This study aimed to analyse the influence of different physical fitness levels of youth basketball players on match-related physical performance, using Random Forest clustering to distinguish between high-fitness level players and low-fitness level players. Twenty male youth basketball players completed the following physical performance tests in two separate sessions: bilateral and unilateral countermovement jumps, bilateral and unilateral horizontal jumps, single leg lateral jumps, the 20 m linear straight sprint test, the 505 test and a repeated sprint ability test. 1 week after the second testing day, players completed a simulated match while external loads were monitored using an ultra-wide band-based Local Positioning System. A Random Forest clustering was used to create two different clusters composed of players with similar physical fitness attributes (high- and low-fitness level players). Results indicate that the Random Forest clustering adequately discriminated among the players in different groups according to their physical fitness attributes. High-fitness level players covered more distance per min in all intensity thresholds and reached higher maximal speed and acceleration intensity during the simulated matches ( p < 0.05). These results may assist basketball practitioners in understanding running performance variations during matches and can be used to optimise preparation for individual players.