Corners, xG and the decision tree

A goal model is the core, but FinalSkore layers on a few more ideas to round out the picture and sanity-check the result.

Corners get their own model

Corners are predicted the same way goals are, just on different data: each team's historical corner averages (home HC and away AC figures) feed an expected-corners estimate, which prices the corner totals and handicaps. Same machinery, different event.

xG — expected goals as a quality signal

Expected goals (xG) scores every chance by how likely it was to be scored, so it measures the quality of chances a team created and allowed — not just the lucky final score. A side winning on xG while losing on the scoreboard is often underperforming its true level and due to regress. Using xG-style quality rather than raw results makes the form window more honest and less fooled by a single deflected winner.

The transitive "decision tree"

FinalSkore also runs a transitive decision tree: if A reliably beats B, and B reliably beats C, that's evidence about how A should fare against C, even with little direct history. It chains results across the league to fill gaps that a head-to-head record alone can't, and acts as a cross-check on the goal model's view.

Putting it together

The model blends the Poisson goal estimate, the corner model, xG-flavoured form, and the transitive tree, then compares the result to the free 1X2 and Over/Under odds to surface a model pick where its view and the market disagree. No single component decides — they vote.