Factors that give enzymes stability, activity, and substrate recognition result from the combination of few weak molecular interactions, which can be difficult to study through rational protein engineering approaches. We used irrational random mutagenesis and in vivo selection to test if a β-glycosidase from the thermoacidophile Saccharolobus solfataricus (Ssβ-gly) could complement an Escherichia coli strain unable to grow on lactose. The triple mutant of Ssβ-gly (S26L, P171L, and A235V) was more active than the wild type at 85 °C, inactivated at this temperature almost 300-fold quicker, and showed a 2-fold higher kcat on galactosides. The three mutations, which were far from the active site, were analyzed to test their effect at the structural level. Improved activity on galactosides was induced by the mutations. The S26L and P171L mutations destabilized the enzyme through the removal of a hydrogen bond and increased flexibility of the peptide backbone, respectively. However, the flexibility added by S26L mutation improved the activity at T > 60 °C. This study shows that random mutagenesis and biological selection allowed the identification of residues that are critical in determining thermal activity, stability, and substrate recognition.