Define the dismissal correctly
The IFAB fouls and misconduct law distinguishes sending-off offences and disciplinary states. A research record should separate direct red cards, second-yellow dismissals, the dismissed person, team, event time and later correction.
| Field | Why it matters |
|---|---|
| Team receiving red | Separates numerical advantage and disadvantage |
| Score immediately before | Captures tactical and outcome state |
| Match minute and actual time | Defines remaining exposure |
| Home or away | Retains venue context |
| Relative pre-match strength | Separates dismissal from baseline quality |
| Player role | Supports personnel-specific analysis where data allow |
| Substitutions | Captures immediate tactical response |
| Review and correction | Prevents a rescinded or changed event from leaking backward |
What focused research can and cannot show
A peer-reviewed hazard study examined first red cards and subsequent scoring in men's World Cup matches from 1998 to 2014 (Empirical Economics). Its design is relevant to conditional scoring rates, but the sample and tournament setting limit transfer to domestic leagues, women's football, later periods or current live markets.
Do not turn one coefficient from that study into a universal odds adjustment. Local analysis needs the same careful state definition and its own later-period validation.
Remaining exposure matters
A dismissal early in a match leaves more time for its associated state to operate than one in added time. Use actual elapsed time under the IFAB duration law and censor or transition the state when another sending-off changes player counts again.
Illustrative state table
| State before event | State after event | Target for analysis |
|---|---|---|
| Home leads, 11 v 11 | Home leads, 10 v 11 | Next goal and final result conditional on lead and disadvantage |
| Draw, 11 v 11 | Draw, 11 v 10 | Same targets for advantaged home team |
| Away leads, 10 v 11 | Away leads, 9 v 11 | New player-count state, not continuation of first event |
The table prevents pooling materially different situations under "a red card happened".
Estimation and validation
The Wyscout open-data paper documents event data that can support timestamped state transitions. Build intervals before and after the event, model score and player-count states, and account for team and competition differences.
The peer-reviewed Bayesian in-play model illustrates conditional probability updating. A red-card extension should be compared with a score-time-strength baseline, evaluated chronologically, and checked for calibration in sparse subgroups.
Live decision boundary
Mark the incident unresolved during any check or review. Do not use a later confirmation in an earlier feature set, and do not assume an old price remained executable through suspension. Save the first reopened quote, available size and accepted receipt separately.
Next step
Use How Odds Move During Match for the next part of this topic.
Continue learning
- Next guide: Second-Half Goal Statistics
- Related guide: Weather in Live Football Analysis
Assumptions and limitations
Red cards are not random interventions; match pressure, team behaviour and player actions influence when they occur. Observational adjustment may leave residual confounding. Sparse combinations of minute, score and player count produce wide uncertainty, and historical effects do not establish a favourable current price.
GambleAware
GamCare
Gambling Therapy
GAMSTOP