Déjà vu — Predicting the number of players in online games through normalization of historical data

A key factor to delivering a good online gaming experience is to have sufficient server resources relative to the number of players online. In this work, we present a simple profiling technique which allows effective prediction of the number of players in the next 24 hour cycle based on both weekend and workday data. The methodology enables planning ahead so that resources can be scaled to the sufficient amount, thereby reducing cost and power consumption.