One of the main problems to test timed systems is that the tester has to decide when to apply the next input to the system under test. Even though the tester could determine good sequences of inputs to find a big variety of errors, the quality of the test suite usually depends on the time when the different parts of the sequences are applied. In this paper we give a formal methodology to provide good time values to test timed systems. These values are computed by taking into account the time stability of the system, that is, if the system is more likely to remain in its current internal state during a given time interval then no input will be applied during that period. In other words, our method will (probabilistically) find those time values that are closer to a change of state in the system, being these values more suitable to apply the appropriate input to the system.