A number of different methods exist to assess clinical stability, a key component of pneumonia management. We compared the prognostic value of different stability criteria through a secondary analysis of the Edinburgh pneumonia study database. We studied four clinical stability criteria (Halm's criteria, the ATS criteria, CURB and 50% or more decrease in C-reactive protein from baseline). Outcomes included 30-day mortality, need for mechanical ventilation or vasopressor support (MV/VS), development of a complicated pneumonia, and a combined outcome of the above. A total of 1079 patients (49.8% male), with a median age of 68 years (IQR 53–80), were included. Ninety-three patients (8.6%) died by day 30, 91 patients (8.4%) required MV/VS and 99 patients (9.2%) developed a complicated pneumonia. Patients with increasing severity of pneumonia on admission, assessed by both CURB-65 and PSI, took a progressively longer time to achieve clinical stability assessed by any method (p < 0.001 for all criteria). Halm's criteria had the highest area under the curve (AUC) for prediction of 30-day mortality (AUC 0.95 (0.94–0.96)), need for MV/VS (AUC 0.96 (0.95–0.97)) and combined adverse outcome (AUC 0.96 (0.95–0.97)). C-reactive protein had the highest area under the curve for complicated pneumonia (AUC 0.96 (0.95–0.97)). Adding C-reactive protein to Halm's criteria increased the area under the curve, but the difference was only statistically significant for complicated pneumonia. All of the criteria performed well in predicting adverse outcomes in patients with pneumonia. Halm's criteria performed best when identifying patients at low risk of complications.