Accentedness and comprehensibility scales are widely used to evaluate the effectiveness of Computer-Assisted Language Learning (CALL) tools. Such evaluations mainly rely on subjective expert assessment to measure accentedness and comprehensibility. In this study, we applied Automatic Speech Recognition (ASR) measures and used acoustic features to investigate the importance of objective features and measures that may underlie accentedness and comprehensibility. Furthermore, we combined the practice length of CALL tool users with the pre- and post-test data to gain insight into how practice affects progress. The experimental results showed that ASR measures, loudness, and formant-related features were most relevant for accentedness and comprehensibility. Additionally, practice time was positively correlated with progress in accentedness and comprehensibility. We also found that longer practice time contributes more to the changes in the acoustic features selected by the Lasso regression.