Head-to-Head (H2H): Does It Work in Predictions?

Head-to-head is the history of matches between two specific teams. It often looks like the most convincing argument in a prediction ("this team hasn't lost to them in five years"), but in practice it is one of the noisiest signals. Below is how to read H2H stats, when they work and why they should not outweigh current form.

What head-to-head history shows

H2H stats usually include:

• results of the teams' recent meetings;
• goals scored and conceded in those matches;
• where the meetings took place: at one team's ground or the other's;
• sometimes totals, corners and cards in those matches.

The idea is that some combinations of styles are awkward for one of the teams. For example, a possession-based side may struggle against an opponent that defends deep and counters quickly. If that match-up repeats, it can show up in the head-to-head record.

Does head-to-head affect the result?

Short answer: yes, but weakly and not always. Three main reasons to treat H2H with caution:

1. A small sample. Two teams from the same league usually meet twice a season. The last five meetings can stretch over two or three years.
2. Teams change. Coaches, squads and systems change over that time. A match from three years ago was often effectively played by different teams.
3. Randomness. Football is low-scoring, so a single result depends heavily on chance. Over five matches randomness easily creates the illusion of a pattern.

So head-to-head is useful as a supporting signal that confirms or questions a conclusion drawn from form, not as the main argument.

How head-to-head is used in the xG.Football model

In the xG.Football model head-to-head is one of five criteria for a match result prediction. The other four are overall form, home and away form, goal threat and goal difference. So H2H contributes at most one point out of five and cannot decide a prediction on its own.

For head-to-head the model takes the pair's last 5 matches, but it scores the criterion once two meetings have been played. For totals (corners, shots on target, cards) H2H is also one of five criteria: the model checks the share of meetings in which the stat went over the line.

There is also a separate "head-to-head at the home side's ground" slice, which only includes matches at the same stadium. It is used in the strict selection of "Both teams to score" predictions.

How to read H2H stats properly

• Look at the dates. The older the match, the less it means.
• Separate the venue. Home and away meetings are different stories.
• Check squads and coaches. If a team has changed coach since the last meeting, past matches say little.
• Compare with team form. If H2H and current form point the same way, the signal is stronger. If they disagree, current form usually matters more.
• Don't mix competitions. A cup tie with a rotated squad and a league match are different situations.

When head-to-head is actually useful

• The teams have played in the same league for a long time, with similar coaches and squads.
• There is a clear tactical reason why one team is awkward for the other.
• You are looking at totals and corners in matches where both teams' styles are stable.
• The head-to-head history matches what current stats show.

Common mistakes

• Drawing conclusions from one or two meetings.
• Weighing matches from five years ago the same as recent ones.
• Believing in "curses" and "bogey teams" without an explanation of why it happens.
• Putting H2H above current form and line-ups.

Ready-made predictions in which head-to-head is only one of several signals are collected in match predictions. On each match page you can open the head-to-head sample and see the actual matches behind the calculation.

The short version

Head-to-head is a useful but weak signal. Look at recent meetings, account for the venue and changes in the teams, and use H2H to check a conclusion drawn from form, not to replace it.