Who gets booked most in the Premier League — and what for?

29% of cautions are for something with no challenge in it.

What bookings are for

4,704 yellow cards, 2023-24 to 2025-26. Shares are of the 4,583 whose reason the feed recorded; the other 121 are excluded rather than assigned, so these sum to 100.

Season by season

Close to fixed across three seasons. A distribution that jumped about would be describing the recorder, not the referees. The count is the cautions whose reason the feed recorded, which is the denominator the shares use — so each share times that count gives its own caution count, checkable against the rows above.

How far each category is checked

One feed cannot check itself. Only an outside figure is a test, and one exists for two of these five. Dissent: Sky Sports, 15 December 2023 (Grounds & Smith), the opening 16 gameweeks of 2023-24. Time Wasting: The Analyst (Opta), 22 August 2023, the opening ~19 matches of 2023-24. The two checks come from different outlets and different windows, so they are quoted separately rather than combined.

Is "arguing" really dissent?

Partly not. Two checks sharing no assumptions agree: we hold almost exactly double the dissent count reported for the same 16 gameweeks, while time wasting from an outside source matches us exactly — and 27% of these cautions have an opposing one in the very same minute.

224 of 828. The null reshuffles each match's own card minutes, so the bunching of cards in busy spells survives and only the pairing is destroyed; in 200 reshuffles it never exceeded 12.3%.

When each kind of card is shown

The labels are the feed's own, so worth testing rather than trusting. An arbitrary label would show no such pattern — dissent is near-absent in the opening quarter-hour, as an offence that needs a decision to argue with should be.

Red cards

Second yellows counted as the red they became, so a dismissal for two cautions appears under the offence that drew the second.

The most-booked players, 2025-26

A census, not a title. 3 players share the lead on 12, one clear of 11 — and the same has been true in every season we hold. A page that crowned one of them would be asserting a gap the data does not contain. 383 players were booked at least once.

2026-27 is excluded: its feed uses a different booking-reason vocabulary (tripping, roughing, holding, diving appear for the first time; argument and time wasting, 26% of bookings in the three seasons here, are absent from its first 35 cards). Merging them would describe neither.

What this is not

A record of what referees penalised, not of who deserved it. Player-by-referee pairings are deliberately absent: the most any referee has cautioned any player is seven, across every booking on record — far too few to carry the inference. See why nobody can be called the strictest referee.

Where these cards came from

Other questions