This study is an initial attempt of automatically estimating the musical performance quality of single instrument studio recordings. The primary purpose is to develop an algorithm for best-take detection in the context of Digital Audio Workstations. The problem is narrowed down to monophonic lines of electric guitar and singing voice in popular music. The analysed musical rubrics are rhythm and pitch. In contrast to previous work, the exact melody is not known for the assessment, while a synchronized click track and a backing track can be used as references. Timing and intonation features are derived from tempogram and chromagram representations, as well as from an automatically performed melody transcription. The majority of implemented features uses either quantization cost functions or histogram-based relations. Different machine learning techniques for classification and ranking are applied for the final musical quality prediction.