Tennis Player Ranking using Quantitative Models

The Association of Tennis Professionals (ATP) and the Women’s Tennis Association (WTA) generate weekly rankings for professional tennis players by awarding points to each player depending on how far the player has advanced in a countable tournament. Since tournaments are designed such that top players face the lower-ranked players in the earlier rounds, a bias is introduced which favours the top players. In this paper we demonstrate two new algorithms, SortRank and LadderRank, which rank professional tennis players. Both ideas make use of a quantitative tennis model to assess the performance of individual players and then compare them with each other. SortRank uses traditional sorting algorithms to rank the players using the result of a simulated match between the two players as the comparison criterion. LadderRank ranks players using a “sports-ladder” style iterative algorithm, which also compares players based on the result of a simulated match between them. Both algorithms are flexible as they can be implemented using any underlying quantitative model. The ranking systems are demonstrated and assessed based on their ability to predict the outcome of matches played within the period used to rank the players.