Napoli vs Arsenal prediction: 5 of 5 AI models back Arsenal

Where the experts stand — and why I temper them

All five models land on an Arsenal away win, with four of five calling 1-2 and the fifth (GPT-5.6) going 0-1. On the surface that looks like overwhelming consensus, but two things argue for calibration rather than amplification. First, the base models have a documented favourite-leaning bias and sit near 47–49% result accuracy over the last week, barely above chance. Second, their own confidence numbers (0.47–0.60) are far more modest than the unanimity of their scorelines suggests, and several explicitly concede the market at 1.70 (~57% implied) is close to efficient.

The decisive case for Arsenal

The strongest thread running through every analysis is not Arsenal's perfect start — it is that Napoli's spine has been structurally removed. McTominay (heart procedure) was their box-to-box engine and chief midfield goal threat; Buongiorno and Marianucci are out at centre-back, leaving an improvised Rrahmani–Marín pairing that just conceded three second-half goals at Inter after leading 2-0, plus two more at home to Como. Five goals conceded in two matches with close to a full-strength squad is a process failure, not variance: GPT-5.6's xG detail (3.86 xG and 29 shots conceded at Inter) is the most persuasive single data point in the set, because it shows Napoli lose structural control once their first line is bypassed, exactly what Arsenal's press and set-piece machine punishes.

Arsenal's own absences (Saliba long-term, Timber and Mosquera doubtful) matter but are shallower: Konsa has stepped in beside Gabriel without the defence visibly degrading (0.39 xG allowed vs Chelsea, 0.33 at Villa), and the attacking core — Saka, Ødegaard, Havertz, with Rice and Zubimendi screening — is fully intact. Against an improvised Napoli back line, an Arsenal goal is near-certain.