 This paper proposes a new approach to evaluating the accuracy of expected goals, XG, models. It uses machine learning techniques to incorporate additional features into existing models, such as player and team abilities, and compares them to traditional statistics. The results show that the new model outperforms the traditional one and even surpasses the accuracy of an industry leader in the same field. Additionally, it demonstrates that certain features are more influential than others, suggesting that they should be included in future models. This article was authored by James Mead, Anthony O'Hare, and Paul McMenamy.