Higher risk does not necessarily produce greater expected value in football passing

Passing performance is often summarized by completion rate, which can obscure differences in pass difficulty, decision value, and execution. We developed an interpretable risk–reward framework combining a calibrated expected pass-completion model with a cross-fitted pass-chain continuation threat model. The framework separates expected success value, expected failure cost, and execution relative to completion probability. The analysis included 3,396,912 open-play passes from 3,926 matches. On a locked independent test set of 390 matches, the Full xPass model outperformed constant and geometry-only comparators (ROC-AUC 0.872; failure-class PR-AUC 0.646) and showed close calibration, with modest overprediction in the highest-risk tail. Across the full sample, expected success value and expected failure cost increased at different rates as risk rose, producing a nonlinear relationship between risk and expected net value. Under-pressure and non-ground passes were associated with less favorable net-value profiles, although these comparisons were descriptive. Three team-season applications showed distinct risk-allocation patterns. Among role-matched players with similar completion rates, modeled difficulty, and passing volume, decision-value profiles differed because expected success value and expected failure cost were combined differently. Results were broadly stable across alternative horizons, xPass specifications, and event filters, but player rankings changed when failure exposure was restricted to opponent-transition threat alone. The framework provides a transparent decomposition of short-horizon decision value rather than a universal risk threshold or overall measure of player quality.