Networked games are an important class of distributed systems. In order for such applications to be successful, it is important that a sufficient level of consistency is maintained. To achieve this, a high level of network traffic is often required. However, this can cause an increase in network latency due to overloaded network hardware, which, ironically, can have a negative impact on consistency. Entity state prediction techniques aim to combat this effect by reducing network traffic. Although much work has focused on developing predictive schemes, there has been little work to date on the analysis of their true impact on the consistency of the system overall. In this paper, we identify an important performance-related characteristic of packet reduction schemes. It is demonstrated that there exists an optimal packet transmission region. Increasing or decreasing network traffic above or below this level negatively impacts on consistency. Based on this characteristic, it is proposed that predictive schemes exploit this optimal point in order to maximise consistency by efficiently utilising the available resources.