A brief history of the modelling of pipe bursts (leaks) in water supply networks using statistical techniques including discussion of why it is necessary and the benefits arising both strategic and tactical.   From early simple regressions to more recent applications such as proportional hazards model and the "Bayesian Tuner" concept currently used by several of the water companies.  The various predictive parameters include material, size and age together with less obvious influences such as soil type and urban density.  Various analytical approaches are discussed including both panel and snapshot methods.  An innovative method of assessing the quality of models is presented ("effectivity") and the importance of the age predictor is discussed for network management and budget allocation.  Time series approaches are also discussed.