Every stats site shows home and away form splits, and every punter reads them. We tested whether they mean anything: across 30,582 matches in every league we cover, roughly 94% of home/away over-2.5 gaps are statistically indistinguishable from chance — for over 3.5 it is ~99%.
Method: for each team-season we compared the observed venue split against the spread you would expect from pure binomial noise at that team's overall rate. The observed distribution of splits barely exceeds the coin-flip prediction. Even the most extreme split we found — 78% overs at home versus 48% away — was only borderline-real (z ≈ 2.1), the kind of gap chance produces regularly across thousands of team-seasons.
The proof that matters: we built model variants that trusted venue splits more — venue-only rates, venue evidence double-weighted. Every one graded worse in walk-forward testing than pooling venue and overall evidence together. Venue matters for who wins (home advantage is real); it barely matters for how many goals. A travelling side scores a little less and concedes a little more — the total roughly cancels.
Consequences on this site: headline form shows all-venue records, the ratings pool venue and overall evidence, and our index weights moved to 40/40/20 after this study — beating the previous weighting on all three goals markets.
Method: walk-forward testing — every prediction is made using only information available before kick-off, then graded against what happened. Studies run June-July 2026 on the site's own dataset. Questions: contact us.