Forecasting judo medal winners at the Olympic Games: an interaction of the International Judo Federation World Ranking List and the Elo System

Abstract

The aim of this study was to verify which ranking list predicts Olympic results better: one created by Elo system, one using the International Judo Federation (IJF) World Ranking List (WRL), or another using the combination of both. The data utilized comprised the outcomes of 93,728 matches, encompassing 42,844 matches from the 2016 Rio Olympic Games cycle and 50,884 matches from the 2020 Tokyo Olympic Games cycle. These matches were held across 311 events, all of which contribute points to the IJF WRL. The data was sourced from https://judobase.ijf.org. A total of 8,142 male and 4,736 female judo athletes from all weight categories were analyzed. We employed two variables as proxies for athletes’ performance throughout the Olympic cycle: the positions in the IJF WRL and the ratings from the Elo System. A binary-response model was utilized. In this model, “success” denoted an athlete receiving a medal, while “failure” indicated otherwise. A combination of the WRL and Elo system better predicted Olympic performance of judo athletes. Additionally, for each rank position an athlete improved in the IJF WRL, there was an increased probability to win an Olympic medal of approximately 7.50%, while for each 10 Elo rating score improvement, the athlete increased the probability to win an Olympic medal in approximately 9.26%. When both systems were used together, the accuracy of the model was approximately 91%, with a sensitivity of nearly 68-69%, and a specificity close to 95%, for Rio de Janeiro and Tokyo editions isolated or grouped. Such information can serve as a valuable tool for national federations staff in selecting the most suitable athletes to participate in the Olympic judo competition, if both the WRL and an Elo rating system are used together.