Aston Villa vs Liverpool

The model outputs 10.15 predicted corners for this fixture, sitting just below the Premier League baseline of 10.5. That gap is tight. Villa's five-match rolling average lands at exactly 10.5, and the constituent scores, 13, 10, 9, 12, and 11, suggest a team whose corner count is reasonably predictable within a narrow band. The home side's data is coherent, and it points consistently.

Liverpool's data tells a different story. Their last five corner totals are 13, 5, 7, 12, and 13, producing a match average of 9.9 but masking a range that renders that figure far less reliable. Five and thirteen are not neighbours. When a team's output swings that sharply across consecutive fixtures, the average obscures more than it reveals, and any model built on that input is working with genuine uncertainty rather than noise that conveniently cancels out.

That spread is the dominant factor behind the medium-confidence rating here. The model's structural output sits at high confidence, and the predicted total is directionally consistent across both team inputs. But Liverpool's variance is enough to widen the uncertainty band materially. The model's read and the pick confidence are not the same thing. This fixture illustrates why. Stake sizing at medium-confidence levels is the appropriate discipline.

In the Premier League, Spectral's medium-confidence corners picks stand at 101 settled, with a 46.5% win rate.