Degree Name

Doctor of Philosophy


School of Electrical, Computer and Telecommunications Engineering


Born only a few decades ago, the video game industry has now surpassed the revenue of the movie business and shows no sign of decline in its expansion rate. In addition to the large number of exclusively online titles operating today, most regular recent game also provide a networked multiplayer mode. Despite their success, the online experience of online games is extremely sensitive to network conditions.

Network games are realtime and highly interactive. For these reasons they are more affected by the imperfections of the telecommunication network than most other Internet applications. In particular, geographical distances introduce unavoidable delays which degrades the experience of players. This objective of this thesis is to models the impact of network imperfections on game playability and propose novel solutions to improve the quality of experience of participants in online games.

There is no consistent analysis across different game research groups on how network delays affect game users. In order to understand the impacts of network delays on game quality, we introduce a generic framework which can analyse the propagation of network disturbances in a game as a three steps process. First, measurable inconsistencies are derived from the game's network topology and synchronisation scheme. Next, we determine the violations of the ideal laws of the virtual environment generated by these inconsistencies. Finally, the perceptual impact of these violations on players is estimated using fuzzy logic utility functions.

Game quality metrics proposed in the literature either suit a particular setup or do not capture every aspects of the players' experience. Using our framework, we introduce a measurable and objective definition of network playability as the collection of all inconsistencies endured by a player. As fairness is a function of the relative playability amongst game users, we define the network unfairness of a game as the variation coefficient of the participants' playability and validate this definition in an experimental setup.

The thesis also investigates the potential to trade-off different aspects of the playability of participants through the alteration of (the) game synchronisation. We demonstrate that absolutely conservative and fully optimistic synchronisation schemes are the extremes of a continuum of possible strategies. Using a specialised game simulator, we search and find the optimum synchronisation parameters within this continuum of trade-offs. Instead of using the same synchronisation for all actions, a second improvement consists of tuning the synchronisation parameters used for different actions independently and according to the specific requirements of each action type.

After studying and proposing enhancements to the synchronisation scheme of online games, this thesis formulates an integer programming problem aiming at optimising the selection of game servers in a cloud of potential sites. We introduce the notion of critical inconsistency and demonstrate that optimising this criteria tends to improve both playability and fairness. Finally, the resolution of small instances of this problem lead to the development of a recursive heuristic capable of converging towards a near-optimum set of servers providing best network conditions for the players of an online game. Simulations shows the gap between our heuristic solutions and a calculable lower bound average 5.19% (2.95 StdDev).