World-wide multiplayer games present several scalability challenges for large-scale deployment. In recent years, significant research has been devoted to scaling in terms of number of players so that the same virtual world can accommodate larger number of simultaneous avatars. A possible solution is to use grid computing techniques for distribution of processing over a cluster or a distributed set of servers. Scalability with respect to geographical distribution of players is another challenge. Large geographical distances between players introduces propagation delays which are impossible to avoid. Without proper design, responsiveness of the application to user inputs is degraded even when there is abundant processing and network resources available to the game. Currently, most subscription based commercial massively multiplayer games, such as Dark Age of Camelot or Lineage II, deploy independent virtual worlds on different continents to reach their clients. Geographical distribution of game servers with proper state synchronisation in overlay networks can help cope with propagation delay. Relevant publications are reported in references ,  and . Game server distribution can be extended in the extreme towards a peer-to-peer situation where each player's machine becomes a server itself. In this case, the same state synchronisation strategies as classic server distribution, which will be discussed here, can be used. However, specific issues to peer-to-peer game architectures   such as limitations in the users' access bandwidth and increased risks of cheating are outside the scope of this work. This article first introduces, through simple examples, the notions of response time and paradoxes along with conservative and optimistic synchronisation techniques using local lag and timewarps. Next, we briefly present the distributed game simulator we developed and the methodology used for the simulations before presenting the two tailoring local lag techniques along with their associated simulations results.