Ok, I have been trying to think of a way to develop a fast paced action racing game with lets say upto 20 players. There would actually only be four actual players in the game - each on a different PC (the rest of the vehicles in the game would be AI controlled vehicles). I have been reading about physics being localised on every client and therefore physics reactions being seen incorrectly across the network on other players games. What if... ...because this is a standalone exhibit solution (i.e. just 4 PCs networked to each other), I make one player game control the first five vehicles' physics and it passes the physical responses to the other PCs as transform information via the hard wired network. The important bit is that the actual physics calculations for those five vehicles are done only on that one PC. Based on this idea, each of the other players PC's get to control another five vehicles - independently - each. The resultant information is passed over the network as before and so everyone sees the correct responses and only gets to have to deal with five vehicles maximum. Does that make sense? Am I missing anything in my thinking which would prove a probelm for me at some stage please? What do you think? Might that work in this instance? IPete2.