A method of training a test system, including: running a predefined set of test procedures on the testing device in order to obtain a temporal test metric; evaluating the temporal test metric via a machine learning model by a processing circuit; and training the machine learning model by the processing circuit to predict a temporal course of a test stability score based on the temporal test metric obtained, wherein the training of the machine learning model is based on the temporal test metric together with a known binary test result, and wherein the temporal course of the test stability score indicates the probability of a respective binary test result throughout the entire duration of a test. A test system and a method for mobile network testing are also described.