Training a single xPass model on combined men's and women's data with an explicit gender indicator feature, which outperforms both gender-blind models (that ignore systematic differences) and gender-segregated models (that have insufficient women's data for robust training). The model learns general passing relationships while capturing systematic differences like steeper pass-length drop-off in women's football and cross-league pressing variation.
Combine men's and women's event data into one training set. Add a binary gender feature. The model learns shared physics/geometry of passing AND gender-specific calibration. Validate by comparing AUC/log-loss against gender-blind and gender-segregated alternatives.