In the scope of this paper, it is aimed to model the stochastic deterioration behavior of a fielded unit in a complex military system and develop condition based maintenance policies. In order to analyze a real life problem and determine effective maintenance policies, following methodology is implemented. First, measurements of a deterioration characteristic are collected from the field. Functions representing the increase in deterioration are obtained and transformed into the condition states. According to condition state classifications, matrices consisting transition probabilities between condition states are estimated. Then optimal condition based maintenance policies depending on these condition states and transition probabilities are found for various objectives using Markov models.