The present paper describes accelerated corrosion test results and the development of a probabilistic model to predict the depth and variability of pitting for pretensioned prestressing wires and strands subject to pitting corrosion. From the accelerated corrosion tests in a chloride-concrete environment the mode of wire failure is studied, and the spatial distribution of maximum pit depth along wires for various lengths is obtained. The analysis shows that the maximum pit-depth distribution of prestressing wires can be represented by the Gumbel extreme value distribution. This model can be extended to predict the distribution of maximum pit depth for any length of wire, corrosion rate and time of exposure in real concrete structures. This information can be used to predict minimum cross-sectional area of a corroded wire, wire capacity and the effect of pitting corrosion on the safety of pretensioned prestressed concrete structures.