Leeds UnitedvsBrighton and Hove Albion

Leeds' home variance is the issue. Their last five home fixtures produced totals of 5, 11, 8, 11, and 8, a spread that resists a clean read and places meaningful noise around the model's 9.85 predicted total. Brighton's away numbers are tighter: 12, 12, 10, 10, 10 across the same window, with a match average of 9.6. That consistency on the Brighton side helps anchor one half of the projection, but the Leeds side won't settle cleanly, and that asymmetry is the first reason to approach this with caution.

Both teams sit below the Premier League baseline of 10.5. Leeds' home average is 9.5, Brighton's away average is 9.6, and the model's 9.85 projection reflects both inputs pulling in the same direction. That directional alignment is genuine. What complicates the picture is Leeds' demonstrated capacity for outlier outputs: the 5-corner game appears alongside the 11-corner results in the same recent window, and that kind of range means the central estimate carries wider-than-usual uncertainty bands.

The medium-confidence rating reflects that scatter directly. Stake discipline matters here. The model's output confidence is technically high, but the pick is graded medium once team-level variance is factored in, and medium-confidence outputs in this model are sized proportionally. Scaling up on the model signal alone, without accounting for the variance in the underlying data, is where edge gets given back.

In the Premier League, Spectral's medium-confidence corners picks have settled 77 times with a 45.5% win rate.