A key challenge for ecologists is to predict how complex ecosystems will respond to environmental impacts such as species extinction. Markov chain models are a relatively empirically-accessible framework that models transitions among ecological states for defined study objects such as points in space. These models have largely been applied to and tested with assemblages of sessile species open to outside immigration. An unanswered question is whether they can be successfully applied to closed systems with stronger feedbacks, and to a broader suite of ecosystem components, particularly mobile species. Using zooplankton sample data from the experimental lake studies of Carpenter, Kitchell and colleagues, I applied cluster analysis based on Daphnia abundance to define ecological states within the lakes and constructed a model of transitions among these states under low plankitivory conditions. I then used this model to predict how the zooplankton community would change under low piscivory/high planktivory conditions by deleting states with high Daphnia abundance and proportionally adjusting the remaining transitions to meet structural requirements of the model. These predictions were tested against independent data for Tuesday Lake under high planktivory/low piscivory conditions. The model provided significant predictive information (P < 0.00001), capturing 85% of the variation in log-abundance and 75% of the change in log-abundance of different zooplankton functional groups under high planktivory. The results demonstrate that Markov chain models can be usefully applied to relatively closed systems and to communities of mobile species.