Enhancing NeuroEvolution-Based Game Testing: A Branch Coverage Approach for Scratch Programs

Automated test generation for game-like programs presents unique challenges due to their non-deterministic behavior and complex control structures. The NEATEST framework has been used for automated testing in Scratch games, employing neuroevolution-based test generation optimized for statement coverage. However, statement coverage alone is often insufficient for fault detection, as it does not guarantee execution of all logical branches. This paper introduces a branch coverage-based fitness function to enhance test effectiveness in automated game testing. We extend NEATEST by integrating a branch fitness function that prioritizes control-dependent branches, guiding the neuroevolution process to maximize branch exploration. To evaluate the effectiveness of this approach, empirical experiments were conducted on 25 Scratch games, comparing Neatest with Statement Coverage (NSC) against Neatest with Branch Coverage (NBC). A mutation analysis was also performed to assess the fault detection capabilities of both techniques. The results demonstrate that NBC achieves higher branch coverage than NSC in 13 out of 25 games, particularly in programs with complex conditional structures. Moreover, NBC achieves a lower false positive rate in mutation testing, making it a more reliable approach for identifying faulty behavior in game programs. These findings confirm that branch coverage-based test generation improves test coverage and fault detection in Scratch programs.