Before and after the dismissal: an event-window scorecard

A careful way to read shots, goals and possession around a red card without mistaking sequence for causation.

A red card looks like a clean before-and-after story: one side loses a player, so the numbers should move. But a match is never reset at the dismissal. The team may already be ahead, behind, under pressure, or producing fewer attacks than its opponent. A useful scorecard asks a narrower question: what changed in bounded windows around the dismissal, and what was the score when it happened?

This event-window approach separates timing from interpretation. It records shots, goals and possession before and after the card, uses comparable denominators, and keeps match state visible. It can describe a response; it cannot, on its own, prove that the red card caused every later change.

TL;DR: Treat the red card as a timestamp, not a reset. Compare bounded before-and-after windows, preserve score and context, pair counts with minutes observed, and describe sequence without claiming causality.

Why the clock matters

A dismissal is an event, not a new match. Keep the timeline intact and treat the red-card minute as a boundary. Before it, each team has built a scoreline, shot profile, possession pattern, and tactical context. After it, the same match continues under a changed personnel constraint. If those starting conditions disappear, a post-card swing can be mistaken for the whole story.

The scorecard is descriptive by design. It asks whether shot volume, goals, or possession share moved between defined intervals. Since dismissals are not random, read the result as evidence about match sequence, not as an experiment.

Define the window before looking at the result

Start with a consistent event record: dismissal timestamp, dismissed team, scoreline, venue status, and whether the match is complete. Choose the comparison windows before calculating results. A near-event view might use 15 minutes before and after the dismissal. A full-match view compares the pre-dismissal segment with the remainder. The first gives equal time at risk; the second reflects the practical rest-of-match picture but gives late dismissals less exposure.

Neither window is universally correct. Label the choice and keep it stable. An early card cannot supply a full pre-window, while a stoppage-time card cannot supply a meaningful post-window. Mark such cases ineligible instead of padding them with invented zeros. If there is more than one dismissal, analyse the first once and flag later cards, or define a separate multi-event design. Decide whether the recorded minute starts the post interval, whether the event itself is excluded, and how stoppage time is represented.

Keep the score at the dismissal

The score is not decorative. A leading team may protect its result after a dismissal; a trailing team may take more risks; a level match may produce another balance of possession and shots. Those are different match states even when the player-count change is identical.

Store the score for both teams at the boundary and classify each as leading, level, or trailing. Retain home or away status and the dismissed side. When coverage exists, add pre-window shot differential, possession share, match minute, and a transparent strength context such as standings or team aggregates. These controls do not remove confounding, but they stop unlike situations becoming one headline average. Do not reset the score to 0-0: it is the baseline that gives later events meaning.

Choose denominators that make comparisons fair

Raw totals are easy to misuse. A 20-minute remainder naturally contains more shots than a five-minute remainder. Report counts with exposure. Fixed windows have the same minutes for every eligible observation, so show shots per window or per minute. Remainder windows should use minutes actually observed and keep their length visible.

Goals are sparse: show the count before and after, describe the direction, and avoid false precision from a single event. If a rate is used, name its denominator. Express possession as a share of covered time and check that the opponent's share is consistent. If coverage is missing, use only usable minutes and flag the gap. Missing possession is not zero possession, and no recorded shot may be a real zero or incomplete feed.

Build a scorecard that preserves sequence

A compact scorecard begins with the event line: dismissal minute, dismissed team, and score. Add each window's length and coverage status. For each team, show shots for and against, goals for and against, and possession share before and after. Pair counts with denominators, then show the absolute change and, where useful, possession change in percentage points.

Keep the teams separate rather than collapsing them into a match total. One side's attacking volume may fall while the other's rises; the total hides that asymmetry. Possession can move without a matching change in chance quality, and a trailing side can add low-probability shots. If expected goals or field location are not covered consistently, say only what the available shots support. The evidence boundary is completed fixtures, standings, match events, and team or player aggregates when coverage is present.

Read the change as a sequence, not a verdict

Several descriptive patterns matter. A side may lose possession immediately but recover it later. Its shots may fall while the opponent's rise, or both may fall as a leader slows the match. Goals may stay unchanged while volume shifts. The pre-window tells you whether the trend began before the card: a team already conceding shots or losing possession may simply continue that path.

Use grouped summaries when the sample permits: leading, level, and trailing states; early, middle, and late dismissals; home and away contexts; and dismissed side. Mark small groups as descriptive. The purpose is to reveal different match states, not to turn a neat average into certainty.

What the scorecard cannot prove

A red card is connected to the events that produced it. Fouls, second bookings, dangerous challenges, tactical breakdowns, referee decisions, player quality, fatigue, and the existing score can relate to both the dismissal and what follows. Dismissals are therefore not random treatments, and even careful before-and-after differences remain vulnerable to selection and time effects.

The scorecard cannot say that every later shot, goal, or possession swing was caused by the dismissal. It cannot establish a universal red-card effect from a descriptive sample, fill gaps in event coverage, turn standings into a precise current-strength measure, or infer tactical intent that was never recorded. A useful finding may be that attacking volume changed in one state but not another, with incomplete coverage in a third.

A reproducible workflow for every fixture

The workflow is repeatable. First select completed fixtures with the required event and time coverage. Freeze the event definition and timestamp rule before calculating results. Capture score, venue, dismissed side, and pre-event state. Create the fixed or remainder windows, attach explicit exposure denominators, and calculate shots, goals, and possession for each team separately.

Then check coverage: confirm that zeros are real, every rate has a visible denominator, and early or late dismissals were not forced into unsuitable windows. Group by match state and timing only when observations support interpretation. Preserve fixture-level rows so any aggregate can be audited back to the event line. Finally, write the interpretation beside the numbers: what moved, what was already moving, and what remains uncertain. The wider football data behind this framework can be identified as Xtra-Stats, while the article should still state its specific coverage.

A quick audit checklist

• Is the dismissal timestamp and scoreline recorded at the boundary?
• Are the window definition, eligible minutes, and denominator stated?
• Are shots, goals, and possession separated by team and period?
• Are match state, home/away context, strength context, and gaps visible?
• Does the language describe sequence without claiming causality?

The useful reading rule

The best scorecard leaves three answers: what changed after the dismissal, what was already true before it, and how much match time was observed. Treat the red card as a timestamp inside a live contest, not a reset button. Preserve the score, bound the windows, normalize exposure, and put limitations beside the comparison. That makes the question answerable without asking descriptive data to prove more than it can.