread 6pt (bench vs starters disagreement). Modal: 1-1 / COL 1-0. Watch: 60' sub window + SUI bench cavalry.