Probabilistic methods are applied to the study of fatigue wear of sliding surfaces. A variance of time to failure (to occurrence of maximum allowable wear depth) is evaluated as a function of a mean wear rate of normal wear and a size of wear particles. A method of estimating probability of failure-free work during a certain time interval (reliability) is presented. An effect of the bedding-in phase of wear on the reliability is taken into account. Experimental data for Ultra High Molecular Weight Polyethylene (UHMWPE) cups of artificial hip implants is used to make numerical calculations.