We analyse complex hierarchical forecast events: end-of-season sports league outcomes *conditional* on the point in the season. We construct forecasts and analyse the forecasts of bookmakers. Our forecasts are based on a statistical model used to inpute outcomes, update explanatory variables and iterate to the end of the season. We then repeat the process from each point in the season numerous times to generate frequentist probabilistic forecasts of end-of-season outcomes. Analysing such a model helps understand complicated probability questions such as how much forecasts should be updated after each gameweek of a season, and enables us to shed some light on the two types of forecasts competitive bookmakers make: (1) updating to new information, and (2) responding to price competition from other bookmakers. We find that our forecasting model is competitive in that it provides similar forecasts to bookmakers throughout the season, and that competitive pressures appear more important than informational events for bookmakers.