In recent years, researchers have begun paying more attention to the use of augmented reality in education. Augmented reality can make learning processes more engaging and entertaining by superimposing virtual images over the real world. In this paper, we propose the use of augmented reality as a new way to learn how to play Go. Go is a popular board game that is difficult for beginners to learn because of its highly complicated strategies and patterns. Teaching these strategies in an augmented environment is more intuitive and fun than conventional ways of learning through textbooks or computer software on a monitor screen. It can also reduce the gap between learning theories and applying book knowledge in a game. To illustrate this, we introduce ARGoTrainer, a marker-based augmented-reality system that acts like a Go tutor and teaches users the strategies of Go. The system allows users to play against human players as well as artificial opponents (the computer Go engine) without losing the physical interaction of the board game. While playing games, the system provides context-driven and real-time assistants to teach users strategies like fuseki (opening moves), joseki (common corner patterns), and territorial analysis. Those strategic suggestions are generated based on the state of the game and are recompiled in real time after each new move. These personalized learning assistants can reduce the time it takes to adapt and apply theories in actual games. A preliminary user study revealed that this new way of learning Go is entertaining and the system could be more effective than learning only from books.