Unpaved roads may represent more than 80% of a country's road network. Given the socioeconomic importance of unpaved roads to the well-being and development of rural populations, agencies in charge of their management should maintain them in optimum condition. A good management system should consider the use of effective evaluations of road conditions and reliable condition performance models. Available performance models for unpaved roads estimate the progression of one distress type subject to variations of independent variables affecting their performance over time. These variables commonly require detailed evaluations of road materials, geometric design, and traffic, demanding considerable expense and limiting the application of the models to project-level management. The objective of this study is to develop condition performance models for network-level management of unpaved roads on the basis of probabilistic deterioration trends observed in the field. The scope is to design practical models that are applicable to different climatic conditions and various road types and that can be effectively used by agencies in developing countries. The condition of an unpaved road network in Chile was assessed during three evaluation periods by use of the unpaved condition index methodology. Finally, condition performance curves for gravel and earth roads were developed with Markov chains and Monte Carlo simulation by consideration of a 10-year analysis period and three different climates.