Head-to-head (H2H): the public's favourite statistic and the weakest
TL;DR. «They've beaten them the last four times» is an irresistible argument and almost always a weak one. The head-to-head record has a structural problem: the teams of three years ago are not today's teams. There are exceptions, and it is worth knowing which.
Note. Informational and statistical content. It is not betting advice nor a promise of profit. No statistic predicts a single match: they are there to estimate probabilities over the long run. 18+ only. Play responsibly.
The underlying problem: small sample, expired data
Within a league, two teams meet twice a season. Five years of history is ten matches, and the ones from five years ago were played by squads, managers and systems that no longer exist.
So H2H combines the two worst possible properties for a statistic: little sample and old data. Either alone would already weaken a conclusion; together they make it close to useless in team sports.
And there is a third trap: H2H blends home and away matches, different competitions and different points of the season. Ten matches that share no context are not a sample of ten.
Where it does add value
| Sport / case | Does H2H add? | Why |
|---|---|---|
| Tennis | Yes, considerably | The same individuals, and styles counter each other stably |
| Team sports, unchanged squad | Somewhat | If the core has not changed, recent precedent keeps some value |
| Derbies and rivalries | Somewhat | The emotional context repeats year after year |
| Team sports, 3+ years back | No | Almost nothing remains in common between those teams and these |
Tennis is the clear case: direct meetings are between the same people, and a style that troubles another tends to keep troubling it. A left-hander with a particular one-handed backhand can hold a disproportionately favourable H2H against a specific opponent, and that is genuine information.
How to read it without fooling yourself
- Check the dates before the record. Four wins, three of them before 2023, are one recent win and three archive entries.
- Split home and away. A 4-0 H2H in which all four were at home does not describe general superiority.
- Check for changes. A new manager, a change of system or a generational turnover in the squad invalidate the precedent entirely.
- Be suspicious if it is the only argument. If H2H is all that holds an idea up, there probably is no idea.
What we do
H2H is not a predictive input in our team-sport engine, for the reasons above. We do publish it as context on the match page, because readers want to see it and because, presented with its dates, it judges itself.
Since July 2026 it is also frozen alongside the rest of the pre-match context, so the page preserves what was known at the moment of the forecast. In tennis, the weight of direct meetings is channelled through the per-surface rating, which is where that style effect makes sense.
Frequently asked questions
Is the head-to-head record useful?
Little in team sports: the sample is tiny and the older matches were played by squads and managers who are gone. In tennis it adds considerably, because the same individuals meet and styles counter each other stably.
Why does H2H work better in tennis?
Because the meetings are between the same people. A style that troubles another tends to keep troubling it, and that is real information which does not expire with a squad change.
How should a 4-0 H2H be read?
By checking the dates first and whether all four were at home. Four wins spread over five years, three of them at their own ground, do not describe general superiority but a handful of matches without shared context.