The architecture of network games is generally of two main types: Client/Server (C/S) and Peer to Peer (P2P). Distributed network games use the P2P architecture mainly to reduce network latency. This architecture may be based on graph or tree structures. Given an identical amount of data to be transferred, the bandwidth requirement in a tree-based P2P network game is not the same as that of a graph-based game. In a graph-based P2P network game, the bandwidth requirement of each player is a linear function of the number of players. In a tree-based game, by contrast, it is a quadratic function of the number of child nodes. This implies that, due to limited bandwidth, some nodes may suffer from being overwhelmed by the arrival of a large number of packets, leading to a packet-drop. In turn, the packet-drop would trigger packet retransmission; this may result in a repetitive cycle of packet retransmission and packet-drop. Such packet-drop problems may cause severe network latency. In this paper we examine methods to reduce the bandwidth requirements of tree-based P2P network games, with a view to improving system performance.