Chelsea overcame poor form to upset Tottenham 2-1, with Enzo Fernández orchestrating the victory through an early opener in the 18th minute and a crucial assist in the 67th that sent Andrey Santos through. The goals came either side of the interval, establishing a two-goal cushion that Richarlison's 74th-minute response proved insufficient to overturn. It was a performance that contradicted the pre-match narrative almost entirely.

Our model predicted a 1-1 draw with only 25% confidence in a Chelsea victory, heavily influenced by the visitors' recent form collapse and Tottenham's rest advantage heading into the fixture. The analysis flagged Chelsea's motivation concerns at mid-table, the caginess typical of derby football, and rain-dampened conditions as factors pointing toward a low-scoring stalemate. Instead, Chelsea's attack clicked early and decisively, with Neto's involvement in the opening goal proving pivotal. While both sides contributed to the goal tally as expected from their head-to-head history, the distribution favored Chelsea emphatically. Tottenham struggled to impose themselves in the opening phases despite their fresher legs—a significant miss for a team fighting for position.

The result exposes how narrow the margins are in prediction modeling, particularly when form reversals occur. Chelsea's dormant attack awoke at precisely the wrong time for our assessment, while Tottenham's tactical setup or execution failed to capitalize on the preparation advantages. The late Richarlison goal salvaged some respectability but couldn't alter a match decided by Chelsea's early penetration and composure. For the model, this serves as a reminder that form anomalies and rest metrics, while statistically sound, don't always translate to pitch reality.