Co01posite Ordinal Forecasting in Horse Racing - An OptiDiization Approach

Using horse racing data in Hong Kong as an example, this paper looks into the properties of an optimization model for making composite ordinal forecasts based on minimization of the absolute error of the joint distribution of the errors of twelve forecasters of race outcomes. It was found that the optimization model is not only sound theoretically, but it is also robust, and can handle situations when data are sparse.