Crystal Palace 1–4 Manchester City : l'IA avait prédit 0–2 ✓

The Unified Read

All five models agree on the fundamental outcome: Manchester City are the most likely winners at Selhurst Park, with 0-2 as the modal scoreline. But the agreement itself isn't where the value lies — it's in the reasoning behind it, and specifically in what the market hasn't fully absorbed.

Palace's Attacking Collapse is the Decisive Factor

The single thread running through every expert's analysis is Palace's goal-scoring drought. They have failed to score in three of their last six competitive matches — 0-2 at Everton, 0-3 vs City, 0-3 vs Bournemouth — and generated approximately 2.0 xG at Goodison Park without converting. This isn't a finishing blip; it's structural. The confirmed absence of Ismaïla Sarr (21 goals last season, groin injury) removes their primary counter-attacking outlet, and Mateta's contract dispute introduces uncertainty into their most important remaining attacker. Against a City side that kept clean sheets in three of their last six, Palace's route to a goal looks extremely thin.

Resolving the Confidence Gap

The key tension to resolve is between models seeing a comfortable City win (GLM-5.2 at 63%) and those seeing a grindier affair (Grok 4.6 at 55%, GPT-5.6 at 57%). Weighting by historical accuracy — Grok 4.6 (60% 7-day, 52% 30-day) and Claude Opus 5 (59% 7-day, 53% 30-day) are the strongest performers; GLM-5.2's 63% confidence carries less weight given its middling accuracy and the known tendency of base models to be overconfident on favourites.