Sales forecasting is a common problem in economics. Lottery sales are one of the favorite issues of sales forecasting. Sales of lottery tickets depend on many economical issues and such a problem was investigated previously (Beensock et al., 2002) where Genetic Programming is used in order to construct different agent structures that predict the number of ticket sales in Israel lottery. In each application in (Beensock et al., 2002), only a single agent is developed to predict the number of ticket sales at a present drawing. Instead we propose a Local Decision-Making model which performs the sales forecasting job in multiple levels by employing agent structures that operate locally and combine their decisions via a suitable decision fusion technique. It seems that Local Decision-Making in Multiple Levels fits well for the problem. © 2011 IFAC.