Head-to-Head Stats: When They Matter

"They’ve won the last 5 H2H meetings" is one of the most over-quoted lines in football punditry. Sometimes it matters. Often it doesn’t.

When H2H matters - Same-era squads : H2H from the last 2 seasons with mostly the same players is meaningful - Tactical match-ups : Some sides genuinely have systems that bother specific opponents - Draw patterns : 4+ draws in last 6 H2H is a real signal — we weight it at 15% in our draw model

When H2H is noise - Sample sizes under 6 matches - Results from 5+ years ago when both squads were completely different - Cross-competition mixing (a Cup match vs a league match)

How we use it Our engine pulls the last 10 H2H matches and weights them by recency. Older meetings count less; meetings in the same competition count more. We never use H2H in isolation — it’s one of 13 signals in the winner model.

A practical rule If H2H disagrees with form, xG, and Elo — trust the data. If H2H agrees with the other signals, treat it as confirmation, not extra weight.