Natural Language Processing (NLP) is increasingly transforming the legal field, with the potential to revolutionize the processing and analysis of legal decisions. Despite this, there are significant unexplored areas, creating limitations and challenges. Our research reviews the NLP landscape in the legal domain and explores various algorithms for predicting legal outcomes. We propose a novel approach that predicts legal outcomes using fact descriptions alone, without accessing rulings, showcasing the potential of deep learning techniques for accurate forecasting. Our evaluation of Pre-trained Language Models (PLM) for various NLP tasks in the legal domain reveals the superiority of PLM-based methods over baseline systems, with Custom Legal-BERT achieving a predictive accuracy of 94%, outperforming Legal-BERT (82%) and Double-BERT (80%). These findings highlight the high predictive accuracy achievable with advanced methods. By addressing current limitations, we pave the way for more accurate and accessible legal practices. This research is a crucial step toward evolving NLP in the legal field, promising improved efficiency and accessibility in legal processes.