Voters’ expectations of party strengths are a central part of many foundational political science theories that posit a strategic act by the voter. But how do voters develop these beliefs and how is this belief formation affected by polling reports? In this article, we present a dynamic Bayesian learning model that serves as a baseline for how beliefs are formed. This also allows us to infer how and when belief formation deviates from the theoretical ideal. We validate the model and illustrate its potential based on a number of experiments conducted on MTurk. We find encouraging results in this pilot study that validate the baseline model.