Legal Outcome Prediction (LOP) is the process of predicting the possible outcome of a court case based on the case file. The aim of this work is to assess the performance of four deep learning models namely BERT, LegalBERT, DistilBERT and RoBERTa. These models were trained and tested with complete case files and their summaries. The models were then evaluated using metrics like train-test accuracy, loss plots and F-score. Based on these evaluation metrics, RoBERTa was chosen as the model with the best performance. In order to understand the predictions made by the deep learning models, Local Interpretable Model-Agnostic Explanations (LIME) was used. LIME is an eXplainable AI (XAI) tool that highlights the words in a case file which were used in order to predict the outcome of that case.