The aim of this study was to build an XGBoost model to predict the match outcome and analyze match-related technical, tactical and physical performance features that may influence the predicted outcome of the match. This is an observational study which follows a retrospective design. The FIFA post-match summary reports were downloaded at the end of the 2023 Women’s World Cup and used to create a dataset which consisted of match-related technical, tactical and physical performance variables. Then, an XGBoost model was built to predict the match outcome and investigate which performance features might influence the predicted outcome of the match. The overall model achieved accuracy of 0.58 ± 0.05. Losses and wins had similar predictive accuracy (0.67 ± 0.06 and 0.67 ± 0.08, respectively), but the prediction of draws performed was significantly worse with accuracy of 0.32 ± 0.16. The top ten features for predicting wins were: (1) out to in actions by the opponent, (2) attempts at the goal, (3) in-behind actions, (4) interceptions by the opponent, (5) loose ball receptions, (6) sprinting per minute by the opponent, (7) offers received by the opponent, (8) in-front opponent, (9) interceptions, and (10) total distance per minute. The top ten features for predicting losses were: (1) attempts at the goal by the opponent, (2) interceptions, (3) out to in actions, (4) possessions interrupted, (5) loose ball receptions by the opponent, (6) in front movements, (7) distance covered by the opponent, (8) in-behind actions by the opponent, (9) total distance, and (10) sprinting per minute. In conclusion, using an XGBoost model, this is the first study to successfully predict the match outcome for wins and losses from the FIFA Women’s World Cup, but also explain which features significantly influence the prediction. This study may serve as a guide for practitioners regarding the use and application of XGBoost models in high performance.