Estimating office real estate prices can be done through both traditional and advanced approaches. The latter attempts to capture the thought processes of market participants and incorporate insights regarding relationships and/or factors that are difficult to observe or quantify through traditional models. However, the specific determinants often remain relatively unexplored, primarily due to data limitations or their poor quality in many countries. Moreover, the influence of these determinants can vary depending on locational differences. Therefore, the primary goal of this paper is to identify the most relevant market and property-related factors for estimating office real estate prices with machine learning, based on the example of the Polish market. Additionally, the secondary research goal is to investigate what makes these factors more relevant. A comprehensive analysis was conducted based on data obtained from a leading market intelligence provider specializing in commercial property analytics, covering a 20-year observation period. The research results indicate that, while there was only one key market-related factor, there were as many as twenty-four property-related variables.