What happened when the Premier League cracked down on dissent?
Both halves sagged in the first season. Only one came back.
Inside the first season, both fell away
| Matchweeks | Cautions | Dissent | Time wasting |
|---|---|---|---|
| MW1–10 | 472 | 20.6% [17.2, 24.4] | 13.3% [10.6, 16.7] |
| MW11–19 | 406 | 21.9% [18.2, 26.2] | 10.6% [8.0, 14.0] |
| MW20–29 | 396 | 19.9% [16.3, 24.2] | 11.9% [9.0, 15.4] |
| MW30–38 | 341 | 15.0% [11.6, 19.1] | 6.7% [4.5, 9.9] |
Both declines clear their own margin of error across the season: dissent p=0.0413, time wasting p=0.0025.
Then they diverged
| Season | Cautions / match | Dissent | per match | Time wasting | per match |
|---|---|---|---|---|---|
| 2023-24 | 4.25 | 19.6% | 0.832 | 10.9% | 0.463 |
| 2024-25 | 4.09 | 17.1% | 0.697 | 8.2% | 0.334 |
| 2025-26 | 3.72 | 18.5% | 0.689 | 7.8% | 0.289 |
Share and rate are not the same thing, and this is why both are here. Cautions overall fell from 4.25 a match to 3.72, so a category can hold its share while falling in absolute terms. Dissent did exactly that — down from 0.832 to 0.689 a match, but no change in share (p=0.4684). Time wasting fell on both measures, 0.463 to 0.289 a match and 10.9% to 7.8% of cautions (p=0.0034). That is what makes it the half that did not stick.
What this does not tell you
The composition says what referees booked, not why it changed. Three explanations fit the same curve, and nothing here separates them:
- Players adapted — fewer players argued or wasted time once the cautions started.
- Referees relaxed — the same behaviour stopped drawing a card as the season wore on.
- The instruction itself changed — emphasis moved elsewhere after the opening weeks.
Our booking-reason data begins with the clampdown season itself, so every figure on this page describes what happened AFTER the instruction. There is no before in our data, and none is invented. 2026-27 is excluded: it holds 186 reasoned cautions, 5 rounds. Its feed also uses a different booking-reason vocabulary, which db.canonical_reason folds correctly — the problem is the sample, not the labels.