Manchester City vs Crystal Palace

The model's predicted total of 10.75 sits close to the Premier League baseline of 10.5. That gap is narrow, and it shapes the cautious read on this fixture. Both sides carry recent match-level averages close to the baseline, 10.3 for City and 10.1 for Palace, rather than the kind of elevated figures that create a clearer separation between projection and market.

Variance is the core uncertainty. City's last five match totals read 13, 12, 13, 7, 10. The 7 is the outlier, a reading that reflects how quickly game state or a defensive setup can suppress corner counts regardless of underlying tendency. A 10 also sits in that sequence, a result that drags the rolling average below the projected figure. Palace show a similar spread across their last five: 13, 13, 8, 13, 9. The high readings dominate, but the 8 and 9 are present, and they matter when calibrating distribution around the average.

The model registers high confidence in the projected total itself. The pick sits at medium confidence. That's a different question. The distinction matters: a slim margin between projection and baseline leaves less buffer for the model to be directionally right while the position still settles correctly. When the edge is thin and the input data carries this much within-sequence variance, a measured position is the right response.

Stake discipline is appropriate here. In the Premier League, Spectral's medium-confidence corners record stands at 46.5% across 101 settled picks.