Define the outcome and exposure
Cards per match, cards per foul, penalties per box entry and added time answer different questions. Use consistent competition and period rules. IFAB's Laws define the official framework, while competition implementation and VAR periods can still differ.
Context table
| Covariate | Why it matters |
|---|---|
| Teams and rivalry | Foul and dissent opportunities differ |
| Score state | Match behaviour changes |
| Competition and season | Rules and VAR implementation change |
| Crowd and venue | Decision environment can differ |
| Assignment process | Referees are not randomly allocated |
A Bundesliga study models card behaviour alongside team running measures, illustrating that raw referee averages omit match context. Research on matches without supportive crowds found a change in visiting-player cautions in its sample, while some outcome effects remained uncertain.
Reproducible comparison
- Freeze appointments and data at the decision time.
- Require a predeclared minimum exposure, but publish the exact count rather than a universal cutoff.
- Model card counts or rates with teams, venue, score and season.
- Use partial pooling so sparse officials are not ranked by extreme raw averages alone.
- Test the model on later assignments.
Illustrative rate
An official showing 24 cards across 6 matches has 24 / 6 = 4.0 cards per match. That number has a small denominator and does not say what caused the cards. If those matches contained unusually high foul totals, a contextual estimate may differ materially.
Separate referee tendency from match assignment
Referees are not allocated to a random mix of fixtures. Competition, derby status, team style and match importance can influence both appointment and card exposure. A raw cards-per-match table therefore describes assignments and outcomes together.
Build a comparison with these fields:
| Field | Why it matters |
|---|---|
| Matches and minutes | Exposes sample size and abandoned games |
| Fouls and cards | Separates opportunities from sanctions |
| Teams and competition | Controls recurring assignment mix |
| Score state | Captures match pressure and chasing |
| Penalties and red cards | Shows rare outcomes separately |
| Season | Detects rule or enforcement change |
Reproducible rate example
If a referee shows 36 cards across eight eligible matches, the descriptive rate is 4.5 per match. Keep both 36 and eight visible. Compare with matched fixtures or a model that accounts for teams and competition; do not infer that the next match will contain 4.5 cards.
Betfair's current football rules show why statistical event records and product settlement definitions must be checked separately.
For any market comparison, verify the operator's card-point and settlement definitions independently. Statistical event labels and betting product rules may not align. Timestamp appointments and prices, because using a referee feature before the official assignment is known would make a backtest impossible to reproduce live.
Retain a list of matches excluded for missing officials, incomplete cards or unusual duration. Silent exclusions can shift a small referee sample materially and make another analyst's rate impossible to reproduce.
Related resources
Read limits of statistics for confounding or cards market rules before comparing a statistic with a product.
Continue learning
- Next guide: Home Advantage in Football
- Related guide: New Manager Bounce
Assumptions and limitations
The example is illustrative. Public referee assignments and event data can be incomplete or corrected. Observational models cannot fully isolate referee disposition from selection, teams and match conditions, and no average guarantees a future card count.
GambleAware
GamCare
Gambling Therapy
GAMSTOP