Several techniques exist which reduce network bandwidth consumption and thus limit the effects of network latency in networked virtual environments. Dead reckoning is one such technique that has been widely adopted for use with dynamic entities. However, in previous work we have shown that the use of a spatial threshold alone in dead reckoning can result in unbounded absolute inconsistency. A novel hybrid threshold was proposed that combined a spatial threshold together with an absolute consistency metric to impose an upper bound on absolute inconsistency under all circumstances. This was verified through a simulation of typical movement in a computer racing game. This paper extends this work by investigating the problem in more detail and verifying the hybrid threshold solution in live trials across the Internet. In particular the relationship between curvature of movement and absolute consistency is explored. The experimental trials allow a comparative analysis of how users behave when different threshold types are used under varying degrees of curvature. This work provides further justification for the use of a hybrid threshold approach when dead reckoning is employed in networked virtual environments.