This study analyses the impact of learning modes, frequency, and (mis)match of preferred learning strategies on two cohorts of Computer Science students. Using VARK learning strategies, Moodle access data, and grades across learning environments, our finding shows students perform best in offline settings (43.79% achieving first-class grades versus 33.64% in blended and 29.25% in online). Frequent Moodle access correlates with higher performance across all modes. Surprisingly, students with mismatched learning strategies (over 70%) outperformed those with matched strategies. This unexpected result likely stems from the disconnect between lecture-based teaching methods and CS assessments. While lectures align with visual and auditory styles, CS coursework demands kinaesthetic approaches through programming and problem-solving, suggesting adaptability may benefit CS students more than strict adherence to preferred learning styles.