Nothing here clears 43%. Fiorentina edge it at 42% against 25% for the draw and 33% the other way — the inputs point in different directions and the engine is not claiming to know. That is published rather than dressed up.

Note that the projected scoreline of 1–1 points to a draw while the probabilities favour Fiorentina. Those answer different questions. A home win is spread across 1–0, 2–0, 2–1 and more, so no single home scoreline need be the most likely result on the board even when home is the most likely outcome. The engine publishes the modal scoreline and the outcome probabilities side by side rather than forcing them to agree.

Across 10 bookmakers the consensus lands within 1 point of our own on this outcome. Where the model and the market agree, expect the market’s accuracy rather than ours — agreement is not evidence that we are right, only that we are not saying anything unusual.

The projection totals 2 goals. That sits under the 2.5 line, though only just: one more goal from either side and that call is wrong.

Confidence is filed as low at 35/99. That tier is set from the shape of the probability distribution, not from how the fixture feels, and it is published before kick-off and graded against the result afterwards — right or wrong, it stays on the page.