In recent years, artificial intelligence algorithms have gained popularity for analysing sports data from GPS tracking and sensing devices. This data provides trainers with valuable insights to enhance their physical training routines. However, it is challenging for fitness coaches to evaluate football players’ fitness levels and select training regimens that will help them become more fit due to a variety of factors, such as a lack of financing, a lack of data, and challenges with data interpretation and analysis. A hybrid approach (MLP-XGB) was developed to overcome these difficulties, drawing inspiration from the Multilayer Perceptron (MLP) in conjunction with the XGBoost model. The objective is to determine the weakest physical attribute on the squad and forecast the fitness levels of Al-Zawraa Football Club players during the second stage of the Iraqi Stars League. The suggested algorithm’s performance was contrasted with that of other methods, including Gate Recurrent Unit (GRU) and MLP. The study made use of actual data collected during games using GPS trackers. With a recall of 97.26%, accuracy of 97.14%, and precision of 97.26%, the proposed model yielded exceptional results. Through the analysis of sizable datasets and suggestions, this research facilitates the process of determining the team’s degree of fitness.