Probabilistic modelling to give advice about rowing split measures to support strategy and pacing in race planning

In this work, we focused on understanding the required performance levels throughout each section of rowing races to finish in certain positions. We conducted our analysis in terms of each 500-m sector of those races for which historical performance data existed. We considered the ranking and time taken in each sector by a given boat as two important predictor factors/attributes that can be taken into account for strategic pacing planning in standard rowing races. We developed a novel hybrid data mining approach based on probabilistic modelling and combinatorial optimization to find the optimal permutation of split measures (times and rankings) that can maximize the chance of a boat winning certain medals in a standard 2000-m rowing race. We further extended our probabilistic model to analyse rowing data from other perspectives. In this research, we considered race type (fast, medium, slow) as well as country profiles. The latter analysis could be used for strategic planning in terms of combating opposing countries’ strategies by understanding their racing patterns.