First Four matchups are typically between teams with similar efficiency profiles - that's often why they're in the play-in. Minor injuries, travel fatigue, and home crowd effects dominate. Models trained on season-long data have less signal to work with.

Full First Four predictions are visible on the Bracket page once collection runs complete March 18. The consensus pick - if all three models agree - is worth noting. But disagreements may be more interesting: when one model bucks consensus on a play-in game, that's where source-driven reasoning is most visible.