No-gi Brazilian jiu-jitsu: A Markovian analysis of elite-level combat dynamics

Abstract

No-gi Brazilian jiu-jitsu is a grappling-based combat sport performed without the jiu-jitsu uniform. It’s fast-growing popularity has been influenced by the contributions to competitors’ performance in mixed martial arts tournaments. This study aimed to analyze the frequency of actions, the transitions between them, and their reward–risk balance—the probability of submitting minus the probability of being submitted—among top-level competitors. Data encompassed all 93 matches (90 competitors) of the World Submission Fighting Championship-2019. Reliability was assessed by two experts (sports scientists, more than 10 years as no-gi Brazilian jiu-jitsu competitors). Data analyses used Bayesian methods for inference. Posterior distributions of the mean frequencies of actions indicate that the main actions (e.g. takedown, sweep, and back-take) tended to occur less than once per competitor in each match, except by the submission attempt (point estimate of 1.03 per match). The highest estimates for the transition probabilities between match actions were within-competitor—a guard pass action to another guard pass action (0.30); between-competitors—a takedown attempt to a submission attempt (0.15). Most of the actions presented positive reward-risk balance, particularly for the back-take, which showed the highest probability of direct transition to submission (0.45). The dynamic analysis of no-gi Brazilian jiu-jitsu matches evidenced some preferential actions’ transitions associated with winning. Findings should enhance the accuracy of performance prediction and provide evidence-based recommendations for coaches.