Problem solving is an essential skill for students to be successful in life and careers. Students need to use efficient strategies to solve problems effectively. In this basic interpretive qualitative study, we aimed to (a) explore children's problem-solving strategies in a game-based tool (i.e., puzzles), and (b) investigate the troubleshooting strategies they employed while solving the puzzles. We recorded students' puzzle-solving efforts, and using an observation analysis approach, noted important moments, patterns in puzzle-solving, troubleshooting methods, and other noteworthy events. Our analysis showed that while solving computer-based puzzles, students demonstrated the use of three approaches: varying-one-thing-at-a-time (VOTAT), building all-at-once or change-all (CA), and a mixed approach. CA was the approach used most often, followed by VOTAT, and then the mixed approach. Of the two troubleshooting approaches, starting the sequence over was the preferred method. Others opted to search for the faulty tile in the sequence. We discuss how these findings can inform practice and provide some insights as to the usefulness of the game-based tool, Lightbot.