FK Crvena Zvezda and FC Copenhagen served up a dramatic comeback that left our pre-match prediction in pieces. Zvezda looked to be cruising after going two goals up—an own goal from Gashtarov in the 29th minute, then clinical finishes from Elsnik (34') and Kostov (41'), both set up by their midfield. At that point, the designated home side was in complete control and tracking our 2-0 forecast. But Copenhagen refused to fold. Carrizo pulled one back from an Markgraf assist in the 70th minute, and a late equaliser forced the match into a draw that neither side had managed before the final whistle. A red card for Gudelj deep into injury time capped an eventful finish.

Our model had assigned a 24% probability to a draw, making it the second-most likely outcome behind a Zvezda win. That said, we'd clearly underestimated Copenhagen's resilience away from home and overestimated how conclusive Zvezda's early dominance would prove. Going in, the metrics pointed to Zvezda's strength at home—they'd been in good form—and Copenhagen's vulnerability on the road. What we missed was that knockout football does funny things to teams with their backs to the wall. When Copenhagen went two down, instead of folding, they found another gear and dragged themselves level. It's the kind of twist that reminds us predictions are leans, not prophecies. The match went to a draw; our model didn't back it strongly enough.