Everton vs Manchester City

The model produces a predicted total of 10.16, sitting just below the 10.5 league baseline. That's the cautious read. When the model and baseline converge this closely, the signal carries less separation from the mean, and medium confidence is the honest output of that narrow gap, not a downgrade applied as an afterthought.

Both teams' recent data show spread. Everton's home corner totals from recent fixtures read 14, 7, 12, 9, 11, around a 9.8 match average. City's away figures read 14, 14, 16, 16, 8, averaging 9.9 across the same window. The lone 8 in City's sequence sits far from the results preceding it, and the model has to assign probability mass to whether that match reflects structural behaviour or isolated noise in City's away approach. Home and away averages at 9.8 and 9.9 leave no skew to resolve the ambiguity.

Dispersion is wider here. Medium confidence signals that the model's distributional certainty is lower, the inputs carry more variance, and the separation between predicted total and baseline is thin. Stake sizing should reflect that uncertainty rather than chasing a position on the back of headline numbers, which can obscure how much noise sits underneath them. The model's job is to quantify that, not smooth it over.

In the Premier League at medium confidence, Spectral has 74 settled bets at a 44.6% win rate.