Social Capital refers to the tangible or intangible benefits that a person can acquire by the virtue of their position within a social network. Because of difficulties in operationalizing metrics of success and measuring social capital, the use of social capital for predictive modeling has been limited. In this paper we address this deficiency by proposing extensions to the standard metrics for measuring social capital and applying it to a domain where the success of agents in social networks can be measured more concretely. The results show that social capital based approaches give better results as compared to more traditional approaches which use previous performance or player similarity measures for performance.