Disentangling recurrent end-of-period shooting declines in professional basketball: A spatially adjusted analysis

End-of-period declines in basketball shooting have been described previously, but the extent to which they reflect spatial shot selection, very late attempts, and game context remains unclear. We analysed 208,775 regulation field-goal attempts from 1685 EuroLeague games across five seasons. A binomial generalized additive mixed model estimated shot success as a nonlinear function of time elapsed within each 10-min period, separately for two- and three-point attempts. The model adjusted for two-dimensional shot location, quarter, home-away status, competition phase, pregame opponent win percentage, starter status, standardized pre-shot score differential, season, and player and game random intercepts. Spatial adjustment increased deviance explained from 3.43% to 9.62%. The temporal smooths remained significant for two-point (effective degrees of freedom [edf] = 3.94, p < 0.001) and three-point attempts (edf = 5.89, p < 0.001). From 5.0 to 9.9 min within a period, marginal three-point probability declined by 10.66 percentage points without spatial adjustment and by 7.07 points after spatial adjustment, a 33.7% attenuation. The adjusted decline remained after excluding 726 potential heaves (−6.54 points) and after excluding all attempts in the final five seconds (−3.89 points). Quarter-specific analyses showed a three-point temporal association in every quarter. Random elapsed-time slopes for players and games did not improve fit. The findings identify a recurrent end-of-period association rather than a uniquely late-game effect. Adjustment for shot location and exclusion of final-five-second attempts attenuated, but did not eliminate, the pattern; unmeasured shot-clock, defensive, and lineup factors preclude causal or prescriptive interpretation.