This paper investigates some new techniques for software reliability forecasting and also adaptive methods based on statistical measurements in order to choose dynamically an appropriate algorithm. At present each application has to be constantly associated to only one fault tolerance strategy. Illustrative examples with real data are given.