Backup goalkeepers present an extreme version of the low sample size problem — they may play only 5-10 matches per season (cup games, injury cover), producing too few shots faced to generate statistically reliable save metrics. Yet clubs need to evaluate and develop them. The solution: use training data extensively, supplement with cup/league cup match data, apply heavy Bayesian priors from the player's historical samples, and focus evaluation on shot-stopping technique quality rather than outcome-based metrics.
For backup GK evaluation: (1) track training session data including all shots faced and their outcomes; (2) aggregate across cup appearances and loan data if available; (3) use technique-based assessment from tracking (positioning deviation, body shape on shots) rather than outcomes; (4) apply Bayesian shrinkage toward position-group mean to avoid over-reading 8-save samples; (5) build a development profile showing improvement trajectories across skills rather than current-level ratings.