Late-match card-event rates and score-state associations in Barcelona La Liga matches featuring Lionel Messi: a retrospective analysis of StatsBomb open event data

Objective We examined whether card-event rates were higher in the final 15 min of regulation time and second-half stoppage time than in earlier periods (H1), and whether home-trailing status at the start of a time window was associated with subsequent home- and away-team card-event rates (H2). A secondary, exploratory analysis compared temporal associations before and from the 2018/19 season onward (H3), using an era proxy rather than actual video assistant referee (VAR) exposure. Methods This retrospective observational study included 519 FC Barcelona La Liga matches from 2004/05 to 2020/21 in StatsBomb Open Data in which Lionel Messi had a recorded playing position and the required files were available. H1 and H3 used three periods (0–75 min, 75–90 min, and second-half stoppage time), whereas H2 used six 15-min regulation-time windows and a second-half stoppage-time window. Outcomes comprised Yellow Card, Second Yellow, and Red Card events. Poisson regression used exposure-time offsets and match-clustered robust standard errors; the primary H1 model adjusted for recorded foul intensity and other covariates. Results Across 3,114 team-by-period observations, adjusted total card-event rates were higher at 75–90 min (incidence rate ratio [IRR] = 1.487, 95% CI 1.353–1.634) and in second-half stoppage time (IRR = 1.766, 95% CI 1.480–2.108), both p < 0.001. For Foul Committed-linked cards, the corresponding IRRs were 1.324 (95% CI 1.195–1.467) and 0.967 (95% CI 0.776–1.203). The Bad Behaviour share increased descriptively from 16.11% to 43.09%. Home-trailing status at window onset was not significantly associated with subsequent away-team card-event rates (IRR = 0.981, 95% CI 0.858–1.121). Era-by-period interactions provided no clear evidence of change. Conclusion Late-match total card-event rates were higher conditional on recorded foul intensity, and the stoppage-time card composition shifted toward Bad Behaviour. The analyses did not formally establish different temporal associations by event source. These patterns do not demonstrate decision inaccuracy or referee bias.