Playing for Quality: A Competitive Educational Game for Teaching Unit Testing in Computer Science Education

Abstract

Software testing is a foundational competence in computer-science curricula and a notoriously disengaging one for learners; a recurring concern for the games-based learning (GBL) community is whether gamified-testing interventions actually deepen learning or merely raise activity levels. This paper presents the design and a proof-of-concept evaluation of Test Royale, an educational game for teaching unit testing whose mechanics are deliberately bound to learning objectives at Bloom’s apply, analyse and evaluate levels: live mutation analysis (Stryker.NET) and coverage analysis (Coverlet) are surfaced as in-game feedback, and the score function is computed directly from validated quality indicators rather than from activity volume. The design is articulated through the Mechanics–Dynamics–Aesthetics (MDA) framework and grounded in the Octalysis core drives. We report a small mixed-methods pilot study with two complementary participant groups: sixteen undergraduate students in a Software Testing course (8 in a gamified condition, 8 in a non-gamified condition; Study 1) and eleven professional developers at an Egyptian fintech company (Study 2). In Study 1, the gamified condition outperformed the non-gamified condition on every measured quality indicator with large-effect-size differences (mutation score Δ = +8.2, d ≈ 0.81; branch coverage Δ = +10.7, d ≈ 0.96; line coverage Δ = +7.5, d ≈ 0.84; useful test LOC Δ = +12.1, d ≈ 1.23), with p < .05 for each and the pattern surviving Holm–Bonferroni correction across the five outcomes, while execution time did not differ significantly. In Study 2, the system achieved a System Usability Scale score of 80.0 (“excellent”; 95% CI [66.88, 93.12], n = 11), with self-reported motivation among already-motivated practitioners remaining high but not significantly increased. Given the small samples, the entanglement of game mechanics with feedback immediacy in our control condition, and the single-session design, we present these results as encouraging proof-of-concept evidence rather than confirmatory findings, and we describe a three-arm replication plan that disentangles game mechanics from feedback design.