Preprint. A Bayesian hierarchical model separates water-polo scoring into even-strength, man-up and penalty components while allowing team abilities to evolve between seasons. Using 940 Italian Serie A1 matches, the study examines when relegation can be predicted and compares ways to initialize newly promoted teams. The model identifies the directly relegated club earlier than the league table in the analysed data, while the play-out remains much harder to predict. Accounting for promotion improves forecasts, but distinguishing how a team was promoted adds little.