Modern digital games generate extensive gameplay telemetry that can reveal patterns of player behavior and performance. This study proposes an AI-enabled game analytics framework for intelligent player profiling using gameplay data collected from a Unity-based Roll Ball environment. The Roll Ball game is implemented as an experimental platform consisting of three stages of ascending difficulties, such as Tier 1, Tier 2, and Tier 3, that are successively unlocked as the player passes from one stage to the next. While playing, the players interact frequently with the game environment, generating information related to their performances. The empirical analysis is performed by gathering information regarding the gameplay of the undergraduates. The results obtained from the scores of the items obtained during the gameplay are used to measure the performance of the player as they progress through the levels. To this end, two popular machine learning techniques, namely, Random Forest and Feedforward Neural Network (FNN), are adopted to assess the relation between the scores achieved in different levels of the Roll Ball game. The experimental dataset was split into a training set and a test set to examine the effectiveness of our proposed approaches. The experimental results obtained demonstrated the capability of machine learning algorithms in detecting performance-level related patterns from the gameplay data and hence obtaining insights about the player behavior under different levels of game difficulty. This work explores the possibility of employing machine learning with gameplay data collected through gameplay telemetry and aims to demonstrate the benefits of intelligent game analytics, player behavior analysis, and data-driven game design. Possible future works include integrating additional features such as time to complete levels, player movements, and interactions at each level to achieve a more detailed and robust modeling of the players and hence to improve the predictive power of our proposed approaches.