Determining Baselines for Evaluating a Horse Racing Prediction Process

This paper examines the establishment of a baseline that can be used to determine the value of a model for predicting the outcome of an uncertain event. The event of interest is predicting the three horses that will be one of... [ view full abstract ]

This paper examines the establishment of a baseline that can be used to determine the value of a model for predicting the outcome of an uncertain event. The event of interest is predicting the three horses that will be one of first three horses to finish a race and win a “Show” bet. Two baselines are established. One baseline is based on randomly selecting horses from the field. The other is based on a crowdsourcing prediction using the betting odds just prior to the race. To account for the difficulty of predicting becoming more difficult as the number of horses in the field increases, a table is created for tabulating outcomes for races to determine the value of the model relative to the baseline. The table will have a row of results for each of the number of horses in a race for the data in the evaluation data set. A criterion for determining the value of a model is proposed using this table of results for a particular baseline.