Universitatea Craiova vs Levski Sofia prediction: 5 of 5 AI models back Universitatea Craiova

Synthesized Match Read

All four models converge on a Craiova home win, and three of four project the identical 1-0 scoreline. This consensus is meaningful but not self-validating — the models share a documented favourite-leaning bias, and the market already prices Craiova at 1.85 (~54% implied), roughly where the raw agreement would place them. The calibration question is whether the evidence genuinely supports a home-win probability above 50%, or whether the cluster is amplifying a thinner edge.

Resolving the scoreline disagreement

Kimi K3 (the weakest model by 30-day accuracy at 43% and negative ROI) is the outlier at 2-0, arguing that once Craiova score, Levski must open up and a second goal follows. GPT-5.6 — the strongest performer (+22.1% 30-day ROI, 50% result accuracy) — directly counters this with a specific, verifiable factor: Assad Al Hamlawi's confirmed absence removes Craiova's clearest penalty-box finisher, making a controlled 1-0 more likely than a 2-0 rout. Grok and GLM both land on 1-0 without naming Al Hamlawi, but their reasoning (Levski's deep block, cagey qualifier dynamics) points the same direction. I weight GPT-5.6's argument heaviest: the finishing gap is a concrete, match-deciding detail, not a generic narrative. 1-0 is the modal scoreline.

Why confidence stays below 50%

The draw is genuinely live. With away goals abolished, 0-0 or 1-1 sends Levski through — they need only avoid losing. Levski demonstrated in the first leg that they can create transition chances against Craiova's advanced wing-backs (Oko-Flex denied one-on-one, Bouras missed a break, Bala had a goal ruled out). If Craiova chase a second goal late, one counter-attack produces 1-1 and eliminates the hosts. This is not a tail risk; it is a central alternative scenario. The models' own counter-arguments consistently acknowledge this, which is why none rate the home win above 52%.